I just wish they had chosen a GPU from a design firm that was more open than Imagination; for years they just about refused to acknowledge that open source software existed. I suspect it was because they had 3rd party IP in their designs, but ofc they aren't saying. In the past Imagination would say "if you want open source drivers then approach your hardware vendor" but what Imagination failed to mention is that they made it infeasible for their counterparties to provide details and also made it infeasible for their counterparties to even acknowledge that Imagination was making it so difficult. So now that they've had a new CEO for the last two years this change in the way they are engaging with the open source community is a genuine sea change.
@@TankEnMatewell I'll give you great news that some of their GPUs have open source drivers now with more on the way the problem now is they need to redo all firmware for the GPU for the open source version and that's going to take a second
@@vixea I know Imagination has open source drivers (kernel (not merged just yet) and mesa), but from my reading it only supports Vulkan 1.0 at the moment (as opposed to 1.3 for AMD / Intel). Not that this is a problem for most people, but it also "supports" OpenGL via Zink; ... it doesn't currently support any of the Vulkan physical device features and only a couple of the (dozen or so) Vulkan extensions to get Zink to work; which means no open source OpenGL on open source Vulkan drivers. So I would contend that it is going to take far more than a second to get this working reliably (let alone usably) in even the medium time frame. Imagination may be on the way to open source land but they still have a long way to go.
It’s likely the image processing tasks are so slow because the image libraries used have hand-optimized assembly implementations for things like decoding, and they just don’t exist yet for RISC-V. Same thing probably applies to things like video decoding, and possibly even the cryptography you mentioned. Edit: Also, love to see that PCIe works well out of the box!
Either that or compilers for spir-v aren't mature enough yet and produce suboptimal code (no auto-vectorization would be a killer for image libraries for example)
Although this SBC was honestly a pain to get running (I have the v1.2a version which needed to be updated by TTL-UART), RISCV has a few tricks up its sleeve that make it extremely viable in this day and age - the first being the significant amount of microcontrollers being released using the architecture, with some even having enough power to run a full linux install. Having some sort of parity between desktop-grade compute and embedded compute like this is extremely helpful for development. On top of this, being able to emulate RISCV cores in FPGAs without any sort of restrictive licensing has essentially enabled an entirely new avenue for hobbyists and small businesses to design truly high performance hardware applications.
RISC V on FPGAs is something I don't see most people mention and it's a pretty big deal for open source hardware. Just having open RISCV based cores would be huge even if it needs to run on an FPGA. People can iterate on the core designs petty rapidly this way.
The potential problem with RISC-V is how variable and modular it can be. Most microcontrollers will only use a 32 bit core and may not have many extensions, compared to a SBC that will probably be 64 bit and have at least the basic set of extensions. Something that might make this even worse is RV32E which is specifically for embedded devices and changes some main aspects of the architecture. If companies make a habit of adding in custom extensions for things like AI then it could get messy as some software may only run on a specific brand of cores, essentially leading to vendor lock in or needing to rewrite a program to use a different brand, that wouldn’t be a problem if everyone just agreed to work together and only use official extensions but that may not happen. Another potential issue could be large differences in performance based on the architecture of that specific chip, all of the chips may be RISC-V and may have all the same extensions and even run at the same clock speed but differences in how it is implemented could create a large difference in performance, this could make comparing performance of different chips a lot harder than it currently is. On top of that with it being open you might end up with lots of companies all releasing their own RISC-V chips and you end up with a market stuffed full of ever so slightly different chips and cores and it may make it more difficult to find what you actually need.
@@shadow7037932 there are open source RISC-V cores that already exist but unless there is serious work done to them they won’t compete with cores made by dedicated companies and they don’t need to open source their implementation. It is relatively easy to implement a basic RISC-V core but if you want more advanced features like pipelining or out of order processing then it gets a lot more difficult. Companies that make more advanced cores will license them, just like ARM does and there is a good chance those cores will be much better than the open source ones available, so if you want a good core that has been developed by a team of engineers and thoroughly tested and documented then you will probably still have to license it. Most small companies have enough to worry about without having to design or modify the processor before they even start on their actual product, they would also have to hire engineers that specialise in processor design which is a skill on its own and is not the same as implementing other things on FPGAs. So it is far faster, easier and probably overall cheaper for them to license a core that is known to work and then build their product around that. Everyone saying things like, “small companies and hobbyists can make their own RISC-V cores”, are really underestimating how much work and knowledge goes into making a decent processor. If it was so easy then companies like ARM wouldn’t have such a hold on the industry and making your own processor would have been much more of a thing before now.
@@conorstewart2214 this is all true, but one of the pure, crystalline, truly good things about computing is how, if they're passionate and dedicated, a single person can move mountains. you could equally make the same statement about creating a AAA game engine or programming language - but you can do it, and people like Muratori, Blow, Tarn and Zach Adams, and many others are proof positive that you can actually make the journey of a thousand miles if you just keep stepping. the "road never traveled" always looks prohibitively difficult...until the first person decides to walk it. this is a drastic perspective shift to the conversation, but i believe the future of computing as a force for good in the world rests on the chip being "opened up". we already live in a world where top-down control of computing infrastructure is used to facilitate a total surveillance state, down to assassinations of journalists and political dissidents. corporate capture of the internet has created a situation where the global commons we became used to is now centralized and controlled through a few siloed "TV channel" style platforms, with all of our data being scraped constantly, invisible censorship, and complete top-down information shaping (will this comment disappear into the ether because i used the word "assassination" twice? let's find out! ) all attempts to empower regular people with technology will be futile until we are able to start reclaiming the stack from the very ground up, and that starts with chips that are truly open, even if anemic at first. is it a lofty goal? sure, it's an insanely lofty goal. but that journey of a thousand miles looks more and more necessary. the current technological paradigm is quite literally driving people insane. getting out from that will require a drastic adjustment of expectations and willingness to rebuild from the ground level. opening up the chip represents the first and most important lock on the technological cage we find ourselves in.
We all know that Jeff has already bought many of these boards. As with raspberry Pi's, this will create a global shortage but he will be able to continue delivering his so interesting videos. 🤣
As a huge RISC-V advocate, I agree that if you want to get a RISC-V board right now, do it because you're interested in RISC-V development. It's really early, and it's definitely exciting, but if you want to do anything productive, this generation just isn't there _yet_.
Thanks for your honest and technical review, I will be very excited about these as they continue to solidify the ecosystem. We have to build it, and you're an integral part of the community. Your videos show people how not to be daunted by compiling or changing code, we need more brave people to explore it and fix it.
Wait, so they did a complete actual PCI-E implementation? Impressive! The rumor I've heard is that the compatibility on the arm boards is low because the soc vendors often don't really implement all of PCI-E, just enough for one or two specific use cases
One thing I learned over the years is that no matter how clear the spec is, harware vendors will find no shortage of ways to violate it, and no limit of creativity on how to disappoint you. You can't rely on anything that is mandatory according to the spec if the vendor doesn't specifically mention support for that feature and has tested for it. Just lower your standards by like a LOT what is acceptable, and then lower your expectations even further and then you have industry practice.
Nobody who touched one of these should think it will start a revolution in the industry. It's just plain not ready. I would be promoting this board so hard if it was ready. It's just not. Everything Jeff covered here is true. Will RISC-V be a revolution? Honestly if we can get people to value open source hardware like they value open source hardware yes, it could be. The companies don't really live up to it being fully open source hardware if you look into the details. That part doesn't concern me. If we can get people to believe in open-source hardware then it doesn't matter that the companies aren't living up to it today. It's more an educational problem just like open-source software. Once people understand what it does for them they will like it. The problem is the consumer devices are not ready. This device is not ready. We'll have to look at it 6 months to a year from now where a lot of this will be fixed (and thanks for discussing that so much in the video Jeff, I agree 100%). Today they're not ready though. I can't emphasize enough how disappointing that firmware update process was to even run the latest OS. That is just plain unacceptable!
@@JeffGeerling Oh wow, I'm glad you saw that one! That was from my own journey with this board and frankly I reached much of the same conclusions you did. Most importantly this board is not ready for Pi fans or even SBC fans to jump into unless they are experts. I don't recall one with a worse firmware update process in recent memory that I've had to do! Very good testing with the GPU as well. That's not one I had tried on there but I'm not surprised. You couldn't be more right about how difficult it is to get anything GPU accelerated going in Linux and that's a huge problem on a lot of alternative SBCs that aren't ARM as well and it's always worth mentioning. Take care Jeff!
@@realms4219 That's already happening. My Google Pixel 7 Pro uses a RISC-V chip. Companies that want to make their own silicon are doing this. RISC-V is also showing up a lot in microcontrollers that are fantastic to use today (ESP32-C3). The question is when will someone make a good enough single board computer powered by RISC-V. We already have good enough phones. We already have good enough microcontrollers. When will we get a good enough SBC? Not yet unfortunately but it's only a matter of time.
@@JamesAChambers How much is 'open source' hardware going to matter when ccomputer/chip designers are going to be making their own secret sauce parts or custom derivatives?
I currently try to shift over most of my dev work to using Risc-V mcu and hopefully sbc s , its for sure the future and software support and actual application use are critical so every helping hand counts.
Really liked that movie. Don't look at what you see on the screens. But the story and hacking community / world was totally not bad for a film... Maybe add the end, just disconnect de server from the internet :)
That RUclips video you were playing back was using AV1, so definitely wouldn't get accelerated decoding. There's an extension for Firefox to force a specific codec.
yes, this is still a real devboard, the software is not mature enough for general consumption. But the RISC-V scene it's improving fast, far faster than ARM is going.
@@marcogenovesi8570 this is annoyingly true. When Apple released the M series it really seemed like ARM was gonna make leaps and bounds but it just sort of fizzled out.
@@Freshbott2 Quite frankly I didn't share that hope. ARM's licensing structure has always cultivated mostly lazy OEM customers that just do the bare minimum integration of whatever very very underwhelming bs ARM designs and licenses out to the "plebs" and call it a day. Yes a few high end licensees like Apple and Ampere/Nvidia are out there doing their designs but that's it. With Risc-V there is no "higher authority" that drip feeds the same OEMs with a "good enough" product so you get plenty of "ARM-like" CPU design firms that create new designs and compete with each other.
@@Freshbott2 I mean, wtf is Qualcomm even doing at this point. They were supposed to herald the "windows on ARM" revolution and that thing went like a fart in the wind. That's because they are mostly just "licensed CPU integrators" and not doing true CPU design
@@marcogenovesi8570 I agree with you about ARM, I was more naive than you but looking back it’s obvious. But I’m not going to dupe myself over RISC-V. It’ll cultivate the same laziness but cheaper and more accessible. In the end it’s still an off-the-shelf model. Just because you can go custom doesn’t mean OEMs will do it outside the embedded/daughter chip space.
I spent about two hours browsing the forum and documentation before buying one of these SBCs. Things don't look any worse than in the early days of Arduino or Raspberry PI :) I look forward to seeing if this upstart can disrupt the Arm ecosystem. Ironically, I was also passionate about Arm challenging Intel back in the day.
The poor test cores with image processing and AI are likely because there isn't hardware SIMD yet. The U74 core is listed as RV64GBC, P or V is required for SIMD.
A risc-v enter to the bar...a new contender is always great news. However is too early to tell how it will perform, definitively a pre-alpha era. For the early testers, this is a Risky-V-usiness haha. Wonderful video as always!
I got a couple of Ox64 in the mail just yesterday, and while they'll be a fun excursion for learning to use RISC-V, something able to run self contained like this sounds like it would be more useful. Unless you want to limit yourself to an IoT device that you ssh into anyway.
Jeff. With so many boards out could you do a video comparing them as of date xxxx? Maybe once a quarter or twice a year. It is a pain to purchase one of these and have the manufacturer go away, software never mature, etc. thanks!
@@JeffGeerling makes sense. For now, I am glad I have a number of Pi 4s. But those things are 3+ years old at this point. I am hoping to try some of these new units, but hate to go down a dead-end road.
Oooh, I love this! I've never done anything with RISC-V but I have been rooting for it all along! The computing landscape will be so much more interesting with this! And many gadgets will be much easier to produce
My prediction: RISC-V will first eat up the microcontroller and embedded space (because hopefully less licensing hassle, hopefully lots of open source development tooling available and ability to just get the hardware you need without anything more), it will need a good-enough SBC so people can use it for easier development, testing, etc. And it looks like this will soonish be a reality.
Risc V has to start somewhere. It has a fight against ARM but folks seem to value the positives enough where we should see a lot of software development over the next few years. Thanks for the video!
This is pretty exciting - lots of potential and room for improvement! I don't know if I would buy one today but will definitely be watching the RISC-V space for signs of maturity. Thanks for sharing your experience with getting this board running!
I wish it was available in a desktop form factor (M-ATX). I'm dying for a modern, high end, reasonably affordable non-X86 desktop platform for high end low power desktop builds. especially if it has good linux support, and ESPECIALLY if one could get games running on it. If one of these companies want to become a new defacto arcitecture, they really need to get them out into the hands of influential people and get these out for development
The person-hours wasted on us all individually figuring out how to get a board running could be put to so much better use. I keep wondering why one of the medium sized SBC vendors doesn't put a temporary freeze on constantly churning out new hardware revisions and instead seeds the market with developer hardware and co-ordinates & funds some community action to get a half-decent Pi alternative. Then they can return to their current marketing strategy of popping out new hardware, knowing that the demand will be so much greater as the customer will be able to make use of it out of the box. They don't have to run to all the corner cases, just get to a stable mainstream Linux release with a reasonable desktop. We demand consumer choice, but look what it's given us!
I have this card! I've been using since day one, I had to compile the RISC-V tool chain to cross compile U-Boot, OpenSBI and a custom kernel. This took so much time but I got a Linux working on it by compiling everything myself ! However I haven't touched the card since then.
China does a lot of RISC. Propriety Architecture structures in a chip are all good. I'd be interested to see an analysis of the RISC Architecture by NIST for functional operation. The differences in test results are startling and the build provides a lot of insight in to design decisions. Your presentations Rock. Thanks Jeff
I've always pronounced Gaussian as GOW-see-an. Gauss was a physicist/mathematician whose name I've heard pronounced gowss. Source: PhD in physics in Canada. Curious if others have had a different experience.
Honestly the first time I actually said the word out loud was this script. I have a Gauss meter and say "Gowse" like "house." But for "Gaussian" in my head I say "gaw-se" like "sauce", not sure why!
I have the board and there is a patch for the HDMI output issue. Also, their GPU driver is not ready yet…. But it can run a Minecraft server without any issue, unlike VisionFive 1, the new board is powerful enough for a 1-2 player scenario.
When RISC was the standardized architecture before CISC, so it didn’t actually change anything. The only thing worth paying attention to in that movie was a pre-hollyweirdedAfricanadopted Angelina.
I do respect the effort being put into this but in the end - what's driving it? I suspect RiscV will be most popular inside China given the ease of replication and utilisation; but I'm not aware of an awful lot of successful base level optimisation work coming from there for OS builds. I've read that optimisation work is most timely when it's done for existing demand applications (mobile phones and gaming come immediately to mind, data centre optimisation may be after that and then broader OS functions). The RPI's ecosystem seems to be an anomaly given the community level work done here but RiscV would have to develop THAT level of support and will optimisation for one RiscV implementation REALLY translate to other implementations? If so will it be just a percentage of that benefit or in full and total? RPI was new and had nothing really like it before (and being out of the UK there was a sense of tinkerer-in-their-shed national pride going on). I'm not sure what level of optimisation work is community and how much falls on the Foundation's dev team (or perhaps 3rd parties looking to improve their Compute Module performance) but still, that's going to be a difficult flash to make happen again. Right now it really looks way too early to see if it's investment worthy outside of China-centric applications; I have a nasty feeling we'll see variations of RiscV SoCs driving weapons systems pointed at us if things get nasty in the South China Sea.
Risc-V ISA is so brilliant, the design is so perfect and well planned it makes you cringe to return to ARM and X64. To be fair, Risc-V assembly programming is only possible because we have good assembler with pseudos instructions (and the optimisations at linking time). I am glad we had CISC in 80-90's, but now Risc-V is the only possible future. I have the Nezha Sipeed Risc-V SBC and I love it. The hardware guys did their job, it is now the time for programmers to optimize their code for Risc-V.
IBM introduced the IBM PC on August 12, 1981. Compaq had a working BIOS for their first compatible computer, the Compaq Portable, by November 1982. So the time between the introduction of the IBM PC and Compaq having a working BIOS was approximately 15 months. It has been 11 years since the Raspberry Pi was introduced. Since then, no one has created effective competitor. Seems like all the small SBC talent is local to the Raspberry Pi Foundation.
Currently the avaliable RISC-V cores are pretty anemic AND compilers are not optimized for the ISA and software is not designed for this ISA. Add on top of that the crappy SBC Device tree/boot up/linux support all results in a poor user experience. This is not the only RISC-V board. Very similar comments WRT Arm SBC's. Most suck. Hopefully the Intel/SiFive developer board is a MAJOR improvement over the HiFive UnMatched?
I'm just hoping that RISC-V goes far. I'm tired of Acorn & Intel being bullies. Do you remember VIA? I remember them well. They're still manufacturing... But stopped designing cpus over 10 years ago.
0:57 and near-anything open source. Modern compilers tend to use multi-architecture code generators. (Rust, for instance, uses LLVM, along with many other languages and compilers, even for old languages like C/C++. GCC doesn't use LLVM, but has its own similar system.) All you have to do is swap out your compiler, and the exact same program will work the exact same way. Things get more complex with interpreted/JIT languages, but worst-case scenario, your program has to replace its JIT system with an interpreter, which still works correctly but is slower. That said, RISC-V is gradually becoming better supported by these systems. Some of them use LLVM just like a compiler. And unless you have your heart set on one specific, exact program, there are very few reasons why someone would genuinely need an interpreted language for performance-critical code.
6:52 Fun fact: Imagination is what’s left of the company that originally developed the MIPS architecture. Notwithstanding I think it still outships x86 3:1, looks like there’s not much money to be left in that market any more, so the company is now embracing RISC-V, including bringing in some GPU smarts.
That Hackers reference made me laugh. So many mixed feelings about that movie... lol. Great cast, bad script. Awesome soundtrack, horrible effects representing the computer side. But, yeah Risk good. They knew it even back then.
I realize competition is generally seen as a good thing, but I think it can be a bad thing if it fractures industries into many fragmented, incompatible ecosystems. If every chip manufacturer makes their own, proprietary ISA based on RISC-V, you're going to be limited to only the software and hardware made for that ISA. I think a common standard is better. That's why I like RISC-V, because it's open and free so any chip manufacturer can make a RISC-V chip. Companies can compete based on who makes the best RISC-V silicon rather than each chip manufacturer locking customers into a proprietary ecosystem.
Poor old RISC V. Been the "next big thing" for 10 years but I just don't think the architecture is going towards being a general purpose CPU. With the ability to add specialized ISAs it's easy to see how this will become the go to chip for embedded applications.
Love that Rocky Linux t-shirt. I'm a big fan of that distro. It's been great seeing the Linux community rally so quickly in the wake of the CentOS debacle.
Sounds like you watch some of the same channels as me. I know a lot of people will disagree with this, but I really wish someone would design a new CISC architecture and make it open. Get rid of all the garbage that Intel and AMD introduced to x86 processors and streamline things so it was a svelte CISC. Like Bisqwit says, why aren't all the move instructions just named mov.
The answer to the documentation is simple: The manufacturer has a vested interest. StarFive, and other companies in the field, need to create themselves a market - SBCs are the entry point for them to get their hardware in the hands of developers. The companies behind the SoCs on most ARM SBCs don't actually care about the SBC market - they're busy selling into phones, set-top boxes, security cameras, and so forth, where nobody cares about proprietary blobs or complying with licence requirements. This puts the burden of documentation on the SBC manufacturers, who also don't care because people will buy their boards anyway, because it's that or a Raspberry Pi.
f*** it all I'm making my own risc-v sbc now that there are 2 cent chips out i can probably make a cluster of like 12 of them and get something usable-ish, hell maybe it'll be able to host a website edit: to be clear I'm talking about these: CH32V003
I'm really, really excited about what the future of CPU architectures hold for us. RISC-V is already being utilized in a bunch of chips by lots of companies like NVidia and Google, but the main CPU is the last piece! This is even more probable now that Google announced they'd fully support RISC-V as a first class citizen in Android.
@@GoogleDoesEvil Windows is proprietary software and doesn’t respect your freedom. While most android distributions are proprietary, some (like GrapheneOS) are free software and respect your freedoms as a user.
@@speedytruck android is so locked down and limited that you can't even install any adblocker, and if you think privacy is bad on windows then it's universes upon universes worse on android.
Jeff I think in your presentation it would have helped some of us if you made a little recap on what RISC and RISC-V is and what is the significance about the board in this video using a different instruction set compared to e.g: The Pi 4. Or maybe I just need to watch the video again! Thank you.
Heh, I was waffling on how much time to spend on RISC-V itself, and I had an initial script with more... I might go back to it again next time I approach a RISC-V board.
I think our copyright laws are a bit outdated. I might be missinformed but to my knowledge design and implementation of actual chip is "orders of magnitude" more difficult than understanding and implementing specific instruction set (simple instructions have rather simple semantics). Unless it comes with dedicated hardware elements that instruction-set consortium provides this is like someone pattented a diesel car and no-one else can make new ones without paying royalties to original inventor. I would love to hear is there anything other than instruction set coming from the x86/Arm consortia that justify the royalties to be paid for making your own chip? (I am not talking about US embargo targeting China, if anything that embargo was implemented too late, licensing never stopped China copying the technology)
Might as well embed the development board into the GPU board which is where I assume AMD/nvidia are going as a product line option. Quality video as ever :D
0:18 these people stare right into the lens of a videoprojector... Do they think they are in the first Alien movie ?... oh wait, i know this cheesy movie 😛
This seems neat. Hopefully they're interested in working with upstream linux as opposed to every other SBC/SoC manufacturer. Imagination being the maker of the GPU is worrying though. They tend to be remarkably poor for linux support, and there seems to be limited interest in reverse engineering them o developing open drivers for them in general. There are theories that the apple silicon graphics are similar to Imagination GPUs though, so maybe when those get off the ground, there can be some common work there for others.
Imagination is actively working on an open source Vulkan Driver for the Rouge series of GPU's. Still may be a WIP, or simply not integrated into Zink/ffmpeg/gstreamer yet.
At $100 or so, I expect they are going to be selling these to the same kind of crowd that bought the original Pi, and those are the people who break out their Assembler and C skills and fix stuff that isn't working.
Hi Jeff, thanks for the video, very interesting and thought-provoking. Given the lack of RPi's atm this looks like it could gain a lot of traction and interest in the soc space. Look forward to more videos from you on RISC soc's. Perhaps I should look at using these kind of RISC based soc's for the many projects I have planned. I am due to finally get my cytoreductive surgery on the 15th so I won't be around for 2 to 3 months but when I get back home I look forward viewing some more content from you. Great to see you looking so well, cheers Jeff.
what a waste of time, I will go with something that works even if it is not the best SBC. It's like a car, I want one that will run faithfully. Not one on the side of the road in a snow storm.
Thank you for getting in on this topic so early, the whole RISC-V thing seems like it has so much possibility, but it's all "potential energy" right now and needs to be converted to "kinetic", or something. A whole lot of bootstrapping going on! I hope it will help democratize lower-end hardware, and maybe even challenge mid to high-end performance some day. It seems perfect for a lot of embedded uses... maybe. Will have to compete with ARM there. But why *not* compete with ARM there? Anything is possible. ARM's own origin story if very "something out of nothing", low expectations, big results. So, once again, very interesting! Maybe RISC-V can chart that "bolt out of the blue" path again, like ARM once did!
I love how dedicated GPU is easier to setup on a rather new RISC V SBC compared to your previous Raspberry Pi.
risk-V is just shiat
I just wish they had chosen a GPU from a design firm that was more open than Imagination; for years they just about refused to acknowledge that open source software existed. I suspect it was because they had 3rd party IP in their designs, but ofc they aren't saying. In the past Imagination would say "if you want open source drivers then approach your hardware vendor" but what Imagination failed to mention is that they made it infeasible for their counterparties to provide details and also made it infeasible for their counterparties to even acknowledge that Imagination was making it so difficult. So now that they've had a new CEO for the last two years this change in the way they are engaging with the open source community is a genuine sea change.
@@TankEnMatewell I'll give you great news that some of their GPUs have open source drivers now with more on the way the problem now is they need to redo all firmware for the GPU for the open source version and that's going to take a second
@@vixea I know Imagination has open source drivers (kernel (not merged just yet) and mesa), but from my reading it only supports Vulkan 1.0 at the moment (as opposed to 1.3 for AMD / Intel). Not that this is a problem for most people, but it also "supports" OpenGL via Zink; ... it doesn't currently support any of the Vulkan physical device features and only a couple of the (dozen or so) Vulkan extensions to get Zink to work; which means no open source OpenGL on open source Vulkan drivers. So I would contend that it is going to take far more than a second to get this working reliably (let alone usably) in even the medium time frame. Imagination may be on the way to open source land but they still have a long way to go.
@@TankEnMate nope the kernel driver was merged of course you need to use a mainline kernel
This video deserves an award based on thumbnail image lone. 😆
agreed! thats a well awesome thumbnail!
Yep, was about to say the same thing. 😂
True. 😂😂😂
yes
Yes! 😊
Alternatively:
Risc-V
/
Business
(Or wait to they release model y) 😁
It’s likely the image processing tasks are so slow because the image libraries used have hand-optimized assembly implementations for things like decoding, and they just don’t exist yet for RISC-V. Same thing probably applies to things like video decoding, and possibly even the cryptography you mentioned.
Edit: Also, love to see that PCIe works well out of the box!
Either that or compilers for spir-v aren't mature enough yet and produce suboptimal code (no auto-vectorization would be a killer for image libraries for example)
Although this SBC was honestly a pain to get running (I have the v1.2a version which needed to be updated by TTL-UART), RISCV has a few tricks up its sleeve that make it extremely viable in this day and age - the first being the significant amount of microcontrollers being released using the architecture, with some even having enough power to run a full linux install. Having some sort of parity between desktop-grade compute and embedded compute like this is extremely helpful for development. On top of this, being able to emulate RISCV cores in FPGAs without any sort of restrictive licensing has essentially enabled an entirely new avenue for hobbyists and small businesses to design truly high performance hardware applications.
RISC V on FPGAs is something I don't see most people mention and it's a pretty big deal for open source hardware. Just having open RISCV based cores would be huge even if it needs to run on an FPGA. People can iterate on the core designs petty rapidly this way.
The potential problem with RISC-V is how variable and modular it can be. Most microcontrollers will only use a 32 bit core and may not have many extensions, compared to a SBC that will probably be 64 bit and have at least the basic set of extensions. Something that might make this even worse is RV32E which is specifically for embedded devices and changes some main aspects of the architecture. If companies make a habit of adding in custom extensions for things like AI then it could get messy as some software may only run on a specific brand of cores, essentially leading to vendor lock in or needing to rewrite a program to use a different brand, that wouldn’t be a problem if everyone just agreed to work together and only use official extensions but that may not happen.
Another potential issue could be large differences in performance based on the architecture of that specific chip, all of the chips may be RISC-V and may have all the same extensions and even run at the same clock speed but differences in how it is implemented could create a large difference in performance, this could make comparing performance of different chips a lot harder than it currently is. On top of that with it being open you might end up with lots of companies all releasing their own RISC-V chips and you end up with a market stuffed full of ever so slightly different chips and cores and it may make it more difficult to find what you actually need.
@@shadow7037932 there are open source RISC-V cores that already exist but unless there is serious work done to them they won’t compete with cores made by dedicated companies and they don’t need to open source their implementation. It is relatively easy to implement a basic RISC-V core but if you want more advanced features like pipelining or out of order processing then it gets a lot more difficult. Companies that make more advanced cores will license them, just like ARM does and there is a good chance those cores will be much better than the open source ones available, so if you want a good core that has been developed by a team of engineers and thoroughly tested and documented then you will probably still have to license it.
Most small companies have enough to worry about without having to design or modify the processor before they even start on their actual product, they would also have to hire engineers that specialise in processor design which is a skill on its own and is not the same as implementing other things on FPGAs. So it is far faster, easier and probably overall cheaper for them to license a core that is known to work and then build their product around that.
Everyone saying things like, “small companies and hobbyists can make their own RISC-V cores”, are really underestimating how much work and knowledge goes into making a decent processor. If it was so easy then companies like ARM wouldn’t have such a hold on the industry and making your own processor would have been much more of a thing before now.
Nah. I will buy one when it becomes as easy as RPi to install an OS and use it.
@@conorstewart2214 this is all true, but one of the pure, crystalline, truly good things about computing is how, if they're passionate and dedicated, a single person can move mountains. you could equally make the same statement about creating a AAA game engine or programming language - but you can do it, and people like Muratori, Blow, Tarn and Zach Adams, and many others are proof positive that you can actually make the journey of a thousand miles if you just keep stepping. the "road never traveled" always looks prohibitively difficult...until the first person decides to walk it.
this is a drastic perspective shift to the conversation, but i believe the future of computing as a force for good in the world rests on the chip being "opened up". we already live in a world where top-down control of computing infrastructure is used to facilitate a total surveillance state, down to assassinations of journalists and political dissidents. corporate capture of the internet has created a situation where the global commons we became used to is now centralized and controlled through a few siloed "TV channel" style platforms, with all of our data being scraped constantly, invisible censorship, and complete top-down information shaping (will this comment disappear into the ether because i used the word "assassination" twice? let's find out! )
all attempts to empower regular people with technology will be futile until we are able to start reclaiming the stack from the very ground up, and that starts with chips that are truly open, even if anemic at first. is it a lofty goal? sure, it's an insanely lofty goal. but that journey of a thousand miles looks more and more necessary. the current technological paradigm is quite literally driving people insane. getting out from that will require a drastic adjustment of expectations and willingness to rebuild from the ground level. opening up the chip represents the first and most important lock on the technological cage we find ourselves in.
Between YOU and "Explaining Computers?" - You're Just TRYING to kill my pay check!!!! LOL
This review seems like a pitch meeting for this Risc V board. It was super easy, barely an inconvenience.
RISC V-based SBCs are tight!
All the documentation on RISC V sounds more like exposition
Wow wow wow, wow!
I see what you're all doing here !
So, you have a single-board computer idea for me?
We all know that Jeff has already bought many of these boards. As with raspberry Pi's, this will create a global shortage but he will be able to continue delivering his so interesting videos. 🤣
I like the idea that Jeff is singlehandedly responsible for the shortage
As a huge RISC-V advocate, I agree that if you want to get a RISC-V board right now, do it because you're interested in RISC-V development. It's really early, and it's definitely exciting, but if you want to do anything productive, this generation just isn't there _yet_.
Clicking the Like button was super easy, barely an inconvenience
Replying to this comment is _tight_!
Wowowowow. Wow.
Thanks for your honest and technical review, I will be very excited about these as they continue to solidify the ecosystem. We have to build it, and you're an integral part of the community. Your videos show people how not to be daunted by compiling or changing code, we need more brave people to explore it and fix it.
I would love to see a follow up after the RISC-V optimizations start trickling in.
I'm sure it will happen, but it might be 6 months or a 1 year before it happens.
see you in 2024
Wait, so they did a complete actual PCI-E implementation? Impressive! The rumor I've heard is that the compatibility on the arm boards is low because the soc vendors often don't really implement all of PCI-E, just enough for one or two specific use cases
One thing I learned over the years is that no matter how clear the spec is, harware vendors will find no shortage of ways to violate it, and no limit of creativity on how to disappoint you. You can't rely on anything that is mandatory according to the spec if the vendor doesn't specifically mention support for that feature and has tested for it.
Just lower your standards by like a LOT what is acceptable, and then lower your expectations even further and then you have industry practice.
Nobody who touched one of these should think it will start a revolution in the industry. It's just plain not ready. I would be promoting this board so hard if it was ready. It's just not. Everything Jeff covered here is true.
Will RISC-V be a revolution? Honestly if we can get people to value open source hardware like they value open source hardware yes, it could be. The companies don't really live up to it being fully open source hardware if you look into the details. That part doesn't concern me.
If we can get people to believe in open-source hardware then it doesn't matter that the companies aren't living up to it today. It's more an educational problem just like open-source software. Once people understand what it does for them they will like it.
The problem is the consumer devices are not ready. This device is not ready. We'll have to look at it 6 months to a year from now where a lot of this will be fixed (and thanks for discussing that so much in the video Jeff, I agree 100%). Today they're not ready though.
I can't emphasize enough how disappointing that firmware update process was to even run the latest OS. That is just plain unacceptable!
Thanks SO much for that post on the firmware update-it was incredibly helpful, more so than the official docs :)
@@JeffGeerling Oh wow, I'm glad you saw that one! That was from my own journey with this board and frankly I reached much of the same conclusions you did. Most importantly this board is not ready for Pi fans or even SBC fans to jump into unless they are experts. I don't recall one with a worse firmware update process in recent memory that I've had to do!
Very good testing with the GPU as well. That's not one I had tried on there but I'm not surprised. You couldn't be more right about how difficult it is to get anything GPU accelerated going in Linux and that's a huge problem on a lot of alternative SBCs that aren't ARM as well and it's always worth mentioning.
Take care Jeff!
Once companies consider these viable, the entire market will shift.
@@realms4219 That's already happening. My Google Pixel 7 Pro uses a RISC-V chip. Companies that want to make their own silicon are doing this. RISC-V is also showing up a lot in microcontrollers that are fantastic to use today (ESP32-C3).
The question is when will someone make a good enough single board computer powered by RISC-V. We already have good enough phones. We already have good enough microcontrollers. When will we get a good enough SBC? Not yet unfortunately but it's only a matter of time.
@@JamesAChambers
How much is 'open source' hardware going to matter when ccomputer/chip designers are going to be making their own secret sauce parts or custom derivatives?
I currently try to shift over most of my dev work to using Risc-V mcu and hopefully sbc s , its for sure the future and software support and actual application use are critical so every helping hand counts.
Having notifications on for awesome Jeff Geerling and Geerling engineering videos is also super easy, barely an inconvenience ❤️😁
This comment is tight!
@@JeffGeerling yeah yeah yeah ..
@@Palmamontt *TIGHT TIGHT TIGHT*
Wowowowow wow, finding people who get it in the comments is super easy, barely an inconvenience
Wow wow wow............................... Wow
Thanks Jeff, just got mine last week.
Hours of frustration and fun ahead.
Great video, thanks.
recompiling kernel until morale improves
For anyone wondering, the movie is "Hackers" from 1995.
Really liked that movie. Don't look at what you see on the screens. But the story and hacking community / world was totally not bad for a film... Maybe add the end, just disconnect de server from the internet :)
Super easy, barely an inconvenience! I love subtle channel crossovers. 😊
Wow wow wow wow wow. Wow
it's called reference, I decided
Subtle channel crossovers are TIGHT!
"So you have a new SBC for me?"....."Yes sir I do."
@@nathangoddard8115 haha, I like that one :-)
That RUclips video you were playing back was using AV1, so definitely wouldn't get accelerated decoding. There's an extension for Firefox to force a specific codec.
That's Rene or @MoreReneRebe at 8:54. Give Rene some love for messing around with RISC-V over on his RUclips Channel www.youtube.com/@MoreReneRebe.
"super easy barely any inconvenience"
I see what you did there
Super easy, Barely an inconvenience ;) NICE !
Kudos for the clip from Hackers!
I wonder if the pool on the roof still has a leak?!
Only one way to find out!
Appreciate your take on RISC-V. I would agree that it is still early development for RISC-V. Definitely not for a beginner like me.
yes, this is still a real devboard, the software is not mature enough for general consumption. But the RISC-V scene it's improving fast, far faster than ARM is going.
@@marcogenovesi8570 this is annoyingly true. When Apple released the M series it really seemed like ARM was gonna make leaps and bounds but it just sort of fizzled out.
@@Freshbott2 Quite frankly I didn't share that hope. ARM's licensing structure has always cultivated mostly lazy OEM customers that just do the bare minimum integration of whatever very very underwhelming bs ARM designs and licenses out to the "plebs" and call it a day. Yes a few high end licensees like Apple and Ampere/Nvidia are out there doing their designs but that's it.
With Risc-V there is no "higher authority" that drip feeds the same OEMs with a "good enough" product so you get plenty of "ARM-like" CPU design firms that create new designs and compete with each other.
@@Freshbott2 I mean, wtf is Qualcomm even doing at this point. They were supposed to herald the "windows on ARM" revolution and that thing went like a fart in the wind. That's because they are mostly just "licensed CPU integrators" and not doing true CPU design
@@marcogenovesi8570 I agree with you about ARM, I was more naive than you but looking back it’s obvious. But I’m not going to dupe myself over RISC-V. It’ll cultivate the same laziness but cheaper and more accessible. In the end it’s still an off-the-shelf model. Just because you can go custom doesn’t mean OEMs will do it outside the embedded/daughter chip space.
I spent about two hours browsing the forum and documentation before buying one of these SBCs. Things don't look any worse than in the early days of Arduino or Raspberry PI :) I look forward to seeing if this upstart can disrupt the Arm ecosystem. Ironically, I was also passionate about Arm challenging Intel back in the day.
Super easy, barely an inconvenience.......awesome :D
The poor test cores with image processing and AI are likely because there isn't hardware SIMD yet. The U74 core is listed as RV64GBC, P or V is required for SIMD.
The V extension is what's going to end up on general purpose processors. The P extenion looks like it's largely focused on 32-bit DSP applications.
Awesome! Any plans to review the Odroid H3+ board?
Love the Pitch Meeting reference, bravo sir. Keep up the great work!
A risc-v enter to the bar...a new contender is always great news. However is too early to tell how it will perform, definitively a pre-alpha era.
For the early testers, this is a Risky-V-usiness haha. Wonderful video as always!
I hope I'm not taking a RISC watching this so early!
I got a couple of Ox64 in the mail just yesterday, and while they'll be a fun excursion for learning to use RISC-V, something able to run self contained like this sounds like it would be more useful. Unless you want to limit yourself to an IoT device that you ssh into anyway.
Jeff. With so many boards out could you do a video comparing them as of date xxxx? Maybe once a quarter or twice a year. It is a pain to purchase one of these and have the manufacturer go away, software never mature, etc. thanks!
I've considered it. The problem is the landscape changes so much, it's really hard to keep up (and still have a life) :/
@@JeffGeerling makes sense. For now, I am glad I have a number of Pi 4s. But those things are 3+ years old at this point. I am hoping to try some of these new units, but hate to go down a dead-end road.
Oooh, I love this! I've never done anything with RISC-V but I have been rooting for it all along! The computing landscape will be so much more interesting with this! And many gadgets will be much easier to produce
My prediction: RISC-V will first eat up the microcontroller and embedded space (because hopefully less licensing hassle, hopefully lots of open source development tooling available and ability to just get the hardware you need without anything more), it will need a good-enough SBC so people can use it for easier development, testing, etc. And it looks like this will soonish be a reality.
Start with a microcontroller. Esp32c3 is risc-v.
YES!!! I've been waiting for someone to use that clip from Hackers about this topic for SO LONG, I knew I couldn't be the only one to remember it !!!
I have forgotten why Risk V was so good. It was that long ago. The Hackers clip was perfectly on point.
Sad you didn’t slide in to frame with your socks and boxers on 😂
Oh believe me, I considered it...
Love the shoutout to Pitch Meeting by Ryan George - "Super easy, barely an inconvenience!" (if you know, you know)
Risc V has to start somewhere. It has a fight against ARM but folks seem to value the positives enough where we should see a lot of software development over the next few years. Thanks for the video!
This is pretty exciting - lots of potential and room for improvement! I don't know if I would buy one today but will definitely be watching the RISC-V space for signs of maturity.
Thanks for sharing your experience with getting this board running!
This man continues to grew ever-creative on his thumbnails...
Easily the best thumbnails on RUclips. Outstanding.
Only rasberry pi is facing global crisis, everyone else can build and ship whatever they want.
Risc V will take several years to close the gaps with ARM. But eventually, they will be there.
"super easy, barely an inconvenience" - reminded me i haven't watched movie pitch videos in a while!
Thank God we didn't have to see Redshirt Jeff do the sock slide in his underoos to the Bob Seger song .
I wish it was available in a desktop form factor (M-ATX). I'm dying for a modern, high end, reasonably affordable non-X86 desktop platform for high end low power desktop builds. especially if it has good linux support, and ESPECIALLY if one could get games running on it. If one of these companies want to become a new defacto arcitecture, they really need to get them out into the hands of influential people and get these out for development
Watching Jeff Geerling is always a RISC...y business.
The person-hours wasted on us all individually figuring out how to get a board running could be put to so much better use. I keep wondering why one of the medium sized SBC vendors doesn't put a temporary freeze on constantly churning out new hardware revisions and instead seeds the market with developer hardware and co-ordinates & funds some community action to get a half-decent Pi alternative. Then they can return to their current marketing strategy of popping out new hardware, knowing that the demand will be so much greater as the customer will be able to make use of it out of the box. They don't have to run to all the corner cases, just get to a stable mainstream Linux release with a reasonable desktop. We demand consumer choice, but look what it's given us!
PS, Did Redshirt Jeff buy them all, not any in stock in mainstream outlets!
I have this card! I've been using since day one, I had to compile the RISC-V tool chain to cross compile U-Boot, OpenSBI and a custom kernel. This took so much time but I got a Linux working on it by compiling everything myself ! However I haven't touched the card since then.
those are way too expensive, for that kind of money I might as well go get a mini-pc and the price difference in power consumption is not that bad.
China does a lot of RISC. Propriety Architecture structures in a chip are all good. I'd be interested to see an analysis of the RISC Architecture by NIST for functional operation. The differences in test results are startling and the build provides a lot of insight in to design decisions. Your presentations Rock. Thanks Jeff
I've always pronounced Gaussian as GOW-see-an. Gauss was a physicist/mathematician whose name I've heard pronounced gowss. Source: PhD in physics in Canada. Curious if others have had a different experience.
Honestly the first time I actually said the word out loud was this script. I have a Gauss meter and say "Gowse" like "house."
But for "Gaussian" in my head I say "gaw-se" like "sauce", not sure why!
@@JeffGeerling I'm going to have to unsubsribe from your channel.
I have the board and there is a patch for the HDMI output issue. Also, their GPU driver is not ready yet…. But it can run a Minecraft server without any issue, unlike VisionFive 1, the new board is powerful enough for a 1-2 player scenario.
When RISC was the standardized architecture before CISC, so it didn’t actually change anything.
The only thing worth paying attention to in that movie was a pre-hollyweirdedAfricanadopted Angelina.
Thank You. Definitely worth watching out for new RISC boards and improvements to this one.
I do respect the effort being put into this but in the end - what's driving it?
I suspect RiscV will be most popular inside China given the ease of replication and utilisation; but I'm not aware of an awful lot of successful base level optimisation work coming from there for OS builds. I've read that optimisation work is most timely when it's done for existing demand applications (mobile phones and gaming come immediately to mind, data centre optimisation may be after that and then broader OS functions).
The RPI's ecosystem seems to be an anomaly given the community level work done here but RiscV would have to develop THAT level of support and will optimisation for one RiscV implementation REALLY translate to other implementations? If so will it be just a percentage of that benefit or in full and total? RPI was new and had nothing really like it before (and being out of the UK there was a sense of tinkerer-in-their-shed national pride going on). I'm not sure what level of optimisation work is community and how much falls on the Foundation's dev team (or perhaps 3rd parties looking to improve their Compute Module performance) but still, that's going to be a difficult flash to make happen again.
Right now it really looks way too early to see if it's investment worthy outside of China-centric applications; I have a nasty feeling we'll see variations of RiscV SoCs driving weapons systems pointed at us if things get nasty in the South China Sea.
I cant believe they actually made the ISA from the textbooks into a computer
Jeff never disappoints with the thumbnails 10/10
Risc-V ISA is so brilliant, the design is so perfect and well planned it makes you cringe to return to ARM and X64. To be fair, Risc-V assembly programming is only possible because we have good assembler with pseudos instructions (and the optimisations at linking time). I am glad we had CISC in 80-90's, but now Risc-V is the only possible future. I have the Nezha Sipeed Risc-V SBC and I love it.
The hardware guys did their job, it is now the time for programmers to optimize their code for Risc-V.
IBM introduced the IBM PC on August 12, 1981. Compaq had a working BIOS for their first compatible computer, the Compaq Portable, by November 1982. So the time between the introduction of the IBM PC and Compaq having a working BIOS was approximately 15 months.
It has been 11 years since the Raspberry Pi was introduced. Since then, no one has created effective competitor.
Seems like all the small SBC talent is local to the Raspberry Pi Foundation.
Currently the avaliable RISC-V cores are pretty anemic AND compilers are not optimized for the ISA and software is not designed for this ISA. Add on top of that the crappy SBC Device tree/boot up/linux support all results in a poor user experience. This is not the only RISC-V board. Very similar comments WRT Arm SBC's. Most suck. Hopefully the Intel/SiFive developer board is a MAJOR improvement over the HiFive UnMatched?
I'm just hoping that RISC-V goes far. I'm tired of Acorn & Intel being bullies. Do you remember VIA? I remember them well. They're still manufacturing... But stopped designing cpus over 10 years ago.
0:57 and near-anything open source.
Modern compilers tend to use multi-architecture code generators. (Rust, for instance, uses LLVM, along with many other languages and compilers, even for old languages like C/C++. GCC doesn't use LLVM, but has its own similar system.) All you have to do is swap out your compiler, and the exact same program will work the exact same way.
Things get more complex with interpreted/JIT languages, but worst-case scenario, your program has to replace its JIT system with an interpreter, which still works correctly but is slower. That said, RISC-V is gradually becoming better supported by these systems. Some of them use LLVM just like a compiler. And unless you have your heart set on one specific, exact program, there are very few reasons why someone would genuinely need an interpreted language for performance-critical code.
6:52 Fun fact: Imagination is what’s left of the company that originally developed the MIPS architecture. Notwithstanding I think it still outships x86 3:1, looks like there’s not much money to be left in that market any more, so the company is now embracing RISC-V, including bringing in some GPU smarts.
That Hackers reference made me laugh. So many mixed feelings about that movie... lol. Great cast, bad script. Awesome soundtrack, horrible effects representing the computer side. But, yeah Risk good. They knew it even back then.
I realize competition is generally seen as a good thing, but I think it can be a bad thing if it fractures industries into many fragmented, incompatible ecosystems. If every chip manufacturer makes their own, proprietary ISA based on RISC-V, you're going to be limited to only the software and hardware made for that ISA. I think a common standard is better. That's why I like RISC-V, because it's open and free so any chip manufacturer can make a RISC-V chip. Companies can compete based on who makes the best RISC-V silicon rather than each chip manufacturer locking customers into a proprietary ecosystem.
Poor old RISC V. Been the "next big thing" for 10 years but I just don't think the architecture is going towards being a general purpose CPU. With the ability to add specialized ISAs it's easy to see how this will become the go to chip for embedded applications.
Love that Rocky Linux t-shirt. I'm a big fan of that distro. It's been great seeing the Linux community rally so quickly in the wake of the CentOS debacle.
Sounds like you watch some of the same channels as me. I know a lot of people will disagree with this, but I really wish someone would design a new CISC architecture and make it open. Get rid of all the garbage that Intel and AMD introduced to x86 processors and streamline things so it was a svelte CISC. Like Bisqwit says, why aren't all the move instructions just named mov.
The answer to the documentation is simple: The manufacturer has a vested interest. StarFive, and other companies in the field, need to create themselves a market - SBCs are the entry point for them to get their hardware in the hands of developers. The companies behind the SoCs on most ARM SBCs don't actually care about the SBC market - they're busy selling into phones, set-top boxes, security cameras, and so forth, where nobody cares about proprietary blobs or complying with licence requirements. This puts the burden of documentation on the SBC manufacturers, who also don't care because people will buy their boards anyway, because it's that or a Raspberry Pi.
RISC-V is the future!
I look forward to a RISC-V chiplet cpu with an fpga chiplet.
f*** it all I'm making my own risc-v sbc now that there are 2 cent chips out i can probably make a cluster of like 12 of them and get something usable-ish, hell maybe it'll be able to host a website
edit:
to be clear I'm talking about these: CH32V003
I'm really, really excited about what the future of CPU architectures hold for us. RISC-V is already being utilized in a bunch of chips by lots of companies like NVidia and Google, but the main CPU is the last piece! This is even more probable now that Google announced they'd fully support RISC-V as a first class citizen in Android.
@@GoogleDoesEvil Windows is proprietary software and doesn’t respect your freedom. While most android distributions are proprietary, some (like GrapheneOS) are free software and respect your freedoms as a user.
@@speedytruck android is so locked down and limited that you can't even install any adblocker, and if you think privacy is bad on windows then it's universes upon universes worse on android.
Jeff I think in your presentation it would have helped some of us if you made a little recap on what RISC and RISC-V is and what is the significance about the board in this video using a different instruction set compared to e.g: The Pi 4. Or maybe I just need to watch the video again! Thank you.
Heh, I was waffling on how much time to spend on RISC-V itself, and I had an initial script with more... I might go back to it again next time I approach a RISC-V board.
I think our copyright laws are a bit outdated. I might be missinformed but to my knowledge design and implementation of actual chip is "orders of magnitude" more difficult than understanding and implementing specific instruction set (simple instructions have rather simple semantics). Unless it comes with dedicated hardware elements that instruction-set consortium provides this is like someone pattented a diesel car and no-one else can make new ones without paying royalties to original inventor. I would love to hear is there anything other than instruction set coming from the x86/Arm consortia that justify the royalties to be paid for making your own chip? (I am not talking about US embargo targeting China, if anything that embargo was implemented too late, licensing never stopped China copying the technology)
Might as well embed the development board into the GPU board which is where I assume AMD/nvidia are going as a product line option. Quality video as ever :D
0:18 these people stare right into the lens of a videoprojector... Do they think they are in the first Alien movie ?...
oh wait, i know this cheesy movie 😛
RiscV will never beat Arm, in efficiency, usability, or use % in computing.
Love how cheese ball you are, it speaks to me dearly. Hope your bum bum is doing okay man! We love you!
In Russia they are already researching RISC V since many years, but the performance is just terrible . The CPUs are called „Elbrus“.
Love everything you do, but We'd be lying if We said We weren't specifically desiring a Hackers reference...you did not disappoint.
This seems neat. Hopefully they're interested in working with upstream linux as opposed to every other SBC/SoC manufacturer. Imagination being the maker of the GPU is worrying though. They tend to be remarkably poor for linux support, and there seems to be limited interest in reverse engineering them o developing open drivers for them in general. There are theories that the apple silicon graphics are similar to Imagination GPUs though, so maybe when those get off the ground, there can be some common work there for others.
Imagination is actively working on an open source Vulkan Driver for the Rouge series of GPU's. Still may be a WIP, or simply not integrated into Zink/ffmpeg/gstreamer yet.
that thumbnail:
slow clap
smirk smile (I'm sure i've done a typo here)
"well done, jeff. well done"
At $100 or so, I expect they are going to be selling these to the same kind of crowd that bought the original Pi, and those are the people who break out their Assembler and C skills and fix stuff that isn't working.
At this price I'd rather explore RISC-V in QEMU.
First we had red shirt Jeff. Now we have sunglasses Jeff in the thumbnail. I approve of adding characters to the Geerliverse.
Wow.. the AMD Radeon HD 7470. I worked on that chip when I was fresh out of college.
Jeff keeps killing these thumbnails!!! 😎😂🤣😂
Hi Jeff, thanks for the video, very interesting and thought-provoking. Given the lack of RPi's atm this looks like it could gain a lot of traction and interest in the soc space. Look forward to more videos from you on RISC soc's. Perhaps I should look at using these kind of RISC based soc's for the many projects I have planned. I am due to finally get my cytoreductive surgery on the 15th so I won't be around for 2 to 3 months but when I get back home I look forward viewing some more content from you. Great to see you looking so well, cheers Jeff.
Props for decent documentation. That's a good indicator this board will lead to something good eventually, if it's not quite there now.
That thumbnail is the stuff of legends - fine work, Jeff.
what a waste of time, I will go with something that works even if it is not the best SBC. It's like a car, I want one that will run faithfully. Not one on the side of the road in a snow storm.
Perfect timing on this since I'll be working more with my VisionFive 2 this weekend. It's been sitting around for a week as I work on other projects.
Thank you for getting in on this topic so early, the whole RISC-V thing seems like it has so much possibility, but it's all "potential energy" right now and needs to be converted to "kinetic", or something. A whole lot of bootstrapping going on!
I hope it will help democratize lower-end hardware, and maybe even challenge mid to high-end performance some day. It seems perfect for a lot of embedded uses... maybe. Will have to compete with ARM there. But why *not* compete with ARM there? Anything is possible. ARM's own origin story if very "something out of nothing", low expectations, big results.
So, once again, very interesting! Maybe RISC-V can chart that "bolt out of the blue" path again, like ARM once did!
"Until next time, I'm Jeff Geerling."
So who will you be next time?
7:58 I sense some ryan george DNA lurking
Speed of risc v development is quite astonishing. I wonder if they can keep it up for a few more years.
Hey bubbles bud it's PATRICK SWAZEY GET THE PUPPET!
I prefer the Gibson's architecture. I hear DaVinci runs really well on it now, and it comes with a wicked fractal decompiler / screensaver.
I hear people hack it though