Thank you for what you have done. No need to reply, but you are saving the world. Open source architecture will change the game. *Truly, thank you sir*
@@ragincaveman Hi James, first of all, thanks for being open and ready to talk with the community here. Many of us, especially the GNU Linux developers got big hopes, that SiFive and other uprising RISC-V Chip manufacturers take the chance, to not make the same errors like AMD, arm etc. and especially intel. I am speaking about the new SiFive P650 and it's so called 'advanced cryptographic features'. What is meant by that? Private users worldwide, especially big companies and governments don't need 'security hardware chips' which can get compromised and ironically be used as hardwired backdoors. We are all waiting for open source, transparent and trustworthy general purpose CPUs without 'secret functions', not something like intel's controversial IME, AMD's PSP, TPM, TPM 2, so called 'secure boot', EFI and the list goes on. Almost all of them have been compromised. Nobody knows, what is really going on inside security chips, which is a big trusting problem. They are really a stupid idea. The only reasonable option is, to write security chips as virtual devices, better in an isolated vm. In that way developers might verify their code and check for backdoors. But we can not do that with BLOBs. And i am wondering why are there cryptographic features in SiFive chips too? We really don't need those. On Linux we got exactly that, what i meant above as a best solution, a good old /dev/random which is a virtual device, which generates random numbers by collecting quantumnoise induced by radiation on several computer electronic components. This is the best method for unpredictable random numbers. And it just works. GNU Linux is practically the backbone of the entire internet. It is software, readable, verifiable and patchable and upgradeable. A hardware security chip on the other hands generates numbers, based on algorithms, which can be predicted and hacked. What is SiFive's explanation and your stance on this problematic situation. An answer is much appreciated. Thanks
@@aladdin8623 It's shorthand for SiFive Shield and SiFive Worldguard, SiFive's specification for security. It's not a hardware security chip black box. Watch previous presentations from SiFive at The Linley Conference and check out the tech papers on the website for more info (registration wall - name/email required).
I love these types of vids Wendell, I really appreciate you doing these types of things for us especially considering how youtube probably barely monetizes them (meanwhile I heard a preacher who directly counters most of the main doctrines gets literally thousands from ad revenue monthly on vids with 3k views each... and my local one got banned.. youtube is so fkn weird)
I've been messing with embedded Rust on a cheapo GD32 RISC-V microcontroller. It's been pretty competitive with a typical STM32 ARM board (aside from the lack of a FPU), which is impressive for such a new architecture running a new ISA.
@@jfltech For context, Gigadevice started producing the "GD32" line as clones of ST's STM32 line, with ARM SOCs having the same pinout, peripheral address layout, etc... as a STM32F103 (which are honestly somewhat dated designs.) These new GD32's are basically the same thing as those, but with the ARM core replaced with a RISC-V "Bumblebee" core licensed from the Chinese developer Nuclei System. Seeedstudio has some GD32 + ESP8266 development boards on Mouser that I considered buying, would've definitely pulled the plug if they had ESP32's.
With nVidia buying ARM, it's going to be interesting to see if big players like Amazon would back a more open ISA like RISCV. Amazon already have ARM in their data centers (see Graviton), and seem like the kind of company that can implement RISCV hardware and abstract it away for regular devs via high level APIs.
@Hugh G. Rection Yes, they are. The issue here is that they also require a larger volume of work to get an application running. Technology stacks are selected depending on the size of the team, the budget and the deadlines. If it's too difficult to use, it won't gain traction. There are a lot of really good technologies that didn't leave the world of academia or hobbyists simply because of these factors. It's better to get a product working and then continue with optimization rather than starting from a low level approach.
@Hugh G. Rection Sure, to an extent. But not everyone needs the absolute last ounce of performance from their hardware. If it can deliver say 90% of the hardware capability with abstraction and provide a far easier system, people won't care because it's "good enough".
25:34 notification noise had me looking through all the meeting applications i have online between the 3 remote desktop sessions i've got going for work to figure out how messaged me lol.
Being able to make specialized accelerators is exactly why Intel bought Altera and one of the reasons AMD is looking to buy Xilinx. Specialized accelerators (either via ASICs or FPGAs) can provide you some really excellent perf/watt or price/performance depending on the workload.
Great talk. Do more of these. Not enough people have these engaging convos with RiscV officianados, and it's easy to see that now RiscV is becoming the standard in Tertiary education institutions, because of its open-source dedication, this industry will dominate computer architecture industry after one generation. Academic shifts always indicate the future direction of associated industries.
This is the future. Nvidia paid a lot for ARM, but arms entire business model is maintaining control over the design. Same with Intel and AMD. This breaks the model.
I've been really interested in the new RISC-V stuff and the newer Linux capable stuff from sifive is exciting. I'm just hoping that someone can make some silicon available to others in the near term that isn't ludicrously expensive.
Conversation around the 30 minute mark made me think of RISC-V having relatively bolt-on viability in fuel refinery network infrastructure and the like.
Depends where you're coming from and what you need, but one important point is the industry momentum RISC-V is picking up right now. Basically "everyone" is working together to make RISC-V as good as it can possibly be (almost like with Linux), and I think we will see a lot of improvement and adoption over the coming years. It's really cool that PowerPC has been open sourced, but I believe that move might be a bit too late for it to catch on (it's kind of drowning in the hype).
I wonder if RISC-V will ever come to personal computing devices as the CPU (I think some drives have RISC-V controllers). It's all about custom silicon, which doesn't seem to have any overlap with the personal computing space at the moment.
It's just a matter of someone seeing a market worth the several billion dollars of investment needed to make a Skylake or Ryzen class process with any instruction set. Apple has done that with the ARM 64 bit instruction set because they know if they put it in Macs then all Mac users will buy one. If they'd chosen RISC-V instead of Aarch64 it would have worked just as well.
Love it. As ARM SBC user/developer it always feels like 1 step forwards, 2 steps back. Everytime we succeed in adding fuctionality there's always other things that break. That due to a lack of devs/testers/users/money/support from Ubuntu/Debian/Microsoft?. With mainline things seem to improve, slowly. But what would be done a year ago stil ain't working right now. ARM could have even taken mainstraim pc last years if only enough development was done. As long as VPU/GPU drivers are not free open-source it's not getting there. I've got hopes for RISC-V to be able to do better. I don't need the maximum possible performance. I just need the performance for my tasks. I rather have 10 x 5w devices each doing its task well. Than an x86 pc wasting energy constantly. Thank you, I'd love to review a SiFive SoC. Greetings, NicoD.
I’ve been on the fence with my Ubuiqiti stuff, especially with the recent firmware changes. I’m not sure where to go from where I am. Will check out the en genius stuff and see how it is.
At the start he says "so a lot of advantages there the open nature means that it's auditable it's extensible you don't have to you know guess what's inside". This is in contradiction to something he said only moments before, "we've got some other secret source things that we do for our architecture" If they have implemented it with their own secret source how can it be auditable? Like saying if the DX12 API was open source that we can audit how Microsoft had implemented it. Which would not be the case.
I understood it to be process technologies for reduction of build time while being open about their Isa. Everyone can see how the bike is made and the gears turn, but how is it mass-produced quickly?
@@Level1Techs I will have to dig into this a bit more. Looking at their site it is hard to see what of their product range is open. I understand open does not mean for free. The ISA is like the CPU version of a header. With it I can build a compatible CPU with zero licence fees. But in using the ISA do I have to release the silicon design? I do not think so and without that it would say it's difficult to audit. Undocumented instructions is what people are concerned with. Having an Open ISA does not stop this. Don't get me wrong, its a step in the right direction as to compete companies have to build the best implementation of the ISA. No need to pay big licence fees. I need to buy one to poke around with. :)
13:01 RISC-V and standardized do not fit together. The idea behind RISC-V boils down to: "This is the recipe to CPU Anarchy. How about we all agree to that? kthxbye" As soon as this takes off, it will run against a wall.
You need some standards for compatibility or you will end up with a bunch of stuff that does not work together. This modular method seems the best to get it adopted and actually used.
Yeah, if so much is optional and extensible, then it isn't likely to be consistent enough to become a commodity platform. Perhaps it will be used in custom embedded applications or something successfully.
How long before we see larger more performant SOCs. Most risc-v stuff is low power right now? I imagine a lot more engineering work is needed for bigger systems?
As someone from an emerging market, I would love a RISC-V SoC for schools. The problem currently is that computers here are too expensive, prohibiting that technology education in public schools is inviable. If someone here develops a RISC-V SoC at a reasonable price, it would be great to popularise computing throughout the country. Raspberry Pi is crazily expensive here for whatever reason. If local development allows for a viable alternative, I would definitely join the efforts (even though I currently have little Knology in the area)
I doubt that you could get much cheaper than a Raspberry Pi, or similar SBC. The reason why the Raspberry Pi has got its relatively low price is because it is using SoCs that are cheaper because they are considered too outdated to put in cell phones and such. I would look at older models or the smaller Pi Zero.
@@galdutro why does it cost that? taxes? that's a rough price. it's so rough, one could buy them at _retail_ here and turn a hefty profit selling them there. it must be some weird taxation.
@@slipcurve1410 probably! For the longest time in the 80s our government though they could simply stimulate our silicon industry through taxation. Guess what: it didn’t work! Maybe the raspberry is going through some sort of lingering legislation from the time.
@@slipcurve1410 but I really hope that RISC-V brings to a flourish local brands of silicon design. It will certainly accelerate innovation! My country already has somewhat of a lead in SiP design, even so that Qualcomm has a local office here to trade technologies in SiP design with local companies and also universities. I just hope that we can become more self sufficient in the technology sector overall, in case something goes wrong with the global trade “infrastructure” we have today! (Also to increase our “bargain power” in international trade and politics) I’m talking about Brazil btw.
Wendell's evil cousin here sounds pretty smart but it ALSO seems like he's the Shelbyville version of Wendell so i also feel like i shouldnt trust him as much.....like, hes remixed and I kinda always liked the OG W....anyone else get them feels...???? JK for real tho these types of convos are great, really, thx guys!
Don't know why, but I also refer to risc-v as old tech with new tooling :DD But I'm becoming more and more interested in other ISAs than x86 and arm as IoT and other low-powered chip(sets) become more and more useful (and even possible with low-power cores) in different applications.. And Risc-V is really interesting in this that you can design your own core, even cheaper than licensing arm cores and could be only and also everything you need in it, add/take away - config exactly what you need
Hey there I want to know that if I want to make a RISCV processor, then from where can I add DAC ,ADC and DSP unit in them beside Sifive boards cause they are costly
I was around folks 20 years ago that did incredible things with embedded Power32 + FPGA. Like linecards for petabit routers. My hope is that RISC-V will handle those capabilities and bring them up-to-date. And I would really like an OpenBSD box with auditable hardware and software. Yeah. I will echo comments around here: a lot of marketing chatter in this interview. I get it, but it wasn't necessary.
While I'm not a developer, programmer, or engineer, I find RISC-V super fascinating and exciting. I've been into science and engineering since I was a little kid but, somehow, the career I'm in is more artistic (I design and teach competitive marching band shows). I'm more of the overall ideas, creativity, and execution kind of guy. I'm working on a project now which I'm prototyping with Arduino and will eventually get custom PCBs made for final production.... (it's a bit of a weird project, and has to do with marching and performing arts.) What I find SUPER cool about RISC-V is the idea that eventually someone could go onto somewhere like Fiverr (five-er🤔 ) and commission an engineer to design actual silicon that can do exactly what they need for a particular application. Imagine if fabs and foundries end up something like the massive PCB houses are now, where you could send over the design and have prototypes manufactured. I mean... there's probably a LOT more to silicon photolithography than PCB manufacturing, but it could be an interesting way for fabs with outdated nodes to find new business.... Yeah, I'm sure that's all fantasy thinking, but still fun to imagine. Assuming the human race makes it much longer. :P
Overall good stuff but I wouldn't put too much trust in an open ISA being more secure against malicious manufacturers.(...like those backed by the currrent government of mainland Taiwan [ccp]) There isn't much to stop a company from adding a few of their own secret instructions along with the base IS.
James thinks like an embedded developer, which is why he says everyone can succeed without taking anyone else’s lunch - specialized systems with many different types of processors which are chosen based on the merits and little care is given to fancy chip marketing. It has nothing to do with being diplomatic, it’s just a different playing field. Wendell is coming at it from the ARM vs x86 and Linux vs Windows attitude... I think this interview was a miss because of that
I have to know, can I build me a chip,that will let me run an RTX 3090.....in my toaster?? Next level computing bro....RISC-V powered toasters that will mine bitcoin in between breakfasts!
So did ARM boards before the non-profit (and indeed subsidized) Raspberry Pi foundation came along. Existing and working at all is the big step. Driving the price down is then just a matter of time and economics. Note that there is now the $499 Icicle board with the same FU-540 penta-core complex as the HiFive Unleashed embedded in a 250k element FPGA. It's running slower (600 MHz), so close to the original Pi but with five cores instead of one and with a lot more RAM. Or like a Pi 2, but 2/3 the clock speed. Next year we'll probably see $250 boards simply by using a version of the PolarFire SoC with smaller FPGA part, once those become available, and a more focused set of peripherals on the board (the Icicle is pretty kitchen sink, with dual gigE, eMMC, SD, x4 PCIe ngen2, Raspberry Pi connector, mikroBUS, 2x CAN).
Is this the reason of selling more than 15 millions of PI for years ? RISC-V is a new world of CPU architect. It totale depends on from where you are looking ! Can you add new ISA to ARM based Broadcom loaded PI ? Again there is no unity, rather variety.
26:00 are you talking about "opencores" dot org? Well, even Alibaba had to add 50 custom instructions to make riscv work for them. I don't think it stands a chance against a mature ISA like PowerPC. If you want to make your own desktop pc, use power. Compilers support everything anyway. The problem that power has is the same that riscv has: Economy of scale. Raptor's Talos cost 3x what x86 does. And Riscv won't be cheaper. You can't beat billions of cores of Arm/X86 each year price wise. And if you want a new esotheric ISA research, look up "Forwardcom dot info". The guy behind that is the assembler God. I don't buy into the RiscV hype 46:50 Here is the important part for all of you risc-v fanbois that want a risc-v desktop/laptop/edgy. Not gonna happen
Great video....I think youtube thinks this is some sort of porn video....the amount of commercials...oh boy! Good thing the subject matter is interesting...
How am I supposed to design a 65536 bit word length processor with 64k to Exa layer hyper threading per core using RISC-V capable of managing 2^65536 bytes of ram with trillions / Exa CPU cores per processor capable of hardware level emulation of at least x86+64, ARM, and power9/X ISA operating at exa hertz+ optical / quantum system bus. a stupid FPGA capable of emulating a microcontroller is grossly stupid.
All hail the tech janitor!
Awesome seeing a UoP grad' on Level 1 James. 💪🏼
@@MrMolchester thanks craig!
Thank you for what you have done. No need to reply, but you are saving the world. Open source architecture will change the game. *Truly, thank you sir*
@@ragincaveman Hi James, first of all, thanks for being open and ready to talk with the community here. Many of us, especially the GNU Linux developers got big hopes, that SiFive and other uprising RISC-V Chip manufacturers take the chance, to not make the same errors like AMD, arm etc. and especially intel. I am speaking about the new SiFive P650 and it's so called 'advanced cryptographic features'. What is meant by that?
Private users worldwide, especially big companies and governments don't need 'security hardware chips' which can get compromised and ironically be used as hardwired backdoors. We are all waiting for open source, transparent and trustworthy general purpose CPUs without 'secret functions', not something like intel's controversial IME, AMD's PSP, TPM, TPM 2, so called 'secure boot', EFI and the list goes on. Almost all of them have been compromised. Nobody knows, what is really going on inside security chips, which is a big trusting problem. They are really a stupid idea. The only reasonable option is, to write security chips as virtual devices, better in an isolated vm. In that way developers might verify their code and check for backdoors. But we can not do that with BLOBs. And i am wondering why are there cryptographic features in SiFive chips too? We really don't need those. On Linux we got exactly that, what i meant above as a best solution, a good old /dev/random which is a virtual device, which generates random numbers by collecting quantumnoise induced by radiation on several computer electronic components. This is the best method for unpredictable random numbers. And it just works. GNU Linux is practically the backbone of the entire internet. It is software, readable, verifiable and patchable and upgradeable.
A hardware security chip on the other hands generates numbers, based on algorithms, which can be predicted and hacked.
What is SiFive's explanation and your stance on this problematic situation. An answer is much appreciated.
Thanks
@@aladdin8623 It's shorthand for SiFive Shield and SiFive Worldguard, SiFive's specification for security. It's not a hardware security chip black box. Watch previous presentations from SiFive at The Linley Conference and check out the tech papers on the website for more info (registration wall - name/email required).
James is such a good dude. I learned A LOT about system building from him while he was at AMD. Really stoked to see what RISC-V has to offer.
thanks for the kind words! :D
Senpai noticed you. 🤣
WHENDAL ASK HIM WHEN IS SIFIVE GOING INTO THE STOCK MARKET ????
37:20 wouldn't hurt sifive if Nvidia were to mess up with arm though.
oh shit. didnt know another youtuber i follow watched level1techs. awesome
I love these types of vids Wendell, I really appreciate you doing these types of things for us especially considering how youtube probably barely monetizes them (meanwhile I heard a preacher who directly counters most of the main doctrines gets literally thousands from ad revenue monthly on vids with 3k views each... and my local one got banned.. youtube is so fkn weird)
WHENDAL ASK HIM WHEN IS SIFIVE GOING INTO THE STOCK MARKET ????
Wendel ask Evan why HE TYPES IN CAPS
Who's whendal, is it wendals procrastinating twin brother. 🤔
You never know when whendal will be here!
Thanks for increasing exposure to RISC-V. What the guys at SiFive are doing really is going to change the industry and the world. Much love Level1
I've never heard of this but really enjoyed the conversation
I've been messing with embedded Rust on a cheapo GD32 RISC-V microcontroller. It's been pretty competitive with a typical STM32 ARM board (aside from the lack of a FPU), which is impressive for such a new architecture running a new ISA.
Interesting.. haven't heard of GD32 .. been mostly fixated on ESP32 for my IoT at the moment... thanks
@@jfltech For context, Gigadevice started producing the "GD32" line as clones of ST's STM32 line, with ARM SOCs having the same pinout, peripheral address layout, etc... as a STM32F103 (which are honestly somewhat dated designs.) These new GD32's are basically the same thing as those, but with the ARM core replaced with a RISC-V "Bumblebee" core licensed from the Chinese developer Nuclei System.
Seeedstudio has some GD32 + ESP8266 development boards on Mouser that I considered buying, would've definitely pulled the plug if they had ESP32's.
With nVidia buying ARM, it's going to be interesting to see if big players like Amazon would back a more open ISA like RISCV. Amazon already have ARM in their data centers (see Graviton), and seem like the kind of company that can implement RISCV hardware and abstract it away for regular devs via high level APIs.
@Hugh G. Rection And without high level APIs, your platform is dead. It's impractical to implement everything from the ground up, all of the time.
@Hugh G. Rection Yes, they are. The issue here is that they also require a larger volume of work to get an application running. Technology stacks are selected depending on the size of the team, the budget and the deadlines. If it's too difficult to use, it won't gain traction. There are a lot of really good technologies that didn't leave the world of academia or hobbyists simply because of these factors. It's better to get a product working and then continue with optimization rather than starting from a low level approach.
@Hugh G. Rection Sure, to an extent. But not everyone needs the absolute last ounce of performance from their hardware. If it can deliver say 90% of the hardware capability with abstraction and provide a far easier system, people won't care because it's "good enough".
25:34 notification noise had me looking through all the meeting applications i have online between the 3 remote desktop sessions i've got going for work to figure out how messaged me lol.
"add lightness"
if you want to go fast, add lightness. good philosophy i never considered would translate to silicon.
Specialised chips like this sound really cool, and the possibilities are pretty exciting.
Being able to make specialized accelerators is exactly why Intel bought Altera and one of the reasons AMD is looking to buy Xilinx. Specialized accelerators (either via ASICs or FPGAs) can provide you some really excellent perf/watt or price/performance depending on the workload.
The first 30 mins felt like a marketing talk, than a technical explainer/Q&A
I have apparently been living under a rock. Thanks for getting it off, now I can see the way forward!
Enjoyed the conversation. Opened my eyes to RISC-V. Time to do more research
Great talk. Do more of these. Not enough people have these engaging convos with RiscV officianados, and it's easy to see that now RiscV is becoming the standard in Tertiary education institutions, because of its open-source dedication, this industry will dominate computer architecture industry after one generation. Academic shifts always indicate the future direction of associated industries.
Hi RISC-V developers. Just google "SAVVY-V". It will be the first stackable and daisy-chainable board which flexible enough for many application.
This is the future. Nvidia paid a lot for ARM, but arms entire business model is maintaining control over the design. Same with Intel and AMD. This breaks the model.
I've been really interested in the new RISC-V stuff and the newer Linux capable stuff from sifive is exciting. I'm just hoping that someone can make some silicon available to others in the near term that isn't ludicrously expensive.
RISC-V is so much the future. I'm gonna nerd out about it wherever I can.
i JUST started using arm as a daily driver platform, now i've gotta start all over again on risc-v?
Conversation around the 30 minute mark made me think of RISC-V having relatively bolt-on viability in fuel refinery network infrastructure and the like.
Looking forward to the new replacement for the Hi5 Unleashed board. Have a few RISC-V boards but want one that can run Linux.
So why I should use Risc-V not PowerPC ( now open sourced )
Depends where you're coming from and what you need, but one important point is the industry momentum RISC-V is picking up right now. Basically "everyone" is working together to make RISC-V as good as it can possibly be (almost like with Linux), and I think we will see a lot of improvement and adoption over the coming years.
It's really cool that PowerPC has been open sourced, but I believe that move might be a bit too late for it to catch on (it's kind of drowning in the hype).
I wonder, when can we expect cheap and powerful risc-v microcontrollers similar to ESP8266/ESP32?
ESP32-S2 already has one RISC-V core and one xtensa core. The next version might well be pure RISC-V as Espressif seems pretty keen on RISC-V.
I wonder if RISC-V will ever come to personal computing devices as the CPU (I think some drives have RISC-V controllers). It's all about custom silicon, which doesn't seem to have any overlap with the personal computing space at the moment.
It's just a matter of someone seeing a market worth the several billion dollars of investment needed to make a Skylake or Ryzen class process with any instruction set. Apple has done that with the ARM 64 bit instruction set because they know if they put it in Macs then all Mac users will buy one. If they'd chosen RISC-V instead of Aarch64 it would have worked just as well.
Please bring him back, and talk about the new SiFive unmatched development board.
I wonder if AMD is trying to grab a slice of the Risc-V Market by designing their own Risc-V chip with integrated Xilinx FPGA on board.
Love it. As ARM SBC user/developer it always feels like 1 step forwards, 2 steps back. Everytime we succeed in adding fuctionality there's always other things that break. That due to a lack of devs/testers/users/money/support from Ubuntu/Debian/Microsoft?.
With mainline things seem to improve, slowly. But what would be done a year ago stil ain't working right now. ARM could have even taken mainstraim pc last years if only enough development was done. As long as VPU/GPU drivers are not free open-source it's not getting there.
I've got hopes for RISC-V to be able to do better. I don't need the maximum possible performance. I just need the performance for my tasks. I rather have 10 x 5w devices each doing its task well. Than an x86 pc wasting energy constantly. Thank you, I'd love to review a SiFive SoC. Greetings, NicoD.
Good job youtube I am already 5 min into the video when I get a notification 😂😂
You got a notification at least. :)
I thought my speakers were broken for the first 10-12 seconds
Nice chat! But gosh is James Prior's audio rough :(
Wooooo! SiFive baby!!!! Open source hardware!? Running linux? Welp,there goes there goes the neighborhood!
I’ve been on the fence with my Ubuiqiti stuff, especially with the recent firmware changes. I’m not sure where to go from where I am. Will check out the en genius stuff and see how it is.
As compared to RISC-V, how much performance is being left on the table by x86 processors nowadays?
At the start he says "so a lot of advantages there the open nature means that it's auditable it's extensible you don't have to you know guess what's inside". This is in contradiction to something he said only moments before, "we've got some other secret source things that we do for our architecture" If they have implemented it with their own secret source how can it be auditable? Like saying if the DX12 API was open source that we can audit how Microsoft had implemented it. Which would not be the case.
I understood it to be process technologies for reduction of build time while being open about their Isa. Everyone can see how the bike is made and the gears turn, but how is it mass-produced quickly?
@@Level1Techs I will have to dig into this a bit more. Looking at their site it is hard to see what of their product range is open. I understand open does not mean for free. The ISA is like the CPU version of a header. With it I can build a compatible CPU with zero licence fees. But in using the ISA do I have to release the silicon design? I do not think so and without that it would say it's difficult to audit. Undocumented instructions is what people are concerned with. Having an Open ISA does not stop this. Don't get me wrong, its a step in the right direction as to compete companies have to build the best implementation of the ISA. No need to pay big licence fees. I need to buy one to poke around with. :)
You should have asked how much the Sifive PC is gonna cost.
13:01 RISC-V and standardized do not fit together.
The idea behind RISC-V boils down to: "This is the recipe to CPU Anarchy. How about we all agree to that? kthxbye"
As soon as this takes off, it will run against a wall.
You need some standards for compatibility or you will end up with a bunch of stuff that does not work together. This modular method seems the best to get it adopted and actually used.
@@Greenwithao It's not modular though. It's do whatever you want. Which will ultimately result in "a bunch of stuff that does not work together".
Yeah, if so much is optional and extensible, then it isn't likely to be consistent enough to become a commodity platform. Perhaps it will be used in custom embedded applications or something successfully.
@@shanehebert396 *"custom embedded applications"* that's exactly where I see it.
If it grows enough, a standard or a few standards will necessarily emerge.
How long before we see larger more performant SOCs. Most risc-v stuff is low power right now? I imagine a lot more engineering work is needed for bigger systems?
Please check out SAVVY-V cluster board for more processing power. We are 2 guys working on it.
Twitter : @riscV_SAVVY
34:14 "In the semiconductor industry I can see this shrinking thing" --- it's called Moore's Law! 😜
Just waddled into the livingroom glancing at the TV as I walk by, then I thought "What the hell is Wendel talking to CinnamonToastKen for?".
What option will give me a mad beard like James?
Nice. Been watching and reading alot about RISC-V after talking to some guys from ultra soc at DAC this year.
great to hear these guys are getting a foothold
he says they're not trying take on the general purpose computing market, but i hope some day they will.
As someone from an emerging market, I would love a RISC-V SoC for schools. The problem currently is that computers here are too expensive, prohibiting that technology education in public schools is inviable. If someone here develops a RISC-V SoC at a reasonable price, it would be great to popularise computing throughout the country. Raspberry Pi is crazily expensive here for whatever reason. If local development allows for a viable alternative, I would definitely join the efforts (even though I currently have little Knology in the area)
I doubt that you could get much cheaper than a Raspberry Pi, or similar SBC.
The reason why the Raspberry Pi has got its relatively low price is because it is using SoCs that are cheaper because they are considered too outdated to put in cell phones and such.
I would look at older models or the smaller Pi Zero.
@@FindecanorNotGmail yeah. but a Rasberry here costs around 100 dollars. I can get a KaiOS featurephone powered by an Snapdragon SoC for half that!
@@galdutro why does it cost that? taxes? that's a rough price. it's so rough, one could buy them at _retail_ here and turn a hefty profit selling them there. it must be some weird taxation.
@@slipcurve1410 probably! For the longest time in the 80s our government though they could simply stimulate our silicon industry through taxation. Guess what: it didn’t work! Maybe the raspberry is going through some sort of lingering legislation from the time.
@@slipcurve1410 but I really hope that RISC-V brings to a flourish local brands of silicon design. It will certainly accelerate innovation! My country already has somewhat of a lead in SiP design, even so that Qualcomm has a local office here to trade technologies in SiP design with local companies and also universities. I just hope that we can become more self sufficient in the technology sector overall, in case something goes wrong with the global trade “infrastructure” we have today! (Also to increase our “bargain power” in international trade and politics)
I’m talking about Brazil btw.
Wendell is so cool and wicked smart. His voice is so calming. Its like listening to those thunderstorm videos before bed.
Wendell's evil cousin here sounds pretty smart but it ALSO seems like he's the Shelbyville version of Wendell so i also feel like i shouldnt trust him as much.....like, hes remixed and I kinda always liked the OG W....anyone else get them feels...????
JK for real tho these types of convos are great, really, thx guys!
Don't know why, but I also refer to risc-v as old tech with new tooling :DD But I'm becoming more and more interested in other ISAs than x86 and arm as IoT and other low-powered chip(sets) become more and more useful (and even possible with low-power cores) in different applications.. And Risc-V is really interesting in this that you can design your own core, even cheaper than licensing arm cores and could be only and also everything you need in it, add/take away - config exactly what you need
Hey there I want to know that if I want to make a RISCV processor, then from where can I add DAC ,ADC and DSP unit in them beside Sifive boards cause they are costly
Excellent content!
I was around folks 20 years ago that did incredible things with embedded Power32 + FPGA. Like linecards for petabit routers. My hope is that RISC-V will handle those capabilities and bring them up-to-date.
And I would really like an OpenBSD box with auditable hardware and software. Yeah.
I will echo comments around here: a lot of marketing chatter in this interview. I get it, but it wasn't necessary.
It can.
@@aliuzel4211 Just followed your name to the SAVVY-V information. Looks good!
@@stephanematis Thank you Stephane. Hopefully test card will be available Q1 2021.
While I'm not a developer, programmer, or engineer, I find RISC-V super fascinating and exciting.
I've been into science and engineering since I was a little kid but, somehow, the career I'm in is more artistic (I design and teach competitive marching band shows). I'm more of the overall ideas, creativity, and execution kind of guy. I'm working on a project now which I'm prototyping with Arduino and will eventually get custom PCBs made for final production.... (it's a bit of a weird project, and has to do with marching and performing arts.)
What I find SUPER cool about RISC-V is the idea that eventually someone could go onto somewhere like Fiverr (five-er🤔 ) and commission an engineer to design actual silicon that can do exactly what they need for a particular application. Imagine if fabs and foundries end up something like the massive PCB houses are now, where you could send over the design and have prototypes manufactured. I mean... there's probably a LOT more to silicon photolithography than PCB manufacturing, but it could be an interesting way for fabs with outdated nodes to find new business....
Yeah, I'm sure that's all fantasy thinking, but still fun to imagine. Assuming the human race makes it much longer. :P
Not being very into this field I at first thought you were going to talk about SciFi and that RISC-V falls into that topic
Wrapping an 8051 with a RISC V system is like stuffing a cow with A shrimp. (singular)
21:00 - the interesting part. Thank me later
thank you !
Not all heros wear capes, some choose a beard instead.
So can I make my own chip?
my understanding, is that you will probably be renting computationalpower from the cloud in the future. Well my mind is stuck in local power, hehe
wow thats some amazing development
Overall good stuff but I wouldn't put too much trust in an open ISA being more secure against malicious manufacturers.(...like those backed by the currrent government of mainland Taiwan [ccp]) There isn't much to stop a company from adding a few of their own secret instructions along with the base IS.
What if the worst had nvidia arm deal fails & nvidia goes to the newly risc-v?
This is Awesome
I have not been living under Erock (Erack).
I want that new fpga dev board
So many odd cuts with animations in this video.
James thinks like an embedded developer, which is why he says everyone can succeed without taking anyone else’s lunch - specialized systems with many different types of processors which are chosen based on the merits and little care is given to fancy chip marketing. It has nothing to do with being diplomatic, it’s just a different playing field.
Wendell is coming at it from the ARM vs x86 and Linux vs Windows attitude... I think this interview was a miss because of that
I have to know, can I build me a chip,that will let me run an RTX 3090.....in my toaster?? Next level computing bro....RISC-V powered toasters that will mine bitcoin in between breakfasts!
Better idea: using the heat produced by mining chips to toast bread
@@NullByte_-mm4dn actually a space heater for winter season would be nice.
wendell! LOL I haven't watched you since tek syndicate.
Are you twins?
Nice talks, but a HiFive board that replaces my Raspberry Pi costs $999 👎
So did ARM boards before the non-profit (and indeed subsidized) Raspberry Pi foundation came along. Existing and working at all is the big step. Driving the price down is then just a matter of time and economics. Note that there is now the $499 Icicle board with the same FU-540 penta-core complex as the HiFive Unleashed embedded in a 250k element FPGA. It's running slower (600 MHz), so close to the original Pi but with five cores instead of one and with a lot more RAM. Or like a Pi 2, but 2/3 the clock speed. Next year we'll probably see $250 boards simply by using a version of the PolarFire SoC with smaller FPGA part, once those become available, and a more focused set of peripherals on the board (the Icicle is pretty kitchen sink, with dual gigE, eMMC, SD, x4 PCIe ngen2, Raspberry Pi connector, mikroBUS, 2x CAN).
Is this the reason of selling more than 15 millions of PI for years ? RISC-V is a new world of CPU architect. It totale depends on from where you are looking ! Can you add new ISA to ARM based Broadcom loaded PI ? Again there is no unity, rather variety.
Rome wasn't built in one day.
nice!!!
Get me a RISC-Very Pi
James I have a feeling that you could get out a chip from your beard sir...Magic beard 😎😎
26:00 are you talking about "opencores" dot org?
Well, even Alibaba had to add 50 custom instructions to make riscv work for them. I don't think it stands a chance against a mature ISA like PowerPC.
If you want to make your own desktop pc, use power. Compilers support everything anyway.
The problem that power has is the same that riscv has: Economy of scale. Raptor's Talos cost 3x what x86 does. And Riscv won't be cheaper. You can't beat billions of cores of Arm/X86 each year price wise.
And if you want a new esotheric ISA research, look up "Forwardcom dot info". The guy behind that is the assembler God.
I don't buy into the RiscV hype
46:50 Here is the important part for all of you risc-v fanbois that want a risc-v desktop/laptop/edgy. Not gonna happen
Great video....I think youtube thinks this is some sort of porn video....the amount of commercials...oh boy! Good thing the subject matter is interesting...
This sounds like something that Nintendo might be interested in, considering security seems to be one of their big concerns.
Jame's microphone is super awful Wendel, send him one of yours!
If you checked your volume right at the start of the vid #checkvolumesquad
WHENDAL ASK HIM WHEN IS SIFIVE GOING INTO THE STOCK MARKET ????
How am I supposed to design a 65536 bit word length processor with 64k to Exa layer hyper threading per core using RISC-V capable of managing 2^65536 bytes of ram with trillions / Exa CPU cores per processor capable of hardware level emulation of at least x86+64, ARM, and power9/X ISA operating at exa hertz+ optical / quantum system bus.
a stupid FPGA capable of emulating a microcontroller is grossly stupid.