@@GaryExplains what is RISC-V missing in comparison to ARMS RISC? Except developer support of course, but as you said it's not like that is much of a difference nor set in stone
@@GaryExplains Would you EXPLAIN? I just like open source and I really hope that it won't take too long until we see the first RISC-V consumer chips...
Yeah, I am planning a video on RISC-V... The big thing is that the ISA is open source, but that doesn't give you very much. Then even if some company kindly open sources a design, then what? You going to make a contract with TSMC to make a chip.
It isn't about one ISA versus another. You could say what is ARM missing compared to x86, or what is MIPS missing compared to ARM. This isn't the important question. Open source "software" is great, I can grab a compiler and go. But open source "hardware", especially for a chip, is nonsense. Do you have a contract with TSMC to make 7nm chips, I don't! Plus it is only the ISA that is open source, not the actual chip designs.
Funny I develop (PHP + MySQL) on Raspberry PI and deploy to x86 instead, mainly because it’s small and cheap, plus the low performance made it easier to identify potential performance problems before deploying to the live server...
They tried ARM based PCs multiple times over the past 10 years. There are a number running windows 10 ARM, but those laptops (e.g., the Asus ones) are overpriced for the hardware. For example a $1000 laptop will have a Snapdragon 835, 4GB of RAM and 64GB of storage, with steep price premiums for more RAM and storage that can't be upgraded. On the X86 side of things, you will be getting a current gen core i5 CPU, 8-16GB of RAM, 256GB SSD, and a mid range dedicated GPU. The issue is really one of companies making ARM PCs, are expecting to get the Apple 450%+ profit margins, while selling an ARM system that benchmarks lower than a $300 entry level x86 PC.
Wanted to add that if you go through all of that expense, you ultimately end up with a system that will not run all of your desktop productivity applications (especially 64 bit stuff), performance will be lower, thus impacting multitasking. There will also be a lack of standardization thus it is not easy to change operating systems, and you will be largely reliant on the device maker to maintain support (which may at most last 18 months). If you need something basic where a SD835 is enough, then you can spend a tiny fraction of the price to get a upper mid range laptop from 2013, and still end up with better performance.
@@GaryExplains The overall idea of what I was trying to bring is that we have to think beyond the direct links and look at the indirect links. You alluded to it in the video where you stated that you tend to develop for the hardware you are using, but developers are not some mythical creatures, they exist in the real world and buy hardware from the same places in general that those of us who do not code for a living, will buy. For an ARM system to become mainstream, it has to be built in a way that appeals to a wide market, otherwise it will not be worth the warehouse space at a place like newegg and it will not be worth the time for companies to put a ton of R&D into (unless that R&D stands a good chance to shift the market over to ARM). The current ARM PCs don't come even close to that and are relatively niche, and without market share, core issue in your video will be stuck in that catch 22 situation. The system you want is not being developed because the low end versions of what you described, are massively overpriced such that no one wants to buy it, and because no one wants to buy it, the companies don't want to invest in R&D for something that is not generating an ROI. The solution you want, requires ARM to reach a point where the choice between ARM and x86 becomes like a choice between Intel and AMD when building a PC. Simply put, a system designed purely as a high perofrmance dev box, will not be popular or in the least bit cost effective, people would rather emulate an ARM environment on an x86 platform for basic debugging. The billion dollar investments into ARM CPUs are pretty much always for the mobile market, thus you will not get something that will rival an Intel Core i7, unless they can get a market where such a CPU , would sell as much as a Core i7 since unlike other markets,CPUs have a massive margin over BOM due to the fact that they have a very limited time window to sell before the tech becomes old and something new is needed in the market. If Qualcomm were to dump 1-2 billion in developing a high end desktop ARM CPU for use in a dev box, it is unlikely that enough software developers would purchase it every year to sustain that business, unless you are okay with spending $100,000+ on a PC that will perform like a $800 x86 PC, and spending that money every year or 2.
I once developed some software in C++ on my x86 computer, but then was deploying it to an ARM based Raspberry Pi. Cross-compiling was interesting for sure, I ended up giving up on compiling it with my computer and had to compile it on another Raspberry Pi. I ran into a few bugs that occurred only on the Pi and not on my development computer, mostly due to data types and the 32bit vs 64bit difference.
@Richard Vaughn The whole point of the platform was to encourage kids to gain interest and maybe become developers. The fact businesses buy and rely on them is a side benefit the people who make them. It's useful for very low end compute-non intensive apps. I would use it for a security layer to separate connections with the outside from the inside where all the heavy computing is done. It can be too slow for even that, so only until I can afford a more powerful ARM or RISC-V based SBC, then a cluster of them.
Shame the whole ARM ecosystem is a horrific mess. Just getting a recent kernel running on most ARM machines is frustrating. Until ARM enforce ACPI etc for everything and open up their GPU drivers properly it's never going to happen.
correct! It just takes a platform leader (Apple, Msft) to build a cash-cow and guard it. When cloud giants (AWS, Google) develop their individual custom chip no 3rd party gets in on it besides the resulting SAS it delivers. Thus war is all about IP (Intellectual Property)... Free Linux broke UNIX licensing but has not killed proprietary x86 business because it's been fragmented.
This will not happen. And it's a great way for vendor lockin on hardware producer side. They haven't learned that they will fail with this attitude. ARM has no server future. Not against AMDs new technology.
I said this years ago. Currently, there's few experimental ARM-PCs out there in the R&D Labs, but nothing for enterprise or consumers. I made the argument that during 2015, there should have been a collaboration to do this with 45W CPU based on 4GHz 8-core ARM Cortex A72 on 14nm, with 64-bit instruction set. Ofcourse, it would've required collaboration with someone like ASUS to make a Motherboard where you can plop in an ARM CPU. And hopefully support some sort of standard RAM support, and PCIe slots for dedicated graphics cards. They could've later made CPU improvements like stepping up to 10nm, 7nm, and 5nm later on. And microarchitecture improvements to A73, A75, A76, A77, and A78. Even throw in a budget options, with a range varying from A38 and A58 CPUs, in 4-core, 8-core, and 16-core configurations, running anywhere from 2GHz to 6GHz at TDPs from 5W to 90W. Have it running Native Linux Distro such as an Official Debian build, a properly optimised AndroidOS from AOSP/Google, and even a Windows 10S... things will take off. Before you know it, native applications are built on ARM PC's with software for ARM Devices, and developers jumping away from x86 systems. Next thing you know there's actual ARM Tablets, Ultrabooks, and Gaming Laptops out there.
With how ARM does it's licensing, I rather not have ARM becoming mainstream on Desktop/Server tbh. We don't need another Intel That's the main reason why there's such a huge push,supported by Google and other behemoth companies, to develop RISC-V. And the potential there is even greater than ARM tbh.
Then we could maybe have PowerPC and then SPARC, and so forth. Having accessible ARM PCs would greatly increase the potential of more processor architectures competing in the market, plus it would massively impact Microsoft's dominance because their trump card of having all that software library disappears when people are now using a whole new architecture, more so considering that there is already a whole library of linux software for a number of other architectures thanks to Debian
@@jackmcslay SPARC, nice as it looks, is now just a Fujitsu thing, and it likely does not have a lot of life left. It came from Sun and Oracle decided there is not going to be any more done for SPARC.
But anyone can produce ARM CPUs. Yes, the license fee is the part that needs competition, but ARM is very cognisant that it doesn't want to be seen as an Intel, exploiting it's market for all its worth for short term gain. It's fees are pennies per CPU, even though it has very little in the way if competition in smartphones, and the profits are large for some of them. It makes very little in the way of gratuitous profit, most of this cost goes to paying engineers to make the best chips of tomorrow, not to line shareholders pockets. Intel in the other hand has extortionate profit per CPU sold. RISC-V will always be nipping on ARMs heels and make sure they don't abuse their market position, and its clear ARM will allow any company to make reasonably priced, excellent performing CPUs for cost price of silicon. I see no reason to think it won't keep doing that. It only operates in a tiny part of the device technology stack, and is more replaceable as a result, unlike Intel that does fabrication and total arch control. ARM has many partners all competing with each other to implement their designs, many fabs competing to fab those designs. It is no more than a licenser of the minimum logic needed to get a CPU to function and function with the current ecosystem. Its ability to be extortionate like Intel is is much smaller, and is nothing like Intel does, even now when it has smartphone dominance. You've probably paid no more than a few dollars to ARM in your lifetime of owning many many ARM CPUs, but probably hundreds to Intel because they can abuse their position much more easily.
when we're done falling in love with ARM, reality will set in. As long as we can keep spending top box for new gear, the industry will come up with new reasons to switch. Thin-clients, client-servers, SAS feeds big-data for AI... pretty soon you won't need to remember your name, just your ID. LOL
That would need a business case; PC makers need a demand from a big portion of the market to flip to ARM, i honestly believe the "long battery life" laptops based on something like snapdragon socs might be killer application that would popularize ARM PCs adoption.
@@ashlord8190 Microsoft built a Win 10 build and an entire development tool set for ARM. As soon as someone builds and optimizes a modern web browser for ARM on Win 10, besides Edge, it can take off. I saw a side by side on an ARM netbook vs one similar priced with an Intel CPU. The ARM was faster, except when running programs that needed x86 emulation. Edge loaded everything faster on the ARM; it was built for the platform. Chrome was slower; it needed emulation. ARM beat Intel by a large margin for battery life. Chrome OS based netbooks didn't beat everything else because of limited local storage and almost entirely depended on cloud compute and storage. Put enough RAM and storage in one with an OS that enough people are comfortable developing for, then you have a winner.
Apple has finally MADE it a thing this decade It's about time. They were tired of people making Hackintoshes. But we will get there eventually. They just need the dopamine rush of being alone for a while. However, Louis Rossman will break it and send videos on repair accross the world (youtube)
@Dex4Sure "Soon it will be NVIDIA as NVIDIA is in talks to buy ARM!" There is no-way that is getting past the anti-trust officers at the justice department. Oh, wait. I just remembered. Government is corrupt.
@@jefflittle8913 hahahaha Obama Government certainly had its fingerprint following the mottos of the past bad ones along with the everlasting Alphabet (Google) who has been changing the laws for themselves and changing the names of every single part of their hive mind to avoid the same laws XD But Nooooo they have a reckoning to pay and they will pay indeed! Ever wonder why they have been giving out "Free Google Homes" for contests that are dead simple to win? They want to know every single secret of your life. Your phone, computer, everyone else's around you, you are a pawn in their hands. However if you play the game right, your pawn will reach the other side of the board and will immediately become a queen. You can rule as well. nVidia should own arm so that Apple does not have the ability to rule the computer world. I honestly don't care.
@@jefflittle8913 Just sell ARM though a shell company in South Africa or something. You can distribute software anywhere in the world, look at Ubutnu for example, it's run out of South Africa. Governments need to realize that in this modern internet age if they become too annoying for digital companies to deal with, those companies can just leave & then government has absolutely no control, so they better not push their luck. Signal for example said if the US tried to take away the legal protection companies have from the activity of their users that they would leave the US & just operate in the British Vrigin Islands where privacy is a more respected right. If they did that then now the US would have absolutely no control over Signal, so you see how that could be a problem for modern governments if they decide to start enforcing anti-trust laws left & right. Companies will leave, & they wont lose a cent of profit doing so, if anything they'll save money being in countries with lower tax rates. Now the US has no control, & they just lost a bunch of taxable income. This is one of the few good things globalism has created as countries now need to be competitive with each other.
We do have ARM based PCs... you can buy them as laptops usually, but desktop ones are out there... They exist. I had one... problem is they are trash compared to x86 in performance. this video is also a 30 second loop of the same thing over and over, why?
@@karlpj1 you gotta pay attention, let me explain. The video is explaining, how does ARM cut into the market? Small businesses go from regular dev pcs to Intel servers because of the easy transition, despite the outrageous multi thousand dollar cost of server cpus. ARM has to drastically cut the cost, from the thousands of dollars, to hundreds, to convince business to make the jump from regular x86 development to ARM servers. Follow the narrative, your eyes and ears are your most valuable organs.
Yep. And if your a developer you can apply for the $500 ADK it’s got 16GB Ram, 500GB HD and all those ports on the back. Or you can just wait a couple of months till the release of an ARM laptop from Apple and use that.
You're missing the point. One of the main reason ARM was made was for mobile platform so that it would fit as many components as possible in the tiniest space possible. It won't be "slot" but more like SMD part (soldered). Sata is also not a slot, it's an interface. There won't be pcie too, it would be just some connectors like raspberry. Way slower, but more versatile in terms of what you can do with it.
@@ragilmalik That might be what it's used for, but that doesn't it's the only possible usage for it. ARM could be a competitor for X86 & X64 but not unless it's put on boards that allow for expansion and customization.
Right, I've been developing for ARM from x86 PC. To test, I have to upload the resulting cross compiled executable to the ARM machine, sometimes far away remotely and it takes significant amount of time, while the resulting executable doesn't always run, let alone run right. If only I can test it locally, things would be much more efficient.
And then try to do fucking complicated multithreading shit like lock free algorithms. You really need local powerfull testing equipment for this. You can't test MT programs on a single socket system. Write reording is a magic that few programmers worldwide fully understand (i do not). So Linus is right, unless there are powerful desktop machines for developers the ARM will continue to fail in the server world. And with all attempts from the past, the ARM servers were disgustingly bad and overpriced - no surprise they all went out of business or even bankrupt.
@Richard Vaughn the sad irony is that not all programs to create ARM programs are runnable on ARM itself, some are really just made with cross compilation in mind.
I feel like ARM corporation (SoftBank?) lost a golden opportunity back in 2016. Imagine at the end of 2015, they released a motherboard that you can hot-swap in your own ARM CPU. And it would be a modern SFF motherboard that will have two DIMM slots, PCIe connection, SATA, DisplayPort, USB, Wifi, Bluetooth, etc etc. And say the first option was for a 14nm 3GHz Quadcore Cortex A72 (64bit), with 1GHz Mali-G71-MP4. It sure would be competitive for the period. And yes, it would need Active Cooling (probably much weaker than Intel/AMD desktop systems though). And it would have an internal GPU that would be more powerful than Intel/AMD. And the best part would be, people could plug in their RAM sticks, SSD drives, and dedicated GPU cards. Imagine a GTX 980Ti working with an ARM System. Sure it would be weird, but also cool. SteamOS could be a boon. And before the year ends and that Intel or AMD has a chance to respond, you just double the performance. You're now toting a 3GHz Octacore Cortex A72 CPU, with 1GHz Mali-G71-MP8 iGPU. Next year you introduce the Cortex A73 and Mali G72, but its a little slower but more efficient, so you boost the clocks. Now you're rocking 4GHz Octacore A73 CPU, with a 500MHz Mali-G72-MP16 iGPU. And you plug in your GTX 1080. And they've yet to release the AMD Ryzen to the public. Fast-forward another year, and now you are using 10nm Octacore Cortex A73 at 5GHz, matching the best of Intel and AMD.... but cheaper and more efficiently. Your iGPU now rocks a 1GHz Mali-G76-MP16 iGPU, which bests any APU's out there. And can even install Windows10-ARM Operating System, alongside native Android, SteamOS, and a plethora of Linux distros (not to mention Server-aimed OS). People have managed to hack these into very smart cryptomining rigs, so there's a lot more excitement. Now, along official ARM CPU's you have them joining you from Qualcomm, Samsung, Huawei and RockChip... all offering hardware on the same socket/motherboard support. Not to mention the likes of ASUS, Gigabyte, MSI and Biostar offering motherboards. You're basically building your own ARM-Server in your home. And things are looking bad for AMD, and even worse for Intel. The fast-forward another year, beginning of last year. You've refined things further. Now the flagship offering is a +10nm 5.1GHz Octacore Cortex A75, combined with a 1.1GHz Mali-G76-MP32. There's also a low-end/budget model using 16nm 2GHz Octacore Cortex A55, with 0.9GHz Mali-G76-MP4 iGPU. People are developing Apps right at home, deploying it to their ARM-Laptops, ARM-tablets, and the usual ARM-based phones and testing it out. They're also developing program for ARM-based servers. Many enthusiasts are sticking their GTX 1080 Ti's and Vega64's, and running games from steam on their linux system using the new SteamOS/Library. All indie developers are now targeting the ARM Desktops to launch their games. A couple big AAA titles are announced for the platform. So what about the year following, in 2019 ? Well, the obvious choice is for 7nm 16-core Cortex A76 CPU running at the usual 5.1GHz frequency. And this time sporting a 1.1GHz Mali-G76-MP64. Overall its enough to challenge the (base) Xbox One in terms of graphics prowess, and challenge the best Xeon/Threadripper in terms of (some) CPU prowess. There's vendors releasing these overpowered ARM-systems, also using dedicated graphics like a GTX 1660 Ti, as standalone TV Boxes that also come with good controllers to do some Steam Gaming on the couch. There's people installing them into Smart TVs, and other vendors pushing out overpowered ARM Laptops. So now, there's a piece of ARM technology from your watch, to your phone, your convertable tablet-laptop, home console, work desktop pc, and the ARM-Based Cloud servers. And it all started when Intel had their pants down, and AMD was still clawing their way in the R&D Lab..... and all it took was one socket/motherboard support to start it all.
It is my bet. They already design ARM chips and they are good in convince developers to develop for new stuffs they are doing (like the touch bar of macbook)
Not impossible buy my guess it will take atleast 2 years, because even if their ARM chips were on par with Intel they will still have to emulate 99% of time until developers make ARM programs.
Already did ! Iphones, ipads run on pretty good custom arm... To disagree, performance per watt is still slightly higher on intel in server market. Not talking white box, talking 40-896 core.
@@GaryExplains Well, curently there exists only 2 official systems, which you can buy. One is 70dolar Arduino like board but with risc-v. The second is 1000dolar full risc-v, 8gb ram, 4core cpu pc, which can run full GNU/Linux. It even can run Quake. It is clear, that currently there are no good pc option for avarage human, but it is only time question. Development boards already exists. You can already develop software on actual hardware. I doubt that Linus would recomend another proprietary ISA. Linux foundation already support SiFive (the company behind Risc-V). Linux foundations have even given several speeches about Risc-v. I think, that we (and Linus) should wait littlebit for good Risc-v systems instead rushing with ARM. It will be way better in long term.
Thank you, you just proved my point. PS. We aren't "rushing with ARM", it has been around a long time. The architecture is now gaining traction in the server market and the video is about making Arm based development boxes available based on the emergence of Arm servers. If you track how long it has taken Arm to get to this point, it mystifies me why people think that in a few short years RISC-V will get to the same point. There are so many hurdles of overcome and making the ISA open source hardly solves any of them. PPS. All of the key ISAs supported by Linux are proprietary and it hasn't stopped their support in anyway whatsoever. PPPS. Most people are supporting RISC-V because it has the words "open source" associated with it, without actually thinking about the economic and engineering problems of creating a high performance processor.
@@GaryExplains It's not that mystifying when you look at who is backing RISC-V. Throw money at marketing and some times people will trust and believe it. X86 has been around for a long time because of a few good reasons. You know this very well. It's a bit mystifying that people will suddenly turn to ARM or RISC-V( or what ever fad new hardware architecture comes along) rather than the proven x86 architecture. There is a really good reason why people don't talk about Motorola CPUs any more. As far as I can see ARM/RISC-V has a few niche markets but when it comes to general work they suck. Think ATOM vs I3.
@@skilletpan5674 ARM and a few niche markets? My router uses ARM, my phones (plural) uses ARM. Sure I have lots of x86 machines also (only one Intel though, mostly AMD ones). But ARM is not a niche market these days. But anyway... I'm oldskool and want to know the hardware. But my guess today's programmers are pretty much separated from it with all the cross development tools available these days.
1. What makes you think the ARM PC's pricing wont jump substantially when its built for a desktop environment? 2. Why change system architectures for the sake of cost savings if you aren't developing specifically for arm based devices? 3. Will the systems be expandable or will you be stuck with a phone desktop hybrid? 4. You might escape intel or AMD in the short term but somone still has to produce and sell the product even if the liscensing is cheap.
A year ago I would say "No way, it's not happening" and start arguing with everyone on that matter. But after seeing so many benchmarks showing how powerful M1 chips are I might reconsider it. Seems like we are actually getting to the point when x86 platform might get obsolete. I guess this is the where technology is heading and it seems to be a fine idea after all :)
We need ARM PCs for another reason, competition. Intel needs competition to keep it inovating and it's prices reasonable and we need competition for Intel so that were not faced with a monopoly.
@@bobbyb6053 But there's no competition at the i7 upwards end of the market and that's where Intel can charge what it likes for its products and there's no external pressure to innovate.
The issue isn't making a motherboard. It's that there's no ARM cores that has enough single threaded performance to compete with viable x86 development processors. Amdahl's law and all of that... Perhaps make a CPU with something like 4 big ARM cores able to compete with high end x86 cores and 10-14 small and weak but power efficient ARM cores and put that in a development box. Then scale that up for servers with four of those pr socket in a four socket configuration with enough memory bandwidth to feed all of those cores. The four big cores could do the heavy lifting on the non-parallizable parts of the code you need to compile. The large number of smaller cores would allow you to test out the scalability of your code on ARM.
Yes. The only reason swappable motherboards exists for Intel X86 is because at some point of time, IBM compatibility was required. So, all components needed to be compatible. The BIOS made sure of that. There is no such thing on ARM side. Every designer is free to build their own designs which are completely incompatible with other ARM computers. You can't just switch your GPU/RAM between two ARM systems. Also, because of this, the OS is integrated to the platform. An Android image for one PC will NOT work on another similar PC (even running a similar processor). This is easily seen today as ROMs are incompatible between ARM phones.
@@Luredreier "there's no ARM cores that has enough single threaded performance..." where did you get that ? Look at these: browser.geekbench.com/v4/cpu/compare/9948464?baseline=9972001
@dennis Are people still developing AmigaOS? I mean it was pretty cool that they managed to build a fully preemptive multitasking operating system in less than 256 kilobytes of ram.
How bout the Chromebook Plus that Samsung makes with the Cortex A72/A53 processor? I haven't tested it myself, but I know it already supports Linux apps in the developer channel, as well as Android apps by default. It's only $400 on Amazon and seems like it would check most of the boxes you laid out for a personal development machine, with the portability, design aesthetics, and battery life being added bonuses for everyday use.
If they started working on ARM workstations right now, how many single digits of years would they have before RISC-V workstations come out and instantly outcompete them?
Your question doesn't make sense. Since RISC-V is starting from zero and Arm (or Intel) are decades into making chips, then RISC-V development would need to be multiple times quicker to catch and then "instantly outcompete them." Who is going to make this super fast RISC-V CPU? Where is the investment coming from? What secret does this theoretical RISC-V CPU have that means it can compete with Arm? Will Arm development stop during this lead up time?
For years people developed on x86 and deployed on Sun machines, so cross architecture wasn't a barrier. They started deploying on x86 because their performance improved, and they were cheaper than Sun machines.
There is a serious cost to sacrificing dev/prod parity. Notably for android developers who mostly run x86 machines and write for mostly ARM based machines. Many of them have to use an actual ARM-based android device to check their own code. The cost of losing the parity is quite large when spread out over the entire industry and certainly did play a part in the rise of x86 server hardware. Parity allows for shorter development pipelines with lower overhead.
@@TaylorIserman Android is primarily programmed in Java. At that point you're more at the mercy of Google's VM than the underlying hardware. When I worked on a mobile project we did have a device lab, but it was more for reproducing specific bug reports than day to day development. We also developed a set of mocks that allowed testing much of the code during the build rather than deployment on a device.
Yeah I’ll concede that java has helped make a lot of app development on android more agnostic. I’ve had issues with an app I wrote that dealt with too many system components to be that way. I still maintain dev/prod parity is a major factor in the prevalence of x86 servers today. To be honest, the main culprit tends to be the OS: like when the intel-compatible version has certain differences or when an OS only exists for some type of hardware (like AIX *grumble grumble*).
@@TaylorIserman I agree OS differences can be a major annoyance, even when it's just two different Linux distributions on x86. For a few years I had to deal with Oracle Unbreakable Linux on servers with Ubuntu on the desktop. Oracle's Linux is forked from Redhat, while Ubuntu is forked from Debian. I swear it was somebody's unethical psychology experiment to see how much swearing people are capable of.
Most server applications these days run in virtualized environments. It doesn't matter what the underlying hardware architecture is as long as the behavior on the software layer is the same.
Where do you work that it doesn't matter? Resource consumption, cost, ease of use, these all matter, maybe not to you directly, but to the chain and process which allows your container to function, so ultimately to you too. Massively.
That only applies if you have programs running in a VM (Java, .NET, JS...). If a program is compiled for x86, no matter if it is running in a container or not, it will not run on ARM and vice versa. You may use an emulation layer of course, but that will be slow and resource intensive.
Emulation is a temporary fix, while the code based is moved to the new ISA. Just the same way apple managed the transition from 68000 to powerPC and from power PC to intel. The latest windows on ARM has intel emulation for the same reason. On the subject of intel they are having trouble manufacturing at 10nm while TSMC and Samsung are at 7nm and testing at 5nm
I'm not too familiar with software development, but aren't there emulators that can be used to test or develop? So it wouldn't matter what was used to build?
ARM needs wider compatibility/maintainability as well. I can install pretty much any flavor of Linux on almost any x86-64 platform. I can install Windows 10 on a Core 2 Duo and expect it to keep getting updates 5 years down the line. That's not necessarily the case with ARM.
People like Jon Masters at Red Hat have been doing a lot of work to make sure you can install any flavour of Linux on Arm. The new Fujitsu A64FX boots Red Hat linux out of the box, not because Red Hat has a special flavour for Fujitsu, but because of the boring standards that RH has been pushing. As for the updates, my guess is that Microsoft is committed to supplying patches for the always-connected laptops.
@asdfrozen OS support isn't really that much of a issue with ARM anymore as it was in the past. The main issue is application support as the majority of both Windows and Linux applications are still coded on X86.
@@Luredreier Most Linux applications already run on ARM. Take Debian (just looked it up just now): 99.6 % of official packages build on AMD64. And on ARM64 it's at 99.4% of all packages. Can't get much higher than that.
It's better than you think, but definitely could be better and would be if something laptop/desktop grade existed out there. It's not like it's hard with the current tech. It just needs to actually be available, preferably as systems that you can get setup with just a few easy clicks.
+@@autohmae Who out there limits themselves to apps in the reposetory? Oo Back when I was using Linux Mint the wast majority of the apps I installed where *not* from the main distro repository at all. Same on Manjaro. The official packages are only a *small* portion of the apps people actually use even on Linux. On Windows it's even more common to use apps that the OS developers have no controll over at all. There needs to be developer systems out there with proper prosumer performance available on ARM before they can truly take off.
I mean all the big players, Apple, Microsoft, and AMD are looking forward to ARM based systems. We're probably a little ways out from ARM completely taking over the market, but that's where literally all the big players are looking toward. And Qualcomm is set to be the next Intel as far as who will be making the chips that power all our future tech.
Lol. Linus said exactly the opposite. He's very skeptical of ARM architecture and thinks it will sink. Though recently he changed his stance a little towards "we'll see how it ends".
The problem is that ARM only makes sence in a SOC. If you want a PC that has a socketed CPU and is ARM based you would have to give up half of the SOC to a north bridge and a south bridge which dosen't makes sence cost wise, since the main advantage of ARM chips is in low parts count. Maybe with 2 chips? And, can it handle the bandwith of PCIe 2.0 or SATA 3.0?
8:05 I’ve done cross-development(embedded, not server stuff) for almost 6 years and it’s just painful, especially on Linux. Issues like: the target communicating only over serial and the pains with that, not enough space on the target to put a debugger, linux-tools(perf) are not compiled with the kernel and it’s heavily hacked that it’s almost impossible to maintain a build of the linux-tools, firewalls, special flashing steps to open the communication with the target, etc... Developing on ARM would of definitely made all of that easier - although, later I kind of worked around that issue by using QEMU + qemu-user-binfmt and faking some dependencies.
You would think that Qualcomm would have an incentive to release arm dev boards, considering they want arm to replace as much of x86 as it can. Hell, Microsoft is pushing hard for arm windows 10. Why not release a surface arm pc specifically targeting developers.
No we dont. We need risc v everything if we want to avoid backdoors built into processors. They all built one and dont give af they even say they do as if thats good thing but theres no option really until risc v develops further.
I absolutely disagree. RISC-V is an open source ISA. That means a company/group/organization can create a RISC-V CPU and sell it. They are under no obligation to release the "source" for that CPU. And even if they did, they can still produce a different version to the one they publish publicly. RISC-V solves none of these problems.
@@GaryExplains I am not sure what point are you trying to make lol. They will trick you into getting different version of the firmware? Is that what you are saying? :))
Theres nothing that runs invisible. It must use electricity, memory and components and produce binaries that if are unknown there is something wrong. But since all is open source and you can verify by code itself idk what the f you talk about them selling you something else lol
there is also the problem that the arm devices are too fragmenting ... each device requires its specific version of linux with the kernel specifically configured .. and this is boring and difficult ..
Guys, you forgot the actual SOFTWARE. And actually there's a huge question - is there any way to efficiently port a x86 to ARM? It seems if we speak of emulation (compiled x86 binary on ARM) that it's simply impossible. Am I right?
No, you are wrong. Is there any way to efficiently port x86 to ARM? No. But both Windows and Linux (as well as most other operating systems) already run on ARM. As does most software. Only games tend not to. And the difference between a computer and a calculator is the ability of the computer to emulate any other computer, given enough storage. That means that you can emulate an x86 computer on ARM, albeit noticeably slower than it would run natively (mostly due to the complex instruction set of the x86).
If he had said risc V I'd be on board but arm just replaces x86, nothing changes and you will have even less development until apple arm did nothing for a good decade or two
RISC-V isn't what you think it is. I have replied to loads of comments under this video about RISC-V and most people just don't understand it right. I will be making a video about it soon. But to address you specific question, we won't see competitive RISC-V processors for many years yet.
What is really needed is open-source platforms (CPU and Chipsets), not just operating systems. Far too much hardware has only Windows / OS-X driver support and some OEMs don't want to open up/offer documentation to make drivers available.
I want an ARM-based motherboard that supports an ARM RISC based CPU but where is it? That is what we need and then add RAM, etc. to build an ARM-based PC or server.
I agree, for most use cases. It's a lot easier to use a runtime like .NET or Java to compile on-the-fly to the local machine's native architecture than it is to make the hardware and its drivers conform to your expectations. And performance is fine most of the time.
I'd like to see PowerPC desktops make a comeback. Though what an Arm based PC would need is proper PCI-e expansion slots, expandable ram, and SATA ports.
@@laurelsporter nope.nvme . stopped using spinny drives in 2008. Not the point. ARM chips process video about ten times faster than x86.my question is why.
@@Ballistichydrant they don't, if you're comparing CPU to CPU. A nice x86 CPU *will* be many times faster. Now, if your specific hardware has a DSP or GPGPU that can be used for it, while your x86 doesn't, that could make a noticeable difference, especially if you're talking mainly about rendering, where GPUs can often process video many times faster than CPUs.
Why would you prefer a system that doesn't even exist in any real significant terms. Why does RISC-V give you, you personally, than other architectures don't.
Why do we need ARM (RISC) based PC's over Intel based PC's? What are the benefits? Also, why ARM (RISC) over RISC-V? RISC-V is a better alternative over ARM (RISC).
For me it's cost. The development I do is not tied to the architecture as most languages cross compile with few issues (Go, Node, Python, C#, etc). Price to performance is what I need and ARM seems to do better with more concurrent workloads.
Start with an ARM ATX motherboard suitable for a workstation or small server, and go from there with all of the PC peripherals, cards, and attachments.
Sure, a native workstation for developers is very helpful. That's not going to make ARM faster, though. It's still very far behind in real-world performance.
Developers tend to use virtual machines, where I work we have many hundreds of these. We can provision them quickly, vary the spec in minutes, and deploy a VM in production on-prem or in the cloud. You can run VMs in Win 10, or VMware workstation or Oraclebox. It should not matter if you develop on an AMD and then run production on x86 - compilers take care of that. CPU architecture should be abstract in most applications.
I have literally walked out of a bar and found a PC sitting on the curb that I took home with me. That's how ubiquitous X86 based PCs are today. ARM has a long way to go to get to that point.
The whole "ambient computing" concept MS mentioned a few years back, I think we are there now. I cant see the hurdles of working on an arm-based laptop with away from my desk, then using a x64-based multi-core'd beast as a desktop, with all my data synced via the cloud. We are already doing this (to a degree) with smart-phones. Very few of use are processing workloads that are architecture-specific and cannot be accessed remotely. Amazon already have arm-based instances available. Whats the prob again?
Counterpoint to the argument for matching workstation and server chipsets: so long as you're writing for the same OS and using a well-defined programming language, you will still be able to test a functionally equivalent program on the development machine.
@Richard Vaughn I think you misunderstand the nature of RISC-V. Only the ISA is open source. Any company, like NVIDIA that adopts it still has to design and develop a superscaler CPU core and then make it. Neither of those tasks are cheap. And their design remains private and closed. Ultimately it doesn't help or harm consumers in anyway, over other ISAs like ARM.
Problem is performance, ARM's isn't that good compared to x86 in performance and it looked like ARM's was going to make inroads into the PC market with how slow x86 was developing, mostly thanks to Intel but now ARM's development is slowly down whiles AMD with Ryzen is shaking up the x86 market, the gap was wide to begin with but now it could get wider when we look at what AMD is doing with Ryzen. Now don't get me wrong, I'm all for ARM's on the PC for added competion but software compatablity is the number one issue, it's all mostly designed around x86 and that wont be easy to change, performance is the second issue and I know they can scale things up but then software development gets more complecated. So if AMD didn't wake up the x86 market, there was a real chance of ARM's making real inroads intos the PC market because performance leap per year was faster with ARM's but Ryzen seems to be changing that as we are seeing 8 cores at mid level prices and likely 16 cores at mid level pricing over the next 2 or so years, if it was up to Intel, we would still be on 4 cores and that left a big opening for ARM's to take advantage off but all thats changed now.
What you are suggesting will only occur if there is an explosion of mITX "carrier" boards supporting products like the RaspberryPi "Compute" module. To this date only the TuringPi vaguely fits this form factor.
@Lassi Kinnunen It was on Hackaday a few years ago. The article is called the "Building the worst Linux PC ever". The effective clock speed was 6.5 kHz.
@@GaryExplains I have to say, I'm not sure what most developers build, but everything I build runs just fine on low specs. :-) It's actually becoming easier with Docker, etc. microservices really are smaller.
For the new year I bought a quad core computer, Windows 10, Intel 64 bit chip, 4G ram, 32 G flash stuck my SSD drive inside and got it delivered for about $250 US (not including the Samsung SSD). The company that makes this computer (the size of an Apple TV box), with power supply of course was Azulle and it is made in the USA. Not ARM but everything else you would want.
Where I work developers develop Java applications on Windows desktops but the application is deployed in the production environment on Linux-based servers. The Java JVM allows us to successfully use two different operating systems for development and deployment. Would this also work at the hardware level between using x86 or ARM based hardware?
@@nicholasjohnson3542 One thing great about [insert any paid or exclusive OS] is their customer support. Because you've paid some amount of money, you're entitled to get some support. ( No need to too often disturb your computer wizard friend ).
@@rommysoeli only losers need support when there's a thing called the internet. Also you're naive if you think that the money you paid actually gets you actual support.
@@TheZaman_ lol it doesn't have the highest, I wish. If you are making fun of me, go and find some people who are actually happy with Microsoft support or something. And why is stackoverflow and sibling websites a thing if people don't search the internet for their problems and ask for help? Better to get support by asking your peers who are in the same situation, rather than unknkwledgeable support people reading from a manual.
Could containers render the underlying hardware architecture irrelevant? You can build on a x86 machine and deploy it to a container running on a Arm server
Does this problem exist if we're developing software that compiles to an intermediate language running in a virtual machine (eg. Java or .net). A lot of this now runs in an abstracted way in the cloud these days anyway ("serverless" architecture - eg. Azure functions), so we don't even know what hardware it's running on.
It is the same reason why MS Windows and MS Office is still so prolific; it is what people have either at work or at home, so the people tend to keep what they are already used to using elsewhere. I believe Windows Server should not be a thing, but it is, only because people who are familiar with Windows on the desktop are more easily able to transition to the same environment on servers, so that is what they go with. It was not that long ago we had servers running on MIPS, SPARC, and Alpha architectures that all fell out of use because people did not want to split their time learning to code and provide support for multiple types of systems when one standard could do the job. Even if ARM desktops and servers become available, shortly afterwards, either the ARM or the Intel based architecture would would become the sole system in use. Mobile devices had many types of architectures, but they have all been whittled down to two, and I would not doubt that within a decade from now, one of those two will be gone.
getting ARM scalable to x86 is kinda tricky. Things (SATA/PCIe/etc.) are actually build based on x86, so you need to either have pain adapting or taking time building a brand new same kind of things with ARM architecture
I'm expecting some kinf of ARM mobo, ATX format, with all the connections and features one can find on an Intel/AMD motherboard. Only a few of them, impossible to find, very expensive, and desperately slow...
I'm curious if the performance gains on ARM are merely due to the 7 nm aspect. If so, x64 should be within touching distance to that in 18 months. It's not a "Power PC roadmap" event, more like "wait until next year and the performance will be equal" hiatus.
No if it was about the process node (like 7nm) then you coudl take a Pentium II processor from the 90s, make it on 7nm and you have a high performance processor. It is mainly to do with the microarchitecture. The process node helps with power efficiency, but it isn't the whole story there either.
@@vardekpetrovic9716 I read recently that Apple are open to the idea of Windows 10 booting natively on the M1, and it was really up to Microsoft to get on with it. Assuming that's true, it should be the same for booting Linux. (Though obviously this is a huge simplification.)
Yep. Totally agree with you. To name a few, Java an golang and JavaScript try to make it easier to achieve that but developers dont all get that vision. I think that vendor lock-in makes money and thats the biggest driver for not adopting these basic principles
I've been saying this since Cortex A72. And not some generic dev box. Just make the silicon socketable. There are already people out there making embedded boards. IPC parity (or damn close enough to it) is here. It's freakin insane that this hardware is being held back from mainstream computing users.
I think what we really need are not ARM, but definitely some RISC processors with an **open** architecture.
RISC-V
EXACTLY!
power9
YES.
SPARC
I want RISC-V-PCs!
Really you don't!
@@GaryExplains what is RISC-V missing in comparison to ARMS RISC? Except developer support of course, but as you said it's not like that is much of a difference nor set in stone
@@GaryExplains Would you EXPLAIN? I just like open source and I really hope that it won't take too long until we see the first RISC-V consumer chips...
Yeah, I am planning a video on RISC-V... The big thing is that the ISA is open source, but that doesn't give you very much. Then even if some company kindly open sources a design, then what? You going to make a contract with TSMC to make a chip.
It isn't about one ISA versus another. You could say what is ARM missing compared to x86, or what is MIPS missing compared to ARM. This isn't the important question. Open source "software" is great, I can grab a compiler and go. But open source "hardware", especially for a chip, is nonsense. Do you have a contract with TSMC to make 7nm chips, I don't! Plus it is only the ISA that is open source, not the actual chip designs.
Funny I develop (PHP + MySQL) on Raspberry PI and deploy to x86 instead, mainly because it’s small and cheap, plus the low performance made it easier to identify potential performance problems before deploying to the live server...
more power to you then for being able to run in a tiny constrained Env. - You'll be a King in the low power IoT
Same here but i use node js instead of php . It can easily handle the loads during development
I don't know about that. I mean, first it's arms, then legs, next thing you know there's a robotic uprising.
🤣
It could happen
😰😱
lol 😂😂 Seems inevitable
I laughed. Great joke
I perfectly fine with my PC not having arms
I'd love a LEG PC.
I perfectly fine with my horse. I dont need a car
It has already eyes and ears
They tried ARM based PCs multiple times over the past 10 years. There are a number running windows 10 ARM, but those laptops (e.g., the Asus ones) are overpriced for the hardware. For example a $1000 laptop will have a Snapdragon 835, 4GB of RAM and 64GB of storage, with steep price premiums for more RAM and storage that can't be upgraded.
On the X86 side of things, you will be getting a current gen core i5 CPU, 8-16GB of RAM, 256GB SSD, and a mid range dedicated GPU. The issue is really one of companies making ARM PCs, are expecting to get the Apple 450%+ profit margins, while selling an ARM system that benchmarks lower than a $300 entry level x86 PC.
Wanted to add that if you go through all of that expense, you ultimately end up with a system that will not run all of your desktop productivity applications (especially 64 bit stuff), performance will be lower, thus impacting multitasking. There will also be a lack of standardization thus it is not easy to change operating systems, and you will be largely reliant on the device maker to maintain support (which may at most last 18 months). If you need something basic where a SD835 is enough, then you can spend a tiny fraction of the price to get a upper mid range laptop from 2013, and still end up with better performance.
I think you completely missed the point of what I am saying.
@@GaryExplains The overall idea of what I was trying to bring is that we have to think beyond the direct links and look at the indirect links. You alluded to it in the video where you stated that you tend to develop for the hardware you are using, but developers are not some mythical creatures, they exist in the real world and buy hardware from the same places in general that those of us who do not code for a living, will buy.
For an ARM system to become mainstream, it has to be built in a way that appeals to a wide market, otherwise it will not be worth the warehouse space at a place like newegg and it will not be worth the time for companies to put a ton of R&D into (unless that R&D stands a good chance to shift the market over to ARM). The current ARM PCs don't come even close to that and are relatively niche, and without market share, core issue in your video will be stuck in that catch 22 situation. The system you want is not being developed because the low end versions of what you described, are massively overpriced such that no one wants to buy it, and because no one wants to buy it, the companies don't want to invest in R&D for something that is not generating an ROI.
The solution you want, requires ARM to reach a point where the choice between ARM and x86 becomes like a choice between Intel and AMD when building a PC. Simply put, a system designed purely as a high perofrmance dev box, will not be popular or in the least bit cost effective, people would rather emulate an ARM environment on an x86 platform for basic debugging.
The billion dollar investments into ARM CPUs are pretty much always for the mobile market, thus you will not get something that will rival an Intel Core i7, unless they can get a market where such a CPU , would sell as much as a Core i7 since unlike other markets,CPUs have a massive margin over BOM due to the fact that they have a very limited time window to sell before the tech becomes old and something new is needed in the market. If Qualcomm were to dump 1-2 billion in developing a high end desktop ARM CPU for use in a dev box, it is unlikely that enough software developers would purchase it every year to sustain that business, unless you are okay with spending $100,000+ on a PC that will perform like a $800 x86 PC, and spending that money every year or 2.
@@Razor2048 Sensible comment.
Other than raw power x86 CPU not really beating ARM SOC.
I once developed some software in C++ on my x86 computer, but then was deploying it to an ARM based Raspberry Pi. Cross-compiling was interesting for sure, I ended up giving up on compiling it with my computer and had to compile it on another Raspberry Pi. I ran into a few bugs that occurred only on the Pi and not on my development computer, mostly due to data types and the 32bit vs 64bit difference.
Exactly. Now if you had a desktop Arm machine that would have all gone much smoother!
@@GaryExplains Indeed it would have
That's pretty much what Linus said.
You develop on Amdx64, so you're likely going to deploy on that arch as well.
Indeed you can. But it just isn't powerful enough for serious development.
@Richard Vaughn The whole point of the platform was to encourage kids to gain interest and maybe become developers. The fact businesses buy and rely on them is a side benefit the people who make them. It's useful for very low end compute-non intensive apps. I would use it for a security layer to separate connections with the outside from the inside where all the heavy computing is done. It can be too slow for even that, so only until I can afford a more powerful ARM or RISC-V based SBC, then a cluster of them.
Shame the whole ARM ecosystem is a horrific mess. Just getting a recent kernel running on most ARM machines is frustrating. Until ARM enforce ACPI etc for everything and open up their GPU drivers properly it's never going to happen.
correct!
It just takes a platform leader (Apple, Msft) to build a cash-cow and guard it.
When cloud giants (AWS, Google) develop their individual custom chip no 3rd party gets in on it besides the resulting SAS it delivers.
Thus war is all about IP (Intellectual Property)... Free Linux broke UNIX licensing but has not killed proprietary x86 business because it's been fragmented.
This will not happen. And it's a great way for vendor lockin on hardware producer side. They haven't learned that they will fail with this attitude. ARM has no server future. Not against AMDs new technology.
I said this years ago. Currently, there's few experimental ARM-PCs out there in the R&D Labs, but nothing for enterprise or consumers.
I made the argument that during 2015, there should have been a collaboration to do this with 45W CPU based on 4GHz 8-core ARM Cortex A72 on 14nm, with 64-bit instruction set. Ofcourse, it would've required collaboration with someone like ASUS to make a Motherboard where you can plop in an ARM CPU. And hopefully support some sort of standard RAM support, and PCIe slots for dedicated graphics cards.
They could've later made CPU improvements like stepping up to 10nm, 7nm, and 5nm later on. And microarchitecture improvements to A73, A75, A76, A77, and A78. Even throw in a budget options, with a range varying from A38 and A58 CPUs, in 4-core, 8-core, and 16-core configurations, running anywhere from 2GHz to 6GHz at TDPs from 5W to 90W. Have it running Native Linux Distro such as an Official Debian build, a properly optimised AndroidOS from AOSP/Google, and even a Windows 10S... things will take off.
Before you know it, native applications are built on ARM PC's with software for ARM Devices, and developers jumping away from x86 systems. Next thing you know there's actual ARM Tablets, Ultrabooks, and Gaming Laptops out there.
@@llothar68 Hi, I'm here to inform you it happened. The Microsoft Surface Pro X is ARM based.
@@roberttalada5196 And it is a total failure. Even the Surface Pro 7 is better then this 600 Euro mehr expensive book.
With how ARM does it's licensing, I rather not have ARM becoming mainstream on Desktop/Server tbh. We don't need another Intel
That's the main reason why there's such a huge push,supported by Google and other behemoth companies, to develop RISC-V.
And the potential there is even greater than ARM tbh.
Then we could maybe have PowerPC and then SPARC, and so forth. Having accessible ARM PCs would greatly increase the potential of more processor architectures competing in the market, plus it would massively impact Microsoft's dominance because their trump card of having all that software library disappears when people are now using a whole new architecture, more so considering that there is already a whole library of linux software for a number of other architectures thanks to Debian
@@jackmcslay SPARC, nice as it looks, is now just a Fujitsu thing, and it likely does not have a lot of life left. It came from Sun and Oracle decided there is not going to be any more done for SPARC.
Software developers cries
But anyone can produce ARM CPUs. Yes, the license fee is the part that needs competition, but ARM is very cognisant that it doesn't want to be seen as an Intel, exploiting it's market for all its worth for short term gain. It's fees are pennies per CPU, even though it has very little in the way if competition in smartphones, and the profits are large for some of them. It makes very little in the way of gratuitous profit, most of this cost goes to paying engineers to make the best chips of tomorrow, not to line shareholders pockets. Intel in the other hand has extortionate profit per CPU sold. RISC-V will always be nipping on ARMs heels and make sure they don't abuse their market position, and its clear ARM will allow any company to make reasonably priced, excellent performing CPUs for cost price of silicon. I see no reason to think it won't keep doing that. It only operates in a tiny part of the device technology stack, and is more replaceable as a result, unlike Intel that does fabrication and total arch control. ARM has many partners all competing with each other to implement their designs, many fabs competing to fab those designs. It is no more than a licenser of the minimum logic needed to get a CPU to function and function with the current ecosystem. Its ability to be extortionate like Intel is is much smaller, and is nothing like Intel does, even now when it has smartphone dominance. You've probably paid no more than a few dollars to ARM in your lifetime of owning many many ARM CPUs, but probably hundreds to Intel because they can abuse their position much more easily.
when we're done falling in love with ARM, reality will set in. As long as we can keep spending top box for new gear, the industry will come up with new reasons to switch.
Thin-clients, client-servers, SAS feeds big-data for AI... pretty soon you won't need to remember your name, just your ID.
LOL
That would need a business case; PC makers need a demand from a big portion of the market to flip to ARM, i honestly believe the "long battery life" laptops based on something like snapdragon socs might be killer application that would popularize ARM PCs adoption.
Or, build your own: ruclips.net/video/jNnCok1H3-g/видео.html
Business case? Apple. I believe if anyone is going to do it, it will be them. It makes a lot of sense too, and they can keep their ecosystem tighter.
@@ashlord8190 perhaps so, but i certainly believe the ecosystems mr. Travolds is implying is Linux.
@@ashlord8190 Microsoft built a Win 10 build and an entire development tool set for ARM. As soon as someone builds and optimizes a modern web browser for ARM on Win 10, besides Edge, it can take off. I saw a side by side on an ARM netbook vs one similar priced with an Intel CPU. The ARM was faster, except when running programs that needed x86 emulation. Edge loaded everything faster on the ARM; it was built for the platform. Chrome was slower; it needed emulation. ARM beat Intel by a large margin for battery life. Chrome OS based netbooks didn't beat everything else because of limited local storage and almost entirely depended on cloud compute and storage. Put enough RAM and storage in one with an OS that enough people are comfortable developing for, then you have a winner.
@@mas921 *Torvalds ?
now it's here with apple
Apple has finally MADE it a thing this decade
It's about time. They were tired of people making Hackintoshes.
But we will get there eventually. They just need the dopamine rush of being alone for a while. However, Louis Rossman will break it and send videos on repair accross the world (youtube)
@Dex4Sure "Soon it will be NVIDIA as NVIDIA is in talks to buy ARM!"
There is no-way that is getting past the anti-trust officers at the justice department. Oh, wait. I just remembered. Government is corrupt.
@@sam11182 He will probably do it out of spite, don't delay buy today
@@jefflittle8913 hahahaha Obama Government certainly had its fingerprint following the mottos of the past bad ones along with the everlasting Alphabet (Google) who has been changing the laws for themselves and changing the names of every single part of their hive mind to avoid the same laws XD
But Nooooo they have a reckoning to pay and they will pay indeed!
Ever wonder why they have been giving out "Free Google Homes" for contests that are dead simple to win? They want to know every single secret of your life.
Your phone, computer, everyone else's around you, you are a pawn in their hands. However if you play the game right, your pawn will reach the other side of the board and will immediately become a queen. You can rule as well.
nVidia should own arm so that Apple does not have the ability to rule the computer world. I honestly don't care.
@@jefflittle8913 Just sell ARM though a shell company in South Africa or something. You can distribute software anywhere in the world, look at Ubutnu for example, it's run out of South Africa. Governments need to realize that in this modern internet age if they become too annoying for digital companies to deal with, those companies can just leave & then government has absolutely no control, so they better not push their luck.
Signal for example said if the US tried to take away the legal protection companies have from the activity of their users that they would leave the US & just operate in the British Vrigin Islands where privacy is a more respected right. If they did that then now the US would have absolutely no control over Signal, so you see how that could be a problem for modern governments if they decide to start enforcing anti-trust laws left & right. Companies will leave, & they wont lose a cent of profit doing so, if anything they'll save money being in countries with lower tax rates. Now the US has no control, & they just lost a bunch of taxable income. This is one of the few good things globalism has created as countries now need to be competitive with each other.
I miss the old BYTE magazine which used to explain technical stuff like CPU design in their issues.
We do have ARM based PCs... you can buy them as laptops usually, but desktop ones are out there...
They exist. I had one... problem is they are trash compared to x86 in performance.
this video is also a 30 second loop of the same thing over and over, why?
It's a recreation of the Groundhog Day film. He pretends we have forgotten and repeats it multiple times.
Actually they aren't trash compared to x86. They are only trash in emulating x86. They run native apps quite well - which is the point.
These comments didn't age well after Apple introduced the Apple SIlicon M1 and now M2. And now there's a Raspberry Pi shortage
9 minutes to repeat 1000 times the same.
I guess you didn't understand the video then.
True.
Yeah this video was uncharacteristically rambly.
@@deus_ex_machina_ He doesn't agree. Evidently he was improvising and thought being confident with not enough content was enough.
@@karlpj1 you gotta pay attention, let me explain. The video is explaining, how does ARM cut into the market? Small businesses go from regular dev pcs to Intel servers because of the easy transition, despite the outrageous multi thousand dollar cost of server cpus. ARM has to drastically cut the cost, from the thousands of dollars, to hundreds, to convince business to make the jump from regular x86 development to ARM servers. Follow the narrative, your eyes and ears are your most valuable organs.
Yes, Apple announces Apple Silicon ARM processors yesterday :)
bump
Yep. And if your a developer you can apply for the $500 ADK it’s got 16GB Ram, 500GB HD and all those ports on the back. Or you can just wait a couple of months till the release of an ARM laptop from Apple and use that.
I'd be incredibly happy to develop for Android from a full-featured ARM workstation.
What is the cheapest ARM based motherboard with PCIe, SATA and RAM slots like the regular x86 ATX mobo you can buy for around 100$?
@Viktor with a K There already are but they're for development and expensive, I still want to know which one is the cheapest.
@@owowowdhxbxgakwlcybwxsimcwx Thanks.
You're missing the point. One of the main reason ARM was made was for mobile platform so that it would fit as many components as possible in the tiniest space possible. It won't be "slot" but more like SMD part (soldered). Sata is also not a slot, it's an interface. There won't be pcie too, it would be just some connectors like raspberry. Way slower, but more versatile in terms of what you can do with it.
@@ragilmalik That might be what it's used for, but that doesn't it's the only possible usage for it. ARM could be a competitor for X86 & X64 but not unless it's put on boards that allow for expansion and customization.
Yeah, kill modularity. That'll be definitely a good thing to trash a whole machine because a single component died. Screw mobiles platforms, tbh.
A cheap linux arm-based PC which is capable of running for 10 hours or more would be awesome.
Mine run for months on end.
Did you mean “on batteries”?
@@lawrencedoliveiro9104 that's probably what he meant.
That's not a PC, it's a laptop.. unless you're able to swap components around it's not a Personal Computer :)
@Cnn is Fakenews Did you even watch the video lol?
Yeah ARM CPU and an RTX2080.
Right, I've been developing for ARM from x86 PC. To test, I have to upload the resulting cross compiled executable to the ARM machine, sometimes far away remotely and it takes significant amount of time, while the resulting executable doesn't always run, let alone run right. If only I can test it locally, things would be much more efficient.
And then try to do fucking complicated multithreading shit like lock free algorithms. You really need local powerfull testing equipment for this. You can't test MT programs on a single socket system. Write reording is a magic that few programmers worldwide fully understand (i do not). So Linus is right, unless there are powerful desktop machines for developers the ARM will continue to fail in the server world. And with all attempts from the past, the ARM servers were disgustingly bad and overpriced - no surprise they all went out of business or even bankrupt.
@Richard Vaughn Not all ARM chips are created equal. A program created for Cortex-A will not work Cortex-M or ARM7TDMI chips, and vice versa.
@Richard Vaughn the sad irony is that not all programs to create ARM programs are runnable on ARM itself, some are really just made with cross compilation in mind.
I feel like ARM corporation (SoftBank?) lost a golden opportunity back in 2016.
Imagine at the end of 2015, they released a motherboard that you can hot-swap in your own ARM CPU. And it would be a modern SFF motherboard that will have two DIMM slots, PCIe connection, SATA, DisplayPort, USB, Wifi, Bluetooth, etc etc. And say the first option was for a 14nm 3GHz Quadcore Cortex A72 (64bit), with 1GHz Mali-G71-MP4. It sure would be competitive for the period. And yes, it would need Active Cooling (probably much weaker than Intel/AMD desktop systems though). And it would have an internal GPU that would be more powerful than Intel/AMD.
And the best part would be, people could plug in their RAM sticks, SSD drives, and dedicated GPU cards. Imagine a GTX 980Ti working with an ARM System. Sure it would be weird, but also cool. SteamOS could be a boon. And before the year ends and that Intel or AMD has a chance to respond, you just double the performance. You're now toting a 3GHz Octacore Cortex A72 CPU, with 1GHz Mali-G71-MP8 iGPU. Next year you introduce the Cortex A73 and Mali G72, but its a little slower but more efficient, so you boost the clocks. Now you're rocking 4GHz Octacore A73 CPU, with a 500MHz Mali-G72-MP16 iGPU. And you plug in your GTX 1080. And they've yet to release the AMD Ryzen to the public.
Fast-forward another year, and now you are using 10nm Octacore Cortex A73 at 5GHz, matching the best of Intel and AMD.... but cheaper and more efficiently. Your iGPU now rocks a 1GHz Mali-G76-MP16 iGPU, which bests any APU's out there. And can even install Windows10-ARM Operating System, alongside native Android, SteamOS, and a plethora of Linux distros (not to mention Server-aimed OS). People have managed to hack these into very smart cryptomining rigs, so there's a lot more excitement. Now, along official ARM CPU's you have them joining you from Qualcomm, Samsung, Huawei and RockChip... all offering hardware on the same socket/motherboard support. Not to mention the likes of ASUS, Gigabyte, MSI and Biostar offering motherboards. You're basically building your own ARM-Server in your home. And things are looking bad for AMD, and even worse for Intel.
The fast-forward another year, beginning of last year. You've refined things further. Now the flagship offering is a +10nm 5.1GHz Octacore Cortex A75, combined with a 1.1GHz Mali-G76-MP32. There's also a low-end/budget model using 16nm 2GHz Octacore Cortex A55, with 0.9GHz Mali-G76-MP4 iGPU. People are developing Apps right at home, deploying it to their ARM-Laptops, ARM-tablets, and the usual ARM-based phones and testing it out. They're also developing program for ARM-based servers. Many enthusiasts are sticking their GTX 1080 Ti's and Vega64's, and running games from steam on their linux system using the new SteamOS/Library. All indie developers are now targeting the ARM Desktops to launch their games. A couple big AAA titles are announced for the platform.
So what about the year following, in 2019 ?
Well, the obvious choice is for 7nm 16-core Cortex A76 CPU running at the usual 5.1GHz frequency. And this time sporting a 1.1GHz Mali-G76-MP64. Overall its enough to challenge the (base) Xbox One in terms of graphics prowess, and challenge the best Xeon/Threadripper in terms of (some) CPU prowess. There's vendors releasing these overpowered ARM-systems, also using dedicated graphics like a GTX 1660 Ti, as standalone TV Boxes that also come with good controllers to do some Steam Gaming on the couch. There's people installing them into Smart TVs, and other vendors pushing out overpowered ARM Laptops. So now, there's a piece of ARM technology from your watch, to your phone, your convertable tablet-laptop, home console, work desktop pc, and the ARM-Based Cloud servers. And it all started when Intel had their pants down, and AMD was still clawing their way in the R&D Lab..... and all it took was one socket/motherboard support to start it all.
qemu-arm might help ?
Apple might be the first one to cash in the In-House ARM based processors
It is my bet. They already design ARM chips and they are good in convince developers to develop for new stuffs they are doing (like the touch bar of macbook)
Not impossible buy my guess it will take atleast 2 years, because even if their ARM chips were on par with Intel they will still have to emulate 99% of time until developers make ARM programs.
Snapdragon 8cx is more likely to succeed than Apple. Windows on ARM already works, but Win32 apps are emulated
It is on the news already. They are switching Macs by 2020
Already did ! Iphones, ipads run on pretty good custom arm...
To disagree, performance per watt is still slightly higher on intel in server market. Not talking white box, talking 40-896 core.
Whoa. Linus Torvalds recomending ARM over RISC-V. Something is not right here...
Which RISC-V CPU and development system should Linux recommend?
@@GaryExplains Well, curently there exists only 2 official systems, which you can buy. One is 70dolar Arduino like board but with risc-v. The second is 1000dolar full risc-v, 8gb ram, 4core cpu pc, which can run full GNU/Linux. It even can run Quake. It is clear, that currently there are no good pc option for avarage human, but it is only time question. Development boards already exists. You can already develop software on actual hardware. I doubt that Linus would recomend another proprietary ISA. Linux foundation already support SiFive (the company behind Risc-V). Linux foundations have even given several speeches about Risc-v. I think, that we (and Linus) should wait littlebit for good Risc-v systems instead rushing with ARM. It will be way better in long term.
Thank you, you just proved my point.
PS. We aren't "rushing with ARM", it has been around a long time. The architecture is now gaining traction in the server market and the video is about making Arm based development boxes available based on the emergence of Arm servers. If you track how long it has taken Arm to get to this point, it mystifies me why people think that in a few short years RISC-V will get to the same point. There are so many hurdles of overcome and making the ISA open source hardly solves any of them.
PPS. All of the key ISAs supported by Linux are proprietary and it hasn't stopped their support in anyway whatsoever.
PPPS. Most people are supporting RISC-V because it has the words "open source" associated with it, without actually thinking about the economic and engineering problems of creating a high performance processor.
@@GaryExplains It's not that mystifying when you look at who is backing RISC-V. Throw money at marketing and some times people will trust and believe it. X86 has been around for a long time because of a few good reasons. You know this very well. It's a bit mystifying that people will suddenly turn to ARM or RISC-V( or what ever fad new hardware architecture comes along) rather than the proven x86 architecture. There is a really good reason why people don't talk about Motorola CPUs any more. As far as I can see ARM/RISC-V has a few niche markets but when it comes to general work they suck. Think ATOM vs I3.
@@skilletpan5674 ARM and a few niche markets? My router uses ARM, my phones (plural) uses ARM. Sure I have lots of x86 machines also (only one Intel though, mostly AMD ones). But ARM is not a niche market these days.
But anyway... I'm oldskool and want to know the hardware. But my guess today's programmers are pretty much separated from it with all the cross development tools available these days.
1. What makes you think the ARM PC's pricing wont jump substantially when its built for a desktop environment?
2. Why change system architectures for the sake of cost savings if you aren't developing specifically for arm based devices?
3. Will the systems be expandable or will you be stuck with a phone desktop hybrid?
4. You might escape intel or AMD in the short term but somone still has to produce and sell the product even if the liscensing is cheap.
A year ago I would say "No way, it's not happening" and start arguing with everyone on that matter.
But after seeing so many benchmarks showing how powerful M1 chips are I might reconsider it. Seems like we are actually getting to the point when x86 platform might get obsolete.
I guess this is the where technology is heading and it seems to be a fine idea after all :)
I was looking for someone who commented M1 chips, since the video came out before the M1. Too bad Apple is relatively expensive.
@@nathanbanks2354 I have M1 MacBook Air. Love it. Great machine.
@@NomadicDmitryiSheep
@@gamingguy9006 Really? Still love my M1 Macbook Air, a powerful machine 💻😉
We need ARM PCs for another reason, competition. Intel needs competition to keep it inovating and it's prices reasonable and we need competition for Intel so that were not faced with a monopoly.
I have to agree ... monopolies of any sort eventually degrade services, products, and processes.
There is competition, Ryzen is going strong. But i'm with you, more competition is always good.
@@bobbyb6053 But there's no competition at the i7 upwards end of the market and that's where Intel can charge what it likes for its products and there's no external pressure to innovate.
Totally agree. Can it be so hard to build a micro ATX motherboard for ARM processors?
The issue isn't making a motherboard.
It's that there's no ARM cores that has enough single threaded performance to compete with viable x86 development processors.
Amdahl's law and all of that...
Perhaps make a CPU with something like 4 big ARM cores able to compete with high end x86 cores and 10-14 small and weak but power efficient ARM cores and put that in a development box.
Then scale that up for servers with four of those pr socket in a four socket configuration with enough memory bandwidth to feed all of those cores.
The four big cores could do the heavy lifting on the non-parallizable parts of the code you need to compile.
The large number of smaller cores would allow you to test out the scalability of your code on ARM.
Yes. The only reason swappable motherboards exists for Intel X86 is because at some point of time, IBM compatibility was required. So, all components needed to be compatible. The BIOS made sure of that.
There is no such thing on ARM side. Every designer is free to build their own designs which are completely incompatible with other ARM computers.
You can't just switch your GPU/RAM between two ARM systems.
Also, because of this, the OS is integrated to the platform. An Android image for one PC will NOT work on another similar PC (even running a similar processor). This is easily seen today as ROMs are incompatible between ARM phones.
@@Luredreier "there's no ARM cores that has enough single threaded performance..." where did you get that ? Look at these:
browser.geekbench.com/v4/cpu/compare/9948464?baseline=9972001
We need Risc-V PCs
No we don't.
Is there no chance to bring back the 1024-core MIPS desktop?
no risc-v for me, i want power9, its a open ISA and is getting somewhat close to threadripper x86 chips
power9 also supports pcie4.0 and is for sale where x86 chips still need to start begin selling
@dennis
Are people still developing AmigaOS? I mean it was pretty cool that they managed to build a fully preemptive multitasking operating system in less than 256 kilobytes of ram.
Great channel Gary. Love your explanations and subjects you talk about.
How bout the Chromebook Plus that Samsung makes with the Cortex A72/A53 processor?
I haven't tested it myself, but I know it already supports Linux apps in the developer channel, as well as Android apps by default. It's only $400 on Amazon and seems like it would check most of the boxes you laid out for a personal development machine, with the portability, design aesthetics, and battery life being added bonuses for everyday use.
If they started working on ARM workstations right now, how many single digits of years would they have before RISC-V workstations come out and instantly outcompete them?
Your question doesn't make sense. Since RISC-V is starting from zero and Arm (or Intel) are decades into making chips, then RISC-V development would need to be multiple times quicker to catch and then "instantly outcompete them." Who is going to make this super fast RISC-V CPU? Where is the investment coming from? What secret does this theoretical RISC-V CPU have that means it can compete with Arm? Will Arm development stop during this lead up time?
I've got 8 raspberry pies and only 3 x86 PC's. Does that count?
For years people developed on x86 and deployed on Sun machines, so cross architecture wasn't a barrier. They started deploying on x86 because their performance improved, and they were cheaper than Sun machines.
There is a serious cost to sacrificing dev/prod parity. Notably for android developers who mostly run x86 machines and write for mostly ARM based machines. Many of them have to use an actual ARM-based android device to check their own code.
The cost of losing the parity is quite large when spread out over the entire industry and certainly did play a part in the rise of x86 server hardware. Parity allows for shorter development pipelines with lower overhead.
@@TaylorIserman Android is primarily programmed in Java. At that point you're more at the mercy of Google's VM than the underlying hardware. When I worked on a mobile project we did have a device lab, but it was more for reproducing specific bug reports than day to day development. We also developed a set of mocks that allowed testing much of the code during the build rather than deployment on a device.
Yeah I’ll concede that java has helped make a lot of app development on android more agnostic. I’ve had issues with an app I wrote that dealt with too many system components to be that way. I still maintain dev/prod parity is a major factor in the prevalence of x86 servers today. To be honest, the main culprit tends to be the OS: like when the intel-compatible version has certain differences or when an OS only exists for some type of hardware (like AIX *grumble grumble*).
@@TaylorIserman I agree OS differences can be a major annoyance, even when it's just two different Linux distributions on x86. For a few years I had to deal with Oracle Unbreakable Linux on servers with Ubuntu on the desktop. Oracle's Linux is forked from Redhat, while Ubuntu is forked from Debian. I swear it was somebody's unethical psychology experiment to see how much swearing people are capable of.
Martin Heermance Yeah I’ve been in a similar situation: not fun! Haha
Most server applications these days run in virtualized environments. It doesn't matter what the underlying hardware architecture is as long as the behavior on the software layer is the same.
The software layer is tied to that underlying hardware, though...
Where do you work that it doesn't matter?
Resource consumption, cost, ease of use, these all matter, maybe not to you directly, but to the chain and process which allows your container to function, so ultimately to you too. Massively.
That only applies if you have programs running in a VM (Java, .NET, JS...). If a program is compiled for x86, no matter if it is running in a container or not, it will not run on ARM and vice versa. You may use an emulation layer of course, but that will be slow and resource intensive.
Emulation is a temporary fix, while the code based is moved to the new ISA. Just the same way apple managed the transition from 68000 to powerPC and from power PC to intel. The latest windows on ARM has intel emulation for the same reason. On the subject of intel they are having trouble manufacturing at 10nm while TSMC and Samsung are at 7nm and testing at 5nm
At 6:24, how are you sure that ARM based computers would cost less than x86 based computers? Is it speculated based on the cost of Rasberry Pi alone?
I am not saying they should costs less, just not $3000 like they do now.
I'm not too familiar with software development, but aren't there emulators that can be used to test or develop? So it wouldn't matter what was used to build?
There are emulators. But that is no sufficient substitute for the real thing if you want to deploy. They also tend to be slow resource hogs.
ARM needs wider compatibility/maintainability as well. I can install pretty much any flavor of Linux on almost any x86-64 platform. I can install Windows 10 on a Core 2 Duo and expect it to keep getting updates 5 years down the line. That's not necessarily the case with ARM.
People like Jon Masters at Red Hat have been doing a lot of work to make sure you can install any flavour of Linux on Arm. The new Fujitsu A64FX boots Red Hat linux out of the box, not because Red Hat has a special flavour for Fujitsu, but because of the boring standards that RH has been pushing. As for the updates, my guess is that Microsoft is committed to supplying patches for the always-connected laptops.
@asdfrozen
OS support isn't really that much of a issue with ARM anymore as it was in the past.
The main issue is application support as the majority of both Windows and Linux applications are still coded on X86.
@@Luredreier Most Linux applications already run on ARM.
Take Debian (just looked it up just now):
99.6 % of official packages build on AMD64.
And on ARM64 it's at 99.4% of all packages.
Can't get much higher than that.
It's better than you think, but definitely could be better and would be if something laptop/desktop grade existed out there.
It's not like it's hard with the current tech. It just needs to actually be available, preferably as systems that you can get setup with just a few easy clicks.
+@@autohmae Who out there limits themselves to apps in the reposetory? Oo
Back when I was using Linux Mint the wast majority of the apps I installed where *not* from the main distro repository at all.
Same on Manjaro.
The official packages are only a *small* portion of the apps people actually use even on Linux.
On Windows it's even more common to use apps that the OS developers have no controll over at all.
There needs to be developer systems out there with proper prosumer performance available on ARM before they can truly take off.
I mean all the big players, Apple, Microsoft, and AMD are looking forward to ARM based systems. We're probably a little ways out from ARM completely taking over the market, but that's where literally all the big players are looking toward. And Qualcomm is set to be the next Intel as far as who will be making the chips that power all our future tech.
Lol. Linus said exactly the opposite. He's very skeptical of ARM architecture and thinks it will sink. Though recently he changed his stance a little towards "we'll see how it ends".
The problem is that ARM only makes sence in a SOC. If you want a PC that has a socketed CPU and is ARM based you would have to give up half of the SOC to a north bridge and a south bridge which dosen't makes sence cost wise, since the main advantage of ARM chips is in low parts count. Maybe with 2 chips? And, can it handle the bandwith of PCIe 2.0 or SATA 3.0?
8:05 I’ve done cross-development(embedded, not server stuff) for almost 6 years and it’s just painful, especially on Linux. Issues like: the target communicating only over serial and the pains with that, not enough space on the target to put a debugger, linux-tools(perf) are not compiled with the kernel and it’s heavily hacked that it’s almost impossible to maintain a build of the linux-tools, firewalls, special flashing steps to open the communication with the target, etc...
Developing on ARM would of definitely made all of that easier - although, later I kind of worked around that issue by using QEMU + qemu-user-binfmt and faking some dependencies.
You would think that Qualcomm would have an incentive to release arm dev boards, considering they want arm to replace as much of x86 as it can. Hell, Microsoft is pushing hard for arm windows 10. Why not release a surface arm pc specifically targeting developers.
I loved the era when someone said Linus, and we all knew it was the guy who actually made something.
No we dont. We need risc v everything if we want to avoid backdoors built into processors. They all built one and dont give af they even say they do as if thats good thing but theres no option really until risc v develops further.
How will you be able to tell if there is, or is not, a back door built into a RISC-V processor?
I personally wont be able to, however im 100% certain there are people who can verify integrity of the device, or component. You disagree?
I absolutely disagree. RISC-V is an open source ISA. That means a company/group/organization can create a RISC-V CPU and sell it. They are under no obligation to release the "source" for that CPU. And even if they did, they can still produce a different version to the one they publish publicly. RISC-V solves none of these problems.
@@GaryExplains I am not sure what point are you trying to make lol. They will trick you into getting different version of the firmware? Is that what you are saying? :))
Theres nothing that runs invisible. It must use electricity, memory and components and produce binaries that if are unknown there is something wrong. But since all is open source and you can verify by code itself idk what the f you talk about them selling you something else lol
there is also the problem that the arm devices are too fragmenting ...
each device requires its specific version of linux with the kernel specifically configured .. and this is boring and difficult ..
That is chaging fast.
Guys, you forgot the actual SOFTWARE.
And actually there's a huge question - is there any way to efficiently port a x86 to ARM? It seems if we speak of emulation (compiled x86 binary on ARM) that it's simply impossible. Am I right?
For Windows on Arm it should just be a re-compile, that is why you can get native Firefox for Windows on Arm.
No, you are wrong.
Is there any way to efficiently port x86 to ARM? No. But both Windows and Linux (as well as most other operating systems) already run on ARM. As does most software. Only games tend not to.
And the difference between a computer and a calculator is the ability of the computer to emulate any other computer, given enough storage. That means that you can emulate an x86 computer on ARM, albeit noticeably slower than it would run natively (mostly due to the complex instruction set of the x86).
If he had said risc V I'd be on board but arm just replaces x86, nothing changes and you will have even less development until apple arm did nothing for a good decade or two
Why not RISC-V servers? If we're going to move away from x86_64, RISC-V is the next step, so why don't we just skip ARM and go streight to RISC-V?
RISC-V isn't what you think it is. I have replied to loads of comments under this video about RISC-V and most people just don't understand it right. I will be making a video about it soon. But to address you specific question, we won't see competitive RISC-V processors for many years yet.
I'm up in arms over this...we need ARM based PCs.
What is really needed is open-source platforms (CPU and Chipsets), not just operating systems.
Far too much hardware has only Windows / OS-X driver support and some OEMs don't want to open up/offer documentation to make drivers available.
I want an ARM-based motherboard that supports an ARM RISC based CPU but where is it? That is what we need and then add RAM, etc. to build an ARM-based PC or server.
Use a virtual machine or emulator. Much cheaper than custom hardware.
And much slower too...
I agree, for most use cases. It's a lot easier to use a runtime like .NET or Java to compile on-the-fly to the local machine's native architecture than it is to make the hardware and its drivers conform to your expectations. And performance is fine most of the time.
@@God-yb2cg Just encourages you to optimise performance.
Guess Apple was listening
I would go ARM since I use linux anyway so all my software would easily port over.
I'd like to see PowerPC desktops make a comeback.
Though what an Arm based PC would need is proper PCI-e expansion slots, expandable ram, and SATA ports.
Why is video editing so fast on my phone and so slow on my 6 core intel?
Are you really still running your PC on a hard drive in 2019?
@@laurelsporter nope.nvme . stopped using spinny drives in 2008. Not the point. ARM chips process video about ten times faster than x86.my question is why.
@@Ballistichydrant they don't, if you're comparing CPU to CPU. A nice x86 CPU *will* be many times faster. Now, if your specific hardware has a DSP or GPGPU that can be used for it, while your x86 doesn't, that could make a noticeable difference, especially if you're talking mainly about rendering, where GPUs can often process video many times faster than CPUs.
@@laurelsporter nope. Please let Gary explain. You don't know what you are talking about.
I'd rather RISC-V, but ARM is def. an improvement on x86.
Why would you prefer a system that doesn't even exist in any real significant terms. Why does RISC-V give you, you personally, than other architectures don't.
Looks like Apple took that seriously
This aged like fine wine
Why do we need ARM (RISC) based PC's over Intel based PC's? What are the benefits? Also, why ARM (RISC) over RISC-V? RISC-V is a better alternative over ARM (RISC).
Risk5 is just for big but not giant companies who can't afford to put fpga into every single board. Arm is more broad
@@BusAlexey Are FPGAs still more expensive than ASIC processors on large production runs?
For me it's cost. The development I do is not tied to the architecture as most languages cross compile with few issues (Go, Node, Python, C#, etc).
Price to performance is what I need and ARM seems to do better with more concurrent workloads.
Start with an ARM ATX motherboard suitable for a workstation or small server, and go from there with all of the PC peripherals, cards, and attachments.
You do realise that this is just a plea for hardware developers to take on additional risk to satisfy software developers desire to avoid it?
Sure, a native workstation for developers is very helpful. That's not going to make ARM faster, though. It's still very far behind in real-world performance.
Doesn't seem to be the case for the Honeycomb. Are you comparing desktop-class x86 to mobile ARMs?
Apple: 🤑 Gary hold my beer.
Was hard to hear audio over the flapping of the arms
Developers tend to use virtual machines, where I work we have many hundreds of these. We can provision them quickly, vary the spec in minutes, and deploy a VM in production on-prem or in the cloud. You can run VMs in Win 10, or VMware workstation or Oraclebox. It should not matter if you develop on an AMD and then run production on x86 - compilers take care of that. CPU architecture should be abstract in most applications.
But you agree that all the VMs are on one architecture, x86.
No, we do not need an ARM desktop, we need RISC-V desktop.
Why?
We need both. And MIPS as well.
I have literally walked out of a bar and found a PC sitting on the curb that I took home with me. That's how ubiquitous X86 based PCs are today. ARM has a long way to go to get to that point.
Did you just... did you just yoink someone's PC?
@@TheGodEmperorOfMankind_ here items left on the curb are trash. If taking that is yoinking then I suppose so?
Need is a strong word. I'm fine with windows on arm as long as it isn't windows on qual¢omm.
Unfortunately, Windows on ARM is only Windows on Qualcomm.
@@lawrencedoliveiro9104 thanks captain.
The whole "ambient computing" concept MS mentioned a few years back, I think we are there now. I cant see the hurdles of working on an arm-based laptop with away from my desk, then using a x64-based multi-core'd beast as a desktop, with all my data synced via the cloud. We are already doing this (to a degree) with smart-phones. Very few of use are processing workloads that are architecture-specific and cannot be accessed remotely. Amazon already have arm-based instances available. Whats the prob again?
Counterpoint to the argument for matching workstation and server chipsets: so long as you're writing for the same OS and using a well-defined programming language, you will still be able to test a functionally equivalent program on the development machine.
I agree! It would be nice if we can get graphics cards and other PCI-e stuff to work with ARM. Then maybe ARM will be all set for desktop use
8cx has all those capabilities built it.
You can't compare the graphics capabilities of the GPU in the 8cx with the capabilities of a dedicated video card.
@@GaryExplains Something like this ruclips.net/video/iaY3Ffor-8c/видео.html , perhaps?
@Richard Vaughn RISC-V is just a definition of an instruction set, it isn't an ARM (the eco-system) alternative.
@Richard Vaughn I think you misunderstand the nature of RISC-V. Only the ISA is open source. Any company, like NVIDIA that adopts it still has to design and develop a superscaler CPU core and then make it. Neither of those tasks are cheap. And their design remains private and closed. Ultimately it doesn't help or harm consumers in anyway, over other ISAs like ARM.
Nice red hat!
Brought to you by IBM
Call me crazy but... I think theres a punch line here :v
@@Nsfwstar yea should have been nice: red hat hat
Problem is performance, ARM's isn't that good compared to x86 in performance and it looked like ARM's was going to make inroads into the PC market with how slow x86 was developing, mostly thanks to Intel but now ARM's development is slowly down whiles AMD with Ryzen is shaking up the x86 market, the gap was wide to begin with but now it could get wider when we look at what AMD is doing with Ryzen.
Now don't get me wrong, I'm all for ARM's on the PC for added competion but software compatablity is the number one issue, it's all mostly designed around x86 and that wont be easy to change, performance is the second issue and I know they can scale things up but then software development gets more complecated.
So if AMD didn't wake up the x86 market, there was a real chance of ARM's making real inroads intos the PC market because performance leap per year was faster with ARM's but Ryzen seems to be changing that as we are seeing 8 cores at mid level prices and likely 16 cores at mid level pricing over the next 2 or so years, if it was up to Intel, we would still be on 4 cores and that left a big opening for ARM's to take advantage off but all thats changed now.
Does it really make a difference? Cross-compiling isn't exactly very complicated. One could even simply compile on the arm-based server itself.
What you are suggesting will only occur if there is an explosion of mITX "carrier" boards supporting products like the RaspberryPi "Compute" module. To this date only the TuringPi vaguely fits this form factor.
I think "will only occur" is a little too strong. That is one possible path, but by no means the only way.
I want an AVR-based PC.
... And by "want", I think I mean to say, "probably don't want"
@Lassi Kinnunen It was on Hackaday a few years ago. The article is called the "Building the worst Linux PC ever". The effective clock speed was 6.5 kHz.
Isn't it simpler to just watch Linus' video than to listen to a re-tell?
Linus didn't make a video about this. LOL.
Gary mumbles. Plenty of stuff to read im sure.
If you want an ARM-based laptop right now: buy a Chromebook and install desktop Linux.
Yes, except most Arm based Chromebooks have low specifications. I have several videos about Chromebooks and Linux on Chromebooks on this channel.
@@GaryExplains I have to say, I'm not sure what most developers build, but everything I build runs just fine on low specs. :-) It's actually becoming easier with Docker, etc. microservices really are smaller.
There is also windows based snapdragon laptop
@@ianfahrizal85 Ohh, but then you end up paying for a Windows license. ;-)
Also I think there are no Linux drivers for the graphics chip ???
@@autohmae why I do need to paying license? Laptop come with windows 10
For the new year I bought a quad core computer, Windows 10, Intel 64 bit chip, 4G ram, 32 G flash stuck my SSD drive inside and got it delivered for about $250 US (not including the Samsung SSD). The company that makes this computer (the size of an Apple TV box), with power supply of course was Azulle and it is made in the USA. Not ARM but everything else you would want.
Where I work developers develop Java applications on Windows desktops but the application is deployed in the production environment on Linux-based servers. The Java JVM allows us to successfully use two different operating systems for development and deployment. Would this also work at the hardware level between using x86 or ARM based hardware?
What if Samsung puts Windows in a device like the new fold and an Exynos CPU
BOOM continiuty on steroids
Windows is fucked up in so many ways. We need to all adapt to Linux and that can run on almost any architecture.
@@nicholasjohnson3542 One thing great about [insert any paid or exclusive OS] is their customer support.
Because you've paid some amount of money, you're entitled to get some support. ( No need to too often disturb your computer wizard friend ).
@@rommysoeli only losers need support when there's a thing called the internet. Also you're naive if you think that the money you paid actually gets you actual support.
@@nicholasjohnson3542 No wonder Linux has the biggest desktop OS market share right?
@@TheZaman_ lol it doesn't have the highest, I wish. If you are making fun of me, go and find some people who are actually happy with Microsoft support or something. And why is stackoverflow and sibling websites a thing if people don't search the internet for their problems and ask for help? Better to get support by asking your peers who are in the same situation, rather than unknkwledgeable support people reading from a manual.
4 vids in 4 days . nice :)
Thx!
Easy to say when you don't need to PAY to develop an OS, device drivers, and app. LOL!
what do you mean?
@@burnstick1380 Supporting multiple hardware architectures is a little bit more difficult than just re-compiling your code.
@@Entrophius ahhh yeah thats a fair point but well theoretically you can do it with virtual translation but that loses performance oc
Could containers render the underlying hardware architecture irrelevant? You can build on a x86 machine and deploy it to a container running on a Arm server
No, because containers only solve the environment/deployment issues, not the testing/validating/etc issue.
Does this problem exist if we're developing software that compiles to an intermediate language running in a virtual machine (eg. Java or .net). A lot of this now runs in an abstracted way in the cloud these days anyway ("serverless" architecture - eg. Azure functions), so we don't even know what hardware it's running on.
wow Linus Tech Tips is so wise :)
You do know that Linus Torvalds (creator of linux) and Linus Sebastian (squeaky RUclips entertainer) are different people?
Potato-based servers for potato-based development environments
^ This.
It is the same reason why MS Windows and MS Office is still so prolific; it is what people have either at work or at home, so the people tend to keep what they are already used to using elsewhere. I believe Windows Server should not be a thing, but it is, only because people who are familiar with Windows on the desktop are more easily able to transition to the same environment on servers, so that is what they go with. It was not that long ago we had servers running on MIPS, SPARC, and Alpha architectures that all fell out of use because people did not want to split their time learning to code and provide support for multiple types of systems when one standard could do the job. Even if ARM desktops and servers become available, shortly afterwards, either the ARM or the Intel based architecture would would become the sole system in use. Mobile devices had many types of architectures, but they have all been whittled down to two, and I would not doubt that within a decade from now, one of those two will be gone.
getting ARM scalable to x86 is kinda tricky. Things (SATA/PCIe/etc.) are actually build based on x86, so you need to either have pain adapting or taking time building a brand new same kind of things with ARM architecture
All those things already exist on Arm platforms.
The controllers for SATA and PCIe and other busses are usually MIPS, not x86.
I'm expecting some kinf of ARM mobo, ATX format, with all the connections and features one can find on an Intel/AMD motherboard. Only a few of them, impossible to find, very expensive, and desperately slow...
Yes they should take some RISCs
😂😂🙄
Arm is risc.
Waiting for the Snapdragon 8CX desperately!
nice redhat
linux
Maybe I missed it, but I'm not getting what the value of using ARM chips in cloud based servers has over using Intel/AMD.
Energy efficiency
I'm curious if the performance gains on ARM are merely due to the 7 nm aspect. If so, x64 should be within touching distance to that in 18 months. It's not a "Power PC roadmap" event, more like "wait until next year and the performance will be equal" hiatus.
No if it was about the process node (like 7nm) then you coudl take a Pentium II processor from the 90s, make it on 7nm and you have a high performance processor. It is mainly to do with the microarchitecture. The process node helps with power efficiency, but it isn't the whole story there either.
Linus: "We need ARM based PCs!"
Apple: "Here you are. Thank us later."
@@vardekpetrovic9716 I read recently that Apple are open to the idea of Windows 10 booting natively on the M1, and it was really up to Microsoft to get on with it. Assuming that's true, it should be the same for booting Linux. (Though obviously this is a huge simplification.)
I am writing this from an ARM Chromebook running Arch Linux.
On a web based, x86 server run by google.
ARM isn't anything special... there is no real reason for ARM vs Intel.
Words like competition or monopoly or innovation meaning anything to you?
@@GaryExplains Common sense? People are always going to think SHORT term.
Exactly, Intel has 30+ years of instruction level profiling to make applications faster not just synthetic benchmarks. I will take Intel over any ARM.
Yep. Totally agree with you. To name a few, Java an golang and JavaScript try to make it easier to achieve that but developers dont all get that vision. I think that vendor lock-in makes money and thats the biggest driver for not adopting these basic principles
I've been saying this since Cortex A72. And not some generic dev box. Just make the silicon socketable. There are already people out there making embedded boards. IPC parity (or damn close enough to it) is here. It's freakin insane that this hardware is being held back from mainstream computing users.