Chris, it is not easy to mediate such a low level discussion "for the masses" and you did a fantastic job pulling apart the detail and rephrasing as needed. Thank you for this content!
It'd be interesting to see if they can stay true to RISC or if they will fall from the temptation ARM fell for, starting to add more and more instructions/features
@@backacheache Would it matter if they added more. Isn't the appeal with RISC-V that you can choose which functions you want to embed into your system? Risc-v basically give you the option to configure your own core, and thats what's so great with it
@@haraldbryn8319 whats so great with it is the small instruction set which makes it multiple times as efficient than than x86 and x64. You can always add more but you cant go lower than the spec, which is what the commenter you replied to was talking about
@@backacheache the base ISA is frozen and the base standard can't be extended. But if some company wants to make a chip with more instructions they can add them themselves.
It took ARM some 15 years I think to go from embedded devices to bigger devices ? So it will take RISC-V at least 10 years. But they haven't even started yet getting into embedded, this is now slowly happening. PS ARM is also RISC, just not RISC-V
I feel that this also has a significant impact on career choice, when looking at software to hardware development there has always been the limitation that these few select vendors are the only ones with access to develop. An architecture like this enables freedom in your career as well which is very appealing.
That also limits exclusivity and increases numbers of capable user, then supply meeting demand and lowering your wage when you get where you plan to now.
I was just thinking today, game graphics would be a lot farther along if multiple open source projects were in use. A lot of proprietary and private software in use... I’m a mobile app developer and so many tools are open source and it keeps everything moving forward.
@@blueversace4447 this is exactly the problem, software should be FREE not open source. Open source is a schemey manipulation of the original values initiated by Richard Stallman and the FSF
@@Ryan-xq3kl If it was free, then how do you motivate anyone to create it? Those guys are academics, they are paid on universities from taxes, it's in their interest to have a platform and tools like that, but there are millions of other people that without profit wouldn't even touch it
@@singularonaut Yeah, don't think it works like that, you could at most offer support for it, but even that without licensing might be lacking for financing an organisation
One special advantage i see in RISC-V being free, is its flexibilty and reduced complexity. Unlike ARM, intel and AMD which inevitably degenerated to bloated CISCs by the time, RISC-V does not need to stick to old proprietary standards like the big companies do by adding closed extensions periodically just to keep alive their patents, power and influence on the whole IT industry. Similar to revolutionary Software like Linux the ISA of RISC-V can develop at fast pace by implementing new and better ideas and replacing outdated ones. I think IT was at a much advanced level now if big companies would not slow down future developments and innovations with their proprietary standards. Who knows how today's IT landscape would have looked like, if intel did not silence transmeta for example. PS: Flexibilty and modularity are keypoints to keep up with the fast pace of IT innovations. And in this regard i furthermore would like to see RISC-V systems paired with FPGA's. Guess what Intel and AMD did to keep up...
@@ShaneFagan There are already tiny processors that use much less power. The ESP32 and many ULP cores run on almost nothing. They are not very powerful though so a spec like 5ghz means NOTHING without knowing the number of operations executed per second. Using group IV semiconductors you can make 400ghz transistors (Used in RF and lasers, networking etc.) theoretically you could build adders and all the pieces of an ALU out of those and have a really fast processing core. A regular processor has many ALUs stringed together to make them useful though. A small 1 watt processor running at 5ghz still won't be anywhere near as powerful as a regular processor if it doesn't have the whole execution pipeline. Michigan U made the world smallest computer that runs on uW of power.
Unlikely that would mean scraping a large part of the RPi ecosystem. Unless there is some real cheap and real powerful chips that put ARM to shame such a scenario would be impossible IMHO(in the short term atleast, say 4-5 years...) There is a sort of a competitor called Beagle V with RPi like features in works. Let's see what it can do. Edit: Beagle also has specialized AI stuff... Once I get my hands on it, it will be alot of fun.
RISC-V will only reach that level of development/market or mind-share/software support in 5-10 years, so no. Maybe they will release a couple of RISC-V boards on the side, but even then I'd expect them to live side-by-side.
In my view the appeal of the Raspberry Pi’s is the ease of getting started due to how easy and abundant software and libraries are for the ARM SOC and the flavor of Debian most Pi owners will be using. A fair amount of Pi owners just use it as a cheap HTPC, or print server, or as a DNS server for stuff like Pi-Hole and will never utilize or tinker around with it. The kind of people that possess just a basic knowledge of navigating the Linux terminal and GUI. RISC-V is not there yet in terms of that kind of support and ease of use.
I'm very excited about esperanto chips. Hope they end up with consumer products also eventually :) If they are more powerful than CUDA cores which I imagine they are. I've only dabled a bit in machine learning and this is huge.
One item of interest. I was going to place an order for a RISC-v developer board here in Canada. I was informed that I would not be able to place an order without a technology export license from the US department of commerce. So much for developing my stealth bathtub technology!😎 I am hoping that this changes, I used to have an NS32032 developer board many many years ago. It was great fun having a device that was so one off item. Unfortunately, the NS line didn't end up becoming a retail product. I am now waiting for a high-performance SBC that I can cluster.
The whole point that RISC-V is open-source hardware and software. Thus beyond the reach of geopolitical shenanigans. Since it is in the public domain. So get your dev kits elsewhere. The world is a much bigger place.
Here's a video suggestion - overview of the number of processors and co-processors in everyday tech. Start with PCs/laptops, but also look at TVs, consoles, etc
Great video! Thank you for putting this together. This movement is obviously going to have a huge impact on the future and it’s very smart of you to put together this panel at this point. Investors who think Invidia is going to dominate the world once it buys ARM should take careful note. I would not be at all surprised if we see Apple move from ARM to this open standard in the next few years.
RISC-V is NOT an architecture, it is an ISA. You can use different architectures to implement that ISA, you can make a pipelined or non pipelined architecture. You can make a superscalar or a scalar architecture. You can make an out of order or in-order architecture. All RISC-V is a specification of what instructions your processor should consume, but it makes no assumptions about the architecture of the chip.
ISA = instruction set architecture There are many levels of abstraction in computing, and architecture has a different meaning in all of the levels. RISC-V IS an architecture, just a different kind of one.
In addition to all general efficiency factors, doing floating-point on short words has a mathematical advantage more than proportional to the shortening. AI often needs only short (8 to 12 bit) words.
This seems like something the Raspberry Pi foundation could do. Make their own RISCV chips. They already are making their own ARM chip for the new Raspberry Pi Pico.
TLDR: summary: x86 clock speeds havent really budged over the last 10 years and the only improvements are in power consumption and heat dissipation with smaller transistors. Therefore the future of performance is in going to the past and using a more efficient ISA along with a top down redesign of the entire PC ecosystem architecture which will be RISC based and open source. Open source in this context does not mean "free" as in free code to download and run yourself. Making a processor still costs a lot of money and the only difference with RISC-V is that chip designers don't have to pay a license to use the RISC standard. An important factor here is that the standard architecture for PCs is over 40 years old and derived from the old IBM PC Compatible standard created 40 years ago. This open standard is the basis for most PCs using x86 CPUs to this day and controls how all devices connected to the motherboard communicate with each other. Apple of course has always had its own proprietary standards as it designs and builds its own hardware and has used RISC architectures in many of its products. So in moving to a RISC-V architecture potentially as an alternative to x86 on the desktop, you are talking about making a new standard to replace the derivatives of the old IBM PC compatible architecture. This open standard plug and play standard and other proprietary standards, such as chipsets, that are controlled by CPU manufacturers would be replaced by a new open standard covering all aspects of the entire hardware ecosystem for PCs. And by doing this, it allows for radically new approaches to the design of PC architectures that are based on more modern technologies and hardware based on the RISC-V standard. Of course all of that will take some time to define the new standards and get buy in by the various hardware vendors and OS vendors. One example of what could be done is creating new PCs where everything is soldered onto the same motherboard as coprocessors which all share memory and data between each other over a new bus architecture based around non volatile fast memory and solid state onboard storage. Of course most of these kinds of architectures would start out on the server side because most of those kinds of custom components are only useful for business applications vs personal computing at home. And I see most of the impact of RISC-V to x86 potentially being in the server market (just as RISC dominated business and server computing 40 years ago).
I know it's not that trivial, yet, to gear up for manufacturing custom ICs, but imagine writing a program in Assembly and running it on a CPU made fot that specific purpose. Your software on a chip processes would be going straight to output.
Im Still waiting 3d chip printers... So only buy a silicon waffer and print any chip I need... We could add chip to any home devices and change the cpu of the smartphone like changing the microsd
I don't think it'll be feasible any time soon, unfortunately. Because of the size of process nodes, very high precision machines working in extremely sterile environments are required to make processor chips. I don't think we'll be able to engineer out those constraints soon.
@@tri-a7655 I similarly designed and implemented a RISC-V processor for a college Computer Architecture class. I think its ISA is fairly straightforward and is very clean. It has some cool tricks that are designed to make the hardware used for decoding instructions really small (especially compared to even a super basic x86 one). Programming one was relatively easy. I loved working on fetching the instructions, decoding them, then working on the ALU to actually execute these instructions, it just was super cool. There were some more hairy parts later on, but I thought it was great. I built the simplest possible option in the ISA, an RV32I processor. There are extensions in the ISA to add instructions for more processor functions. Mine could basically do simple integer addition/subtraction and any compiled code would specifically support that, e.g. multiplication would become a series of addition steps. But, you can add multiplication and division, floating point numbers, 64 (or 128) bit support, and so much more right on the processor. You can even add custom instructions. Common instructions are all very nicely specified and the ISA lends itself to easily adding and removing these extensions. Making custom designs with only the necessary features for the use case is really simple as a result. Based on this experience, I believe a company with some chip design resources and specific needs should be very interested in RISC-V, especially if they would otherwise be locked down to available ARM, x86, or other preexisting chips. Developing your own ARM chips is simply too expensive due to the licensing, I don't even know how you could get x86/64 licenses. If you want any option beyond what others provide, you would otherwise be out of luck. As such, I think RISC-V is filling a great need and could be a serious competitor in the future in many growing sectors.
@@tri-a7655 Yeah RISC-V was very enjoyable to use and I'd say I enjoyed using it more than x86. I just have a hard time doing lower-level programming but RISC-V I found myself enjoying it even though I wasn't that great at it.
@@jokinglimitreached1503 Sure, but they have tons of money, properties and most importantly patents invested in x86 and ARM, which will all be made pretty much worthless if RISC-V succeeds. Also intels and amds duopoly for x86 and arms monopoly would be shattered. Good for everyone except them.
@@elukok Hmm. I think they already capitalized on their patents for decades, they got their investment back. x86 is just slowly becoming less relevant. If they adopt RISC-V, it will be good for them, I think it will even save their business
@@jokinglimitreached1503 Yeah, but they wont be protected by patents so the competition will be much higher. No more milking the customers as it is now. Of course I agree and hope that implementing RISC-V will be the only way forward for them eventually.
Still, the next big thing in consumer computers is ARM because of Apple M1. But I have a serious question. Is it even possible now to create a general purpose CPU on Risc-V to be close in performance to what typical CPU offer? I know it is great to make, specialized co-processors and stuff, but that long that anyone do not take this approach and architecture and do something for mass consumers it will stay in the limbo of co-processors.
As a layperson, I don't understand what you mean by open-source computing architecture. Does it mean for instance if i buy a computer using such a chip I will need software specifically compiled to run on it? Kind of how windows 64 bit software don't run on older windows 32 machines.
Yes, you'll need specifically compiled software, but recently there has been a lot of development in translation programs which allow running software from other architectures without much performance loss. However, that has nothing to do with the open source part. It being open source just means that everyone designing hardware has access to the basic feature set of RISC-V and they don't have to pay anyone for it. This facilitates faster and cheaper development and helps innovation.
RISC-V has only gained prominence over the last few years. iPhone came out before RISC-V got on the radar. Apple also bought a very expensive license from ARM so they are free to customize as they see fit while most vendors buy a finished core design.
You know, the 1st time AMD could really compete with Intel was in 1996 with their K5 CPU - which used a RISC processor and a x86 decoding front-end that was faster clock-for-clock compared to the [original] Pentium...
How This business Model is any difference than ARM? The People who do custom cores still end up selling to other vendors. Very few would need a super custom architecture, most just want to build a Phone or something ...
It's targeting a different slice of the market so I doubt it will be a direct competitor anytime soon. I could see this blowing up with new companies innovating in the Internet of Things (Iot) market
@@grossly820 In the corporate world you can side-step the issue by having legacy code run on remote (cloud) machines using technologies such as "citrix"
The RISC-V ISA documentation is a case study to teach students how NOT to write documentation. Pages and pages of details and never a clear picture of the general idea. Typically case of head in people for who of course this big picture is obvious and what they are proud of is this little detail, but they don't realize a technical documentation is intended to people with the totally opposite perspective. This result in a case of the open source fallacy, "open" is not really "open" if the documentation is badly done. Seriously make some effort there !!! For exemple on how to write great ISA documentation, check out old Motorola 68HC11 programmer's guides. Those were masterpieces of the field. Both super detailed and super clear.
It's designed to be extended. If I remember right, official extensions and custom ones don't overlap, so you can add your own instructions without asking permission.
It’s all great until you need a 10 billion dollar FAB to materialize anything, and at that point you’re completing against every other technology. Seems to be competitive on any level out spending the other guys is an absolute necessity.
Is Intel sleeping or are they going to leverage these developments for future products? RISC processors have been around for a long time. So it is not like Intel is being taken by surprise.
interesting but this thing will not easily going to be useful compute device for users any soon and the first ever will be really slow and backwards. but no brainer on storage, network devices, cameras, etc. though I could be wrong, if something as powerful as China starts using this for consumer and government
What?! This doesn't have much to do with politics. Mostly allowing for more competition in stale market resulting from large companies stifling innovation.
Try moving in more frantic/spastic ways as you talk... it makes it super easy to avoid psychosis while trying to watch you speak (sarcasm; had to turn off the video not 5 sec into your intro)
Chris, it is not easy to mediate such a low level discussion "for the masses" and you did a fantastic job pulling apart the detail and rephrasing as needed. Thank you for this content!
The masses are revolting.
*\(^___^)/* 🐕
*\(^___^)/* 🦒
RISC-V have full potential to become No.1 RISC ISA, But take some more time to complete with ARM.
It'd be interesting to see if they can stay true to RISC or if they will fall from the temptation ARM fell for, starting to add more and more instructions/features
@@backacheache Would it matter if they added more. Isn't the appeal with RISC-V that you can choose which functions you want to embed into your system? Risc-v basically give you the option to configure your own core, and thats what's so great with it
@@haraldbryn8319 whats so great with it is the small instruction set which makes it multiple times as efficient than than x86 and x64. You can always add more but you cant go lower than the spec, which is what the commenter you replied to was talking about
@@backacheache the base ISA is frozen and the base standard can't be extended. But if some company wants to make a chip with more instructions they can add them themselves.
It took ARM some 15 years I think to go from embedded devices to bigger devices ? So it will take RISC-V at least 10 years.
But they haven't even started yet getting into embedded, this is now slowly happening.
PS ARM is also RISC, just not RISC-V
I feel that this also has a significant impact on career choice, when looking at software to hardware development there has always been the limitation that these few select vendors are the only ones with access to develop. An architecture like this enables freedom in your career as well which is very appealing.
That also limits exclusivity and increases numbers of capable user, then supply meeting demand and lowering your wage when you get where you plan to now.
*\(^___^)/* 🐒
Open-source always wins on long run)
I was just thinking today, game graphics would be a lot farther along if multiple open source projects were in use. A lot of proprietary and private software in use... I’m a mobile app developer and so many tools are open source and it keeps everything moving forward.
@@blueversace4447 this is exactly the problem, software should be FREE not open source. Open source is a schemey manipulation of the original values initiated by Richard Stallman and the FSF
@@Ryan-xq3kl If it was free, then how do you motivate anyone to create it? Those guys are academics, they are paid on universities from taxes, it's in their interest to have a platform and tools like that, but there are millions of other people that without profit wouldn't even touch it
@@deedoodeedoo6382 You will create open source for free, but maintains it for fee 😏
@@singularonaut Yeah, don't think it works like that, you could at most offer support for it, but even that without licensing might be lacking for financing an organisation
One special advantage i see in RISC-V being free, is its flexibilty and reduced complexity. Unlike ARM, intel and AMD which inevitably degenerated to bloated CISCs by the time, RISC-V does not need to stick to old proprietary standards like the big companies do by adding closed extensions periodically just to keep alive their patents, power and influence on the whole IT industry.
Similar to revolutionary Software like Linux the ISA of RISC-V can develop at fast pace by implementing new and better ideas and replacing outdated ones. I think IT was at a much advanced level now if big companies would not slow down future developments and innovations with their proprietary standards. Who knows how today's IT landscape would have looked like, if intel did not silence transmeta for example.
PS: Flexibilty and modularity are keypoints to keep up with the fast pace of IT innovations. And in this regard i furthermore would like to see RISC-V systems paired with FPGA's. Guess what Intel and AMD did to keep up...
*\(^___^)/* 🐕
*\(^___^)/* 🐒
The competition for RISC-V compliant core IP would be a very interesting topic.
1000 cores, 20w -you have my attention.
You have to remember highly specialized chips. You won't be running any desktop on it or even regular server software ever.
@@autohmae They have linux desktop running on risc-v chips already, but you are right the chip in question is highly specialized.
@@Waitwhat469 I know but those AI chips are for running AI, they are like co-processors, or like GPUs, they are not for running an OS.
@@ShaneFagan There are already tiny processors that use much less power. The ESP32 and many ULP cores run on almost nothing. They are not very powerful though so a spec like 5ghz means NOTHING without knowing the number of operations executed per second.
Using group IV semiconductors you can make 400ghz transistors (Used in RF and lasers, networking etc.) theoretically you could build adders and all the pieces of an ALU out of those and have a really fast processing core. A regular processor has many ALUs stringed together to make them useful though. A small 1 watt processor running at 5ghz still won't be anywhere near as powerful as a regular processor if it doesn't have the whole execution pipeline. Michigan U made the world smallest computer that runs on uW of power.
@@autohmae very true
I hope my next Raspberry Pi SBC will be high performance RISC-V based.
Unlikely that would mean scraping a large part of the RPi ecosystem. Unless there is some real cheap and real powerful chips that put ARM to shame such a scenario would be impossible IMHO(in the short term atleast, say 4-5 years...)
There is a sort of a competitor called Beagle V with RPi like features in works. Let's see what it can do.
Edit: Beagle also has specialized AI stuff... Once I get my hands on it, it will be alot of fun.
RISC-V will only reach that level of development/market or mind-share/software support in 5-10 years, so no. Maybe they will release a couple of RISC-V boards on the side, but even then I'd expect them to live side-by-side.
In my view the appeal of the Raspberry Pi’s is the ease of getting started due to how easy and abundant software and libraries are for the ARM SOC and the flavor of Debian most Pi owners will be using.
A fair amount of Pi owners just use it as a cheap HTPC, or print server, or as a DNS server for stuff like Pi-Hole and will never utilize or tinker around with it. The kind of people that possess just a basic knowledge of navigating the Linux terminal and GUI.
RISC-V is not there yet in terms of that kind of support and ease of use.
damn dude, your videos are always the best
It's a girl and she's
engadget! ))
Very well done, Chris. Thanks for convening and moderating this roundtable -- fascinating.
*\(^___^)/* 🐒
I'm very excited about esperanto chips. Hope they end up with consumer products also eventually :) If they are more powerful than CUDA cores which I imagine they are. I've only dabled a bit in machine learning and this is huge.
*\(^___^)/* :😺
i started watching engadget a few weeks ago, and is already one of my favorite tech channel, your videos are great.
this video was a good surprise for me. I was expecting something more similar to their articles, which I don`t dig much, but this video was great
IT professional here, all way AMD but looking forward to RISC-V and ARM. ;) Have a great day!
Arm got eaten. Risc is the only real way now.
One item of interest. I was going to place an order for a RISC-v developer board here in Canada. I was informed that I would not be able to place an order without a technology export license from the US department of commerce. So much for developing my stealth bathtub technology!😎 I am hoping that this changes, I used to have an NS32032 developer board many many years ago. It was great fun having a device that was so one off item. Unfortunately, the NS line didn't end up becoming a retail product. I am now waiting for a high-performance SBC that I can cluster.
It takes time to build up an ecosystem, I'm certain in time this will be solved.
The whole point that RISC-V is open-source hardware and software. Thus beyond the reach of geopolitical shenanigans. Since it is in the public domain. So get your dev kits elsewhere. The world is a much bigger place.
*\(^___^)/* 🖐
Between RISC-V and OpenPOWER, I'm really hopeful about open source ISA's going forward! This feels very much like where Linux was in the late 1990's.
very true
Here's a video suggestion - overview of the number of processors and co-processors in everyday tech. Start with PCs/laptops, but also look at TVs, consoles, etc
*\(^___^)/* 🐘
Excellent interviewing skills. You asked the questions that I wanted to hear the answers to.
*\(^___^)/* 🐹
Nicely done. Thanks for all the participants too.
*\(^___^)/* 🐹
Great video! Thank you for putting this together. This movement is obviously going to have a huge impact on the future and it’s very smart of you to put together this panel at this point. Investors who think Invidia is going to dominate the world once it buys ARM should take careful note. I would not be at all surprised if we see Apple move from ARM to this open standard in the next few years.
I didn't need this talk for me. I needed this so bad for explaining risc-v and why it's awesome.
*\(^___^)/* 🐧
You guys are awesome - RISC-V is a gamechanger
*\(^___^)/* 🐨
RISC-V is NOT an architecture, it is an ISA.
You can use different architectures to implement that ISA, you can make a pipelined or non pipelined architecture. You can make a superscalar or a scalar architecture. You can make an out of order or in-order architecture. All RISC-V is a specification of what instructions your processor should consume, but it makes no assumptions about the architecture of the chip.
ISA = instruction set architecture
There are many levels of abstraction in computing, and architecture has a different meaning in all of the levels. RISC-V IS an architecture, just a different kind of one.
@@simjans7633 By that logic, English is a dictionary.
@@simjans7633i i believe he meant risc v isn't an architecture in the physical sense.
The academic beginning of RISC V sounds a lot like a CPU parallel of Tanenbaum's Minix OS.
In addition to all general efficiency factors, doing floating-point on short words has a mathematical advantage more than proportional to the shortening. AI often needs only short (8 to 12 bit) words.
The Future of Tech is ARM & RISC-V SoC's/CPU's/APU's as well as Micro/Exo-Kernels!!!!
really interesting stuff. thank you for the excellent panel and discussion
I wish I knew Verilog -- I feel like it'd be really interesting to try and implement a RISC-V compliant core on an FPGA
*\(^___^)/* 🐘
*\(^___^)/* 🐧
Risc V is the future \m/
Can we have transcript of videos like these meetings....
This makes Western Digital a good stock to Invest in this upcoming quarter :) QC
Does Western Digital have a big exposure to risc-v?
Qualcomm bought nuvia so they have the best cpu engineers now.
This seems like something the Raspberry Pi foundation could do. Make their own RISCV chips. They already are making their own ARM chip for the new Raspberry Pi Pico.
*\(^___^)/* 🐘
Really interesting and cool talk, thanks!
TLDR: summary: x86 clock speeds havent really budged over the last 10 years and the only improvements are in power consumption and heat dissipation with smaller transistors. Therefore the future of performance is in going to the past and using a more efficient ISA along with a top down redesign of the entire PC ecosystem architecture which will be RISC based and open source.
Open source in this context does not mean "free" as in free code to download and run yourself. Making a processor still costs a lot of money and the only difference with RISC-V is that chip designers don't have to pay a license to use the RISC standard.
An important factor here is that the standard architecture for PCs is over 40 years old and derived from the old IBM PC Compatible standard created 40 years ago. This open standard is the basis for most PCs using x86 CPUs to this day and controls how all devices connected to the motherboard communicate with each other. Apple of course has always had its own proprietary standards as it designs and builds its own hardware and has used RISC architectures in many of its products.
So in moving to a RISC-V architecture potentially as an alternative to x86 on the desktop, you are talking about making a new standard to replace the derivatives of the old IBM PC compatible architecture. This open standard plug and play standard and other proprietary standards, such as chipsets, that are controlled by CPU manufacturers would be replaced by a new open standard covering all aspects of the entire hardware ecosystem for PCs. And by doing this, it allows for radically new approaches to the design of PC architectures that are based on more modern technologies and hardware based on the RISC-V standard. Of course all of that will take some time to define the new standards and get buy in by the various hardware vendors and OS vendors.
One example of what could be done is creating new PCs where everything is soldered onto the same motherboard as coprocessors which all share memory and data between each other over a new bus architecture based around non volatile fast memory and solid state onboard storage. Of course most of these kinds of architectures would start out on the server side because most of those kinds of custom components are only useful for business applications vs personal computing at home. And I see most of the impact of RISC-V to x86 potentially being in the server market (just as RISC dominated business and server computing 40 years ago).
RISC-V is wonderful, but i wouldn't describe it as new. It was launched in 2010.
1000 cores CPU only using 20W of power.. Nice..
Then thats not cpu.. Its an optimized GPU lol
The only thing I can think of is that Intel will never gain the kind of dominance it once enjoyed up until 2019.
As I understand it, Intel’s cutting edge still reigns supreme, it’s just the yields that are lagging behind.
Thank you.
I know it's not that trivial, yet, to gear up for manufacturing custom ICs, but imagine writing a program in Assembly and running it on a CPU made fot that specific purpose. Your software on a chip processes would be going straight to output.
if this thing release i can make my own ryzen 5950x?
Im Still waiting 3d chip printers... So only buy a silicon waffer and print any chip I need... We could add chip to any home devices and change the cpu of the smartphone like changing the microsd
I don't think it'll be feasible any time soon, unfortunately. Because of the size of process nodes, very high precision machines working in extremely sterile environments are required to make processor chips. I don't think we'll be able to engineer out those constraints soon.
What could be done is custom ordered chips, though! Design a chip and get it shipped to you for (relatively) cheap.
@@simjans7633 I thought ou have to buy those custome Silicon in the tens or hundreds of thousands to get low cost per chip.
Is there a place to print my RISC-V designs? I'm very curious to know!
Incredible video!
This isn’t new new I had to use RISC-V for my system architecture class in college. It was really hard for me to grasp it at first 😪
@@tri-a7655 I similarly designed and implemented a RISC-V processor for a college Computer Architecture class. I think its ISA is fairly straightforward and is very clean. It has some cool tricks that are designed to make the hardware used for decoding instructions really small (especially compared to even a super basic x86 one). Programming one was relatively easy. I loved working on fetching the instructions, decoding them, then working on the ALU to actually execute these instructions, it just was super cool. There were some more hairy parts later on, but I thought it was great.
I built the simplest possible option in the ISA, an RV32I processor. There are extensions in the ISA to add instructions for more processor functions. Mine could basically do simple integer addition/subtraction and any compiled code would specifically support that, e.g. multiplication would become a series of addition steps. But, you can add multiplication and division, floating point numbers, 64 (or 128) bit support, and so much more right on the processor. You can even add custom instructions. Common instructions are all very nicely specified and the ISA lends itself to easily adding and removing these extensions. Making custom designs with only the necessary features for the use case is really simple as a result.
Based on this experience, I believe a company with some chip design resources and specific needs should be very interested in RISC-V, especially if they would otherwise be locked down to available ARM, x86, or other preexisting chips. Developing your own ARM chips is simply too expensive due to the licensing, I don't even know how you could get x86/64 licenses. If you want any option beyond what others provide, you would otherwise be out of luck. As such, I think RISC-V is filling a great need and could be a serious competitor in the future in many growing sectors.
@@tri-a7655 Yeah RISC-V was very enjoyable to use and I'd say I enjoyed using it more than x86. I just have a hard time doing lower-level programming but RISC-V I found myself enjoying it even though I wasn't that great at it.
RISCV is much easier than x86 and even ARM from everything I’ve tried so far. Coding in x86 assembly is a nightmare
16:10 Clip this, we'll need it in the future when this dominates and people ask why it did.
@Kharthigeyan Satyamurthy ryzen now is around 100 watts 16 core ?
so, basically i can get some of these? and either have them made for whatever i want them for? or what?
NVIDIA purchase of ARM will accelerate RISC-V IHMO. 1000 cores on a single chip and 20 Watts?!?! Wow Esperanto
Wow 1000+ RISC-V cores 12:00
This is good for all.
Apart from Intel, Amd and Arm.
@@elukok They can use RISC-V too. Just like Nvidia uses it.
@@jokinglimitreached1503 Sure, but they have tons of money, properties and most importantly patents invested in x86 and ARM, which will all be made pretty much worthless if RISC-V succeeds.
Also intels and amds duopoly for x86 and arms monopoly would be shattered.
Good for everyone except them.
@@elukok Hmm. I think they already capitalized on their patents for decades, they got their investment back. x86 is just slowly becoming less relevant. If they adopt RISC-V, it will be good for them, I think it will even save their business
@@jokinglimitreached1503 Yeah, but they wont be protected by patents so the competition will be much higher. No more milking the customers as it is now.
Of course I agree and hope that implementing RISC-V will be the only way forward for them eventually.
Someone needs to make a GPU like Intel Larrabee, made of thousands of RISC V cores, that can be programmed just like a CPU.
Can RISC-V adoption help deal with the current chip supply issues companies are all facing?
Nope...its an architecture which you'll still need to fabricate...thers no alternative to the dependency on TSMC and Samsung sadly...
Still, the next big thing in consumer computers is ARM because of Apple M1. But I have a serious question. Is it even possible now to create a general purpose CPU on Risc-V to be close in performance to what typical CPU offer? I know it is great to make, specialized co-processors and stuff, but that long that anyone do not take this approach and architecture and do something for mass consumers it will stay in the limbo of co-processors.
not yet. but soon...
As a layperson, I don't understand what you mean by open-source computing architecture. Does it mean for instance if i buy a computer using such a chip I will need software specifically compiled to run on it? Kind of how windows 64 bit software don't run on older windows 32 machines.
If u r asking a basic question like this, u should stick with apple products. U will be happier that way.
Yes, you'll need specifically compiled software, but recently there has been a lot of development in translation programs which allow running software from other architectures without much performance loss. However, that has nothing to do with the open source part. It being open source just means that everyone designing hardware has access to the basic feature set of RISC-V and they don't have to pay anyone for it. This facilitates faster and cheaper development and helps innovation.
5:37 swiss is not a language
I don't understand why Apple push ARM... why he doesn't invest money into the developing RISC-V. RISC-V have much better efficiency.
Why would Apple have chosen ARM rather than RISCV then? Just curious.
RISC-V has only gained prominence over the last few years. iPhone came out before RISC-V got on the radar. Apple also bought a very expensive license from ARM so they are free to customize as they see fit while most vendors buy a finished core design.
You know, the 1st time AMD could really compete with Intel was in 1996 with their K5 CPU - which used a RISC processor and a x86 decoding front-end that was faster clock-for-clock compared to the [original] Pentium...
Hit TSMC up...
Yeah. RISC is good.
What can i do to get something like this to develop and test with?
RISC V dev boards. Like a RPi.
look up the BeagleV
How This business Model is any difference than ARM? The People who do custom cores still end up selling to other vendors. Very few would need a super custom architecture, most just want to build a Phone or something ...
I for one welcome our RISC-V robot overlords.
Is it gonna beat ARM out of the gate or would this take lots of time?
It's targeting a different slice of the market so I doubt it will be a direct competitor anytime soon. I could see this blowing up with new companies innovating in the Internet of Things (Iot) market
@@devingearing5714 I guess people was right about ARM still being leadership position for quite a long time.
@@justaknight4719 arm can't replace x86 fully. Arm sucks at emulating x86. We need x86 for legacy programs.
Unless there's a mass product that use RISC-V like smartphone (from major company) it will be long road
@@grossly820 In the corporate world you can side-step the issue by having legacy code run on remote (cloud) machines using technologies such as "citrix"
The RISC-V ISA documentation is a case study to teach students how NOT to write documentation. Pages and pages of details and never a clear picture of the general idea. Typically case of head in people for who of course this big picture is obvious and what they are proud of is this little detail, but they don't realize a technical documentation is intended to people with the totally opposite perspective.
This result in a case of the open source fallacy, "open" is not really "open" if the documentation is badly done.
Seriously make some effort there !!!
For exemple on how to write great ISA documentation, check out old Motorola 68HC11 programmer's guides. Those were masterpieces of the field. Both super detailed and super clear.
@@AIR-Dmnshn indeed it is. And so what ?
Great vid
do you claim emulation, at any level, is restricted?
if you need to go full vertical intergration, you could code the program directly to fpga or asic, with no conflicts
Is adding to the list of set instructions going to be a democracy?
Because this instruction set is an open standard, one wouldn't be beholden to the licensing fees of the few big companies that make processors today.
It's designed to be extended. If I remember right, official extensions and custom ones don't overlap, so you can add your own instructions without asking permission.
Wow pack a 1000 cores in a chip using only 20 watts that be great
I keep hoping AMD will start making Risc-V chips. Too much to ask?
That is real good, finally we gonna be free of Intel monopoly!
Intel and AMD form a duopoly for processors that use the x86 and x64 architectures.
Will ARM software work on RISC-V CPUs?
No, it's new ISA based on RISC Instruction.
No, and Nvidia will turn ARM into a dead-end technology for licensees.
With emulation or recompiling, yes, but it'll be slower.
1000 cores into 20w??? DAAAAAAMNNNN
its developing so slow.. i think i will be old before it becomes main stream trend
No it's actually not. You have to remember that RISC-V was *announced* not even 5 years ago
It’s all great until you need a 10 billion dollar FAB to materialize anything, and at that point you’re completing against every other technology. Seems to be competitive on any level out spending the other guys is an absolute necessity.
Science Fraction
I wonder if this processor is capable to talk for the blind so someone like me could develop our own product.
NIce....
Starts at 0:35
Is Intel sleeping or are they going to leverage these developments for future products? RISC processors have been around for a long time. So it is not like Intel is being taken by surprise.
We want a risc v soc to replace raspberry pi.
bl-604
Hell yeah
interesting but this thing will not easily going to be useful compute device for users any soon and the first ever will be really slow and backwards. but no brainer on storage, network devices, cameras, etc. though I could be wrong, if something as powerful as China starts using this for consumer and government
2:00 GILF
BeagleV! :-D
Great info, but Krste Asanović may need to see a doctor, he appears to be fading away.
I've been working in isolation and i'm so sad that i'm traped in a life that is still a Mystery to me...
Too bad so sad
Open source is not free
Wow! How did you lose so much weight?!!!
How am i so early
Cleveland won this trade. I think Brooklyn is going to implode in the playoffs.
Yay. Hopefully politicians are out of tech for good.
What?! This doesn't have much to do with politics. Mostly allowing for more competition in stale market resulting from large companies stifling innovation.
HP Inc files for Bankruptcy! inshallah!
lol is he trying to lift off?
Oh yeah. Do you love me? 😍💋 💝💖❤️
Try moving in more frantic/spastic ways as you talk... it makes it super easy to avoid psychosis while trying to watch you speak (sarcasm; had to turn off the video not 5 sec into your intro)
How are these people so bad at talking??
Talking heads just show us it working...lol
first
Too long didn't listen much, and hard too understand nyahaha 😹😹😹😹 when didn't include a more understandable simple explanation
very good guests but very bad interviewer