Dr. Ian Cutress Explains The Hype Around RISC-V
ฝัง
- เผยแพร่เมื่อ 6 มิ.ย. 2023
- RISC-V seems to get a lot of hype, so Gordon asked Dr. Ian Cutress from @TechTechPotato what the big deal is.
*This video is sponsored by SilverStone, the leader in small-form factor power supplies for over 10 years. The SX1000R Platinum is the company's SFX 12V 4.0 update to the pioneering SX1000 Platinum released in 2020. It now includes the Gen 5 12VHPWR connector, and features ATX 3.0 performance updates and quieter operation. Pick up the SX1000R Platinum today and give your small form factor PC the power supply it needs to tackle the latest CPUs and GPUs!
Buy PCWorld merch: crowdmade.com/collections/pcw...
Follow PCWorld for all things PC!
--------------------------------
SUBSCRIBE: th-cam.com/users/pcworld?sub_c...
TWITCH: / pcworldus
TWITTER: / pcworld
WEBSITE: www.pcworld.com
#riscv #computex2023 #sponsored - วิทยาศาสตร์และเทคโนโลยี
RISC-V is the Linux of CPU architecture/instruction sets. It's important.
And as said in the video it's fragmented like Linux so it won't be widely successful.
@@bslay4rIt already *is* successful. Just not in the consumer space.
Fragmentation is kind of a strawman. x86 was fragmented too. All it takes is a single vendor deciding to set the standard and then it will happen.
Ubuntu was really almost there, then they went all Canonical and turned off so many people to it. Ubuntu 14.04 made me feel it was the distro to rule 'em all, then 16.04 ruined it and it gets worse every year 😔
@@bslay4r the fragmentation issue on Linux is quite negligible nowadays for an application dev. You just have to make your program or API opensourced or use some distro-agnostic package system like Flatpak or Appimage with whatever UI API you've chosen to use.
And for a driver dev, you simply don't care about fragmentation as you just follow the kernel development.
@@bslay4r Why do you guys keep harping on "fragmentation", as if ARM or X86 aren't fragmented? Those architectures have so much baggage you have no idea what you're talking about.
@@bslay4rIt’s not fragmented, though. People who say that haven’t done their homework.
It was a real loss to the community when Ian left Anand but I'm glad to see he's still out here making content for us.
@Erik, don't think he ever left honestly. :)
I didn’t know he left. Love him and his YT and Twitter.
well I'd agree about the "content for us" part few months ago but now his TTP videos turned to "content for my contractors" unfortunately.
@@PainterVierax I've been doing a lot of internal stuff and travel recently. Almost 60k miles. But have a video editor now. YT is really only meant to be 25% of my time, so it comes in bursts. Also I only take projects I think are genuinely interesting. Otherwise I say no
Anandtech has fallen off the cliff without him and Andrei
For the RISC-V fragmentation bit, there are standards now!
They call them Profiles. That's what software will be expected to target in the future, if successful, and that's the minimum base that most hardware will want to support.
Sort of, they don't work in the same way as a nailed down standard. The issue with profiles is that depending on the usecase, you can mix and match profiles to your hearts content.... What that does is build a chip to your exact needs but that also means that no two RISC-V cores are alike....
Unlike say ARM which defines it's ISA and has a standards base (which they call versions) for it, so Apple (as a licensee of ARM cores) can build what it likes but ARM literally checks for anything that is non-conforming to the ISA. Apple can build on top of that but if it's chip steps out of a version of ARM ISA, it's not an ARM chip anymore and can't be sold (and if they try, out come the lawyers).
Is it a concern for RISC-V adoption? Maybe but for the moment since this is a thing that you are seeing used in micro-controllers and dev units where this kind of fragmentation is actually a good thing. Once you start taking about consumer-facing units, the game will have to change and what RISC-V International does to address that will be VERY important... but for now it doesn't matter.
@@DuvJones Just as a note, it's important not to mix up extensions (which you can mix and match as you want) and the new RV Profiles which are a different thing entirely
A hardware core would target a single profile. For instance, a big Linux capable core might target RVA23, and that defines the base of required, optional, or incompatible extensions that the core must comply with.
What RV profiles are *is* the standard base, the nailed down specification that software can target, without worrying about variation and fragmentation, so that it's all much simpler than before.
If you want to make a big Linux-class application core, you are expected to target the RVA profile.
If it's an embedded core where you still want to benefit from the software ecosystem, but not quite a huge application class core, you target RVB.
If it's a microcontroller, then the base is RVM.
@@DuvJones you apparently don't understand what a "profile" is. A RISC-V applications processor profile e.g. RVA20 (previously known as RV64GC), RVA22, RVA23 ... is a precise set of ISA extensions that your core *must* support in order to be marketed as e.g. RVA22. It is a monotonically-increasing set of extensions, just as ARMv8.1-A, ARMv8.2-A, ARMv8.3-A etc is. No difference.
@@DuvJonesthey work exactly the same way as other specified platform levels, except there have been no deprecations yet so they're all backward compatible.
Does it works the same as Arm SystemReady?
The moment you mentioned Jim Keller I know that whatever he works on will be successful . That guy has the golden touch
Jim Keller is a journeyman computer architect. Patterson has been slapping the monkey in academia his whole life. Patterson has been working on an ISA for forty freakin' years.
Years ago I took a class with Dr. Patterson, one of the guys who came up with RISC-V. I'm sure he's very pleased by the acceptance of the architecture thus far.
Calling Patterson "one of the guys who came up with RISC-V" is the understatement of the century. You've been very lucky to study such an S-tier environment!
He was one of the guys that came up with RISC I, II, III, IV, and now V. They practically coined the term.
They're a huge early proponent and caused the RISC boom in the early 80's that influenced MIPS, SPARC, ARM, and others to exist in the first place.
Balanced and informative. Good questions. Thanks for not talking-over the expert. :)
love this type of coverage, thank you pcworld channel
Many years ago Intel sealed the fate of x86 by ignoring the low power processor market. They were so deep into CISC that they felt that more instructions was the way to continue capturing value. So folks like me and Apple switched to ARM. That was all well and good except for ARM's licensing games, so I've moved to RISC-V where I don't have to worry about this anymore. I currently have microcontrollers running RISC-V and it's been a seamless transition. I expect the same as we get to larger processors.
But why I really think RISC-V will win is as follows:
1. Linux has proven that open source works
2. The rate of innovation is higher with more players
3. Because of its block architecture, fabrication costs are lower because you only need transistors for the modules you include
4. No ISA licensing fees or worries about future shenanigans by the license holder as we see with ARM today
5. Greater security as fewer instructions reduce the attack surface. Generating code with compilers makes security patches easier
6. As mentioned in the video, the Chinese have an intense interest in getting away from licensing restrictions as levers to make them bend to international will
This time has been long coming for a part of the industry that still remains mostly proprietary. I first encountered RISC in the early 1980s and was immediately sold on the architecture. It's taken a long time to get RISC into the mainstream with ARM. But no one wants ARM to become the next Intel by protecting what it has from innovation. As always happens in the computer industry, once an ecosystem becomes large (Windows, x86, IBM, and now ARM) and is controlled by a few players then those entities switch from innovation to defense of their technology to retain their dominance. Open source fixes this problem. And I wouldn't worry about fragmentation. Linux has shown that good ideas stick around and bad ideas get kicked to the curb. It's market forces in action.
Thank you for stating your experience with RISC and listing sensible reasons for why it is important. Your point about ecosystems rings true throughout most major industries internationally and should be more of a talking point as years pass.
Honestly RISC-V does have me curious but I will agree with Dr. Ian in saying that consumer devices using RISC-V are a long way off. I think the use you are going to see a lot of is security enclaves, the ones that you see currently use low-power ARM systems and RISC-V seem just poised to take over given the licensing issues ARM is enducing.
To your points:
1. "if you can keep it." Linux has proven that open source works... so long as the community behind it keeps doing what they do. Which means that the tech industry has to do something that is alien to it: Understand people.
3. You know, as the fab world moves to more 3D architecture... this is going to be a lot more important.
4. That is both good and bad. ARM tends to have leverage to pull the industry in a direction, that kind of power can't be overstated especially now. RISC-V doesn't have that (and might NEVER have that)... we can only see what the future holds but this will likely come up.
5. This is why I look at the growth of security enclaves as something of a boon for RISC-V. You can just grab what you need and build on top of the rest without having to conform to anything else. That will keep attack surfaces lower, which gets interesting when no two RISC-V cores will be the same.
6. Ah yes, China.... You know that with the geopolitics tightening around China (and similarly Russia) that RISC-V International is going to have to navigate some rather choppy waters because the US's approach has been to stall China's tech acquisition and it is getting more intense these days. Simply put, China can't go ignored and neither can the US's reaction to the nation... and what that means for RISC-V is a complete unknown till something happens.
@@DuvJones Depends on what consumer devices we are talking about. Risc-V is already extremely popular among microcontroller manufacturers, even the Westerners. And like Linux with embedded, you probably already use devices powered by some Risc-V core without knowing it.
RISC-V is an open standard like USB, that doesnt mean it necessarily is open source. Lots of close source RISC-V processors.
But you do need an open standard of the instruction set in order for the processor itself can be open source.
It is possible someone makes a specific RISC-V processor and they licens it out like ARM but it will be easier to make it work with a different RISC-V processor where ARM is in control of any ARM based processor, so issue of one company having too much control is less likely with an open standard like RISC-V than the closed ones.
@@AndersHass you make a good point about open standard vs. open source. I will try to be more careful in the future.
Happy that Gordon is getting recommended to me more often. He's so pleasant to watch and listen to.
In the embedded MCU space, Risc-V can also be "interesting", like when Espressif is now making more and more chips based on this architecture, without involving Arm licences. Absolutely not high performance, but quite useful for IoT purposes and such. And hobbyists. I've also got a number of boards with ESP32-C3 which is the first of them, and with suitable firmware, it works great.
I've bought a bag of CH32V003s, tiny _tiny_ 10 cent MCUs that run at 48mhz.
They implement risc-v's RV32EC ISA.
I'm going to use them to drive a digital sound synthesizer IC from analogue inputs.
I'm quite excited, I have dreams of building a teeny little general purpose computer from them too
Also interesting when ur esp32 designs just crash due to shoddy design in the power circuitry and horrible async libs.
Esps are cheap the titan of mcus really.
I’ve learned over years not to trust an esp32. Never had those issues with PIC Atmel Parallax any.
@@MostlyPennyCatten cents does seem compelling. Are they WORM like those 1 cent paduacs?
I've got many ESP32 running, both original, S2, S3 and C3, and they all works fine for my use cases, except for two early sample ESP32-C3 revision 2 no longer supported.
The performance isn't that much worse than ARM chips that have been through way more years of refinement. There's also the issue of poorly optimized software, but that will improve as well. For these sorts of low end applications, it simply might not matter enough, and the financial incentive of avoiding the license fees is worth the switch.
I don't watch many PC hardware videos, but I definitely enjoy videos from Gordon at PCWorld
How fast RISC-V will gain market share in the consumer-facing space is mainly a matter of how greedy ARM will dare to get. ARM ownership changing hands every few years with the looming possibility that it may get bought out by one of your competitors and price you out of the market over the next 5-10 years would be another reason for anyone who depends heavily on ARM for their business to get RISC-V migration plans ready just in case and arrange existing work accordingly.
ARM is certainly not greedy. Considering the ubuiquity of it's IP, they make almost no money. You really think that couple of cents from your purchase going to them is unjustified? You are right about issues with ownership, they should never have been allowed to be sold.
You talk like the riscv is free but it is not, design and make a riscv processor doesn't necessary cheaper than ARM processor.
Every company in the industry is rapidly trying to build up in-house RISC-V expertise on both the hardware and software side. They all see the writing on the wall. The real question is, when is the right time to make the switch. Some companies will take the risk on it earlier, maybe too early. Others will be too late. But the change is inevitable and they all know it.
Listened very intently but had my eyes on the girl in the white dress the whole interview.
Haha, same.
Likely the role of global politics and international trade barriers will be the determining factors in if RISC-V will be a success. Open Source though, Love it! Peace.
Noted, shame if the case.
RISC-V is an instruction set architecture rooted within RISC (reduced instruction set computer) principles. This processor is very unique and also revolutionary as it is a free, common, and open-source ISA where hardware can be developed, the software can be ported & processors can be designed to support it.
THE RISK-V 10 CENT PROCESSOR be great.....
The open source HYPE HYPE HYPE people. Just make sure to hype it up when it counts most, as its forgotton.
@@SebLukaGaming I agree.
Not for or against RISC-V, but I think it'll be down to any adoption in China if RISC-V makes it to the big time. This is down to a move away, forced or voluntarily, from US or Western tech, hardware and software, such as x856 ARM, Microsoft Windows, maybe also Apple and Android. It will be RISC-V and some form of Linux as the OS. If China starts to mass manufacture these and begins to sell to the world, it will be a game changer. China could even donate millions of new RISC-V Linux based PCs to the developing world to kick start it too. I'm sure India could do much the same too. I give it a timeframe of about 5 years though to the mainstream consumer.
China should donate that to who? The US that cut em out of 5G for the most part? Or the rest of the western countrys that deliver one massive shitshow all around the globe? Dream along...
Google is working on Android for RISC-V.
If they CCP gives you something for "free" then best bet there is backdoors installed. The CCP is not some benevolent force. Look what they do to their own citizens. They harvest the organs of dissidents and sell them to ultra filthy rich people.
I have to agree. It's important to take RiscV out of the context of the Chinese-US trade war that it's currently framed in, and to put it into the context of worldwide tech development. We're witnessing the breaking down of a license paywall for enterprises, research initiatives and indie labs for functional and powerful computing hardware. Now you don't have to have robust US-based trade relationships to multiple megacorps to make powerful computers. That's a good thing for everyone.
China definitely wants to switch everyone over to RISC-V. They're investing heavily in it. Europe is too. There are US RISC-V companies, though, and they will keep developing it as well. Nobody actually likes ARM, it was better than Intel garbage but it was a lesser of two evils. RISC-V can be an actual good.
Thank you Dr Ian and Thank you Cameraman!
So then it isn't a hype. RISC-V is cutting out its own specialized market segment.
And increasingly cutting into ARM's market as well. The speed it's doing these things is much faster than ARM managed to do.
Thanks guys, very insightful! Thin clients, of cpurse!
(*"course", but that typo was just too good xD)
it's got lots of competition! it's a super view of where aggressive the chip market stands! good luck!
Horse Creek looks like a really good chip for Network Attached Storage devices to replace the Jasper Lake cores used today.
Great interview. This is what it looks like when knowledgeable people speak. Thanks.
He's being very conservative in his estimates though, and treating RISC-V like it's going to copy ARM's adoption speed. RISC-V is already way ahead of where ARM was this early in its development.
I didn't expect less from PCWorld. RISC-V ISA has a very fast adoption by the industry and a free software stack. When they start fearmongering is when you know is about to happen.
The guy says no major firms behind it, but he conveniently forgets or does not even know that Alibaba are into RISC5
The adoption rate is incredibly fast compared to even ARM, which was already quite fast.
This is the thing though, right. My LSI SATA HBA has a PowerPC core in it, because PowerPC cores are super cheap and they are a form of RISC. RISC-V may easily be what replaces small embedded cores like that for the foreseeable future, and it's going to be really interesting because RISC-V is so much more customisable.
PowerPC is dead and has been for quite a while are you talking about POWER ISA?
@@Leonard_MT PowerPC isn't dead. Inform yourself instead of being narrow-minded! It is dominating in the high end of the Automotive industry. The world is not only the things that you can directly observe in your own sphere.
@@rursus8354 Well from my understanding PowerPC has been replaced by POWER ISA, however POWER ISA is commonly called PowerPC.
Looking sharp Gordon!
Good review, thanks. I do embedded and I look to RISC-V as a possible ARM competitor in embedded.
It already is.
The TH-cam channel "Explaining Computers" 6 month ago used a Risk 5 computer running Linux for a week as his only computer.
He did emails, editing videos, word processing, watching TH-cam videos.
IMO, Risk 5 will be successful in industry and with the public but like Linux it will have a limited market share.
It may be of interest to governments concerned about back doors in X86 and ARM processors.
RISC-V will have a major part of the market share of phones within a couple years. It'll maybe have all of the market share for phones and tablets within 10 years. It will be much slower at taking over the desktop, and particularly workstation markets. Low end desktops will probably see RISC-V increasingly as that market shrinks overall.
It wouldn't terribly surprise me either if the CCP basically forces their whole country to switch to RISC-V based computers running some kind of Linux-based kernel as soon as they can realistically do that.
@@fakecubedI've just watched a video by Alex Ziskind on TH-cam about a new CPU from Qualcomm called the Snapdragon X Elite ZH-WXX.
This CPU that they are producing for laptops is performing better than the M2 chip from Apple.
In the video he said that it's based on ARM architecture but didn't say that it was a RISK 5 processor, however on Linus Tech Tips he was demonstrating a computer that has a Risk 5 processor made by Sophon.
Something that was very interesting in that video was that Qualcomm are getting interested in making RISK 5 processors, hearing that I'm wondering if Qualcomm are working faster on RISK 5 processors than they are admitting.
Could the X Elite ZH-WXX from Qualcomm be a RISK 5 architecture chip?
RISC-V got the vocal minority like Linux has, very often for the same reason (they tend towards open/freedom).
Google is working on making Android to work on RISC-V so I would think that would be the earliest for it to be a thing for regular people. Otherwise perhaps small board computers like Raspberry pie will also be more RISC-V but not everyone is interested in such computer (as mentioned in the video, SiFive already got such a board with Linux being an option).
SiFive is expensive AF, I've got VisionFive2 from Ali, which is way cheaper than RPi4.
@@hombrepepega3472 You're right. Though according to benchmarks, the VF2 is more on par with the RPi3 than the RPi4 (which is not a bad thing considering the gap in power consumption between those two Pies).
Phones will be on whatever qualcomm decides they will be. They ensure that with how they license their mobile radio patents. The only one "willing" to pay more to use their own chips is Apple, and they tried to sue qualcomm for their unfair licensing...
@@esoel mainly in North America. There are other mobile network standards in other places in the world. I do also believe Qualcomm is part of the RISC-V foundation.
@@PainterVieraxdude anything less than anpi 4 sucks to use though
Ive got a riscv64 virtual machine running Ubuntu mantic Minotaur on an amd fx9370 host
Pretty much everything works well enough (haven't tested virtio-gpu yet), except pci passthrough from shot to guest... Tho to be fair, it doesn't seem to work with an aarch64 guest either
Does anyone have any info on cross-architecture pci- passthrough?
It would be very slow, because PCIe relies on memory mapped IO. PCIe passthrough works on same-arch guests because it's specifically enabled in hardware virtualization. There may be a way to do this, by putting the CPU emulation inside a hardware virtualization context, but that's not how it's set up right now.
5:05 I wonder if pinecil and ESP32-C3 (along with some other ESP32s) are a good counter arguments
I have a few of ESP32-C3s. SUPER little uControllers. They even come with a 'builtin' JTAG debugger (not just an interface), which works great. I also bought a bunch of WCH CH32V003 and CH32V203 Risc-V based microcontrollers (less capable than the C3, but REALLY cheap! $8 delivered for a CH32V003 dev board, programmer/debugger and 5 extra chips!)
There are a few risc-v sbc/dev boards starting to show up in Asian markets that advertise Linux support. Not sure how standardised they are though, haven't got any of them.
Can confirm, they are not standardized. But then again how standardized are things in the SBC world
I mean, if what you mean by standardized is that you can run unmodified OS images from distros, then yes they are standardized.
Can confirm, have two and they both have support from distro like fedora and Ubuntu.
The big distros are supporting RISC-V very well.
It's a good idea to se something in action...
ExplainingComputers channel done a demo on RISC-V ; episode title :
VisionFive 2: RISC-V Quad Core Low Cost SBC
Google is actively pushing RISC-V enablement in Android and ChromeOS, I've been saying this is the first way it'll be in consumer products as an application processor since 2015. There are standardized platforms now with combinations of extension profiles (e.g. RVA22) and firmware platforms.
It is poised to make it into competitive consumer products as an application processor in a matter of a couple years.
Though the first step was the push of Linux kernel development for R-V. Without that, the AOSP port was unthinkable.
This is 8 years later: Where are the chromebooks or Android devices with Risc-V? I don't know even one. You can't buy these. Risc-V seems to not be doing this great if almost 10 years later, there is still nothing much to show in the normal power (not micro controller space) CPU space.
@@nigratruo Your comment is absurd.
Do you realise Google announced AOSP support on Risc-V only THIS YEAR?
Do you realise that even though Risc-V is almost 10 years old, it takes time to implement from a concept then solidify a standard to finally make end user products?
And it's not like ARM isn't profiting from tech inertia and didn't improve itself during that decade.
@@PainterVierax So.... why does it take so long then and we still don't have anything to show? Why this rabid and obsessed support of Risc-V making it something it is not?
Noticed how fast Apple switched to ARM? THAT is how a successful new system deploys, not 10 years. The fact is that Risc-V is too weak to use for anything else than micro controllers right now and I doubt that this will change, because if something is lagging for 10 years, the reason for it lagging will remain.
@@PainterVierax there is no such thin as tech inertia, you probably don't even work in the tech sector, otherwise you knew this:
If a technology is good enough, it will get quickly adopted. This is always true, because it brings so many benefits. Risc-V is probably not good enough or, let's say: Not substantially better than ARM, so nobody bothers to use it for larger systems than the tiny lower powered micro controllers.
Jim Keller is a journeyman computer architect. Patterson has been slapping the monkey in academia his whole life. Patterson has been working on an ISA for forty freakin' years.
What was the purple case in the ad? It looks cool.
I have a particular use case which I can use one of the more unusual aspect of the available boards for is an Intrusion detection system based on snort. The feature that attracts me is that these boards have two ethernet ports which means I can force all traffic through the board!
As I listen to this 8 months beyond its publishing on YT, Google has announced a smartphone based on a RISC-V design. Less than a year past this interview. Not 10, 15, or 20 years. As the Chinese ramp up with their fabrication capabilities, I would imagine that they will build very sophisticated, high-performance RISC-V implementations that are intended to completely circumnavigate the choke hold that Western firms like Intel have held over the market since its inception in the 1970s.
Yeah, 15-20 years was an absurd estimate, and we all knew that when he said it.
Thank you James.... if u know, u know.
RISC-V is the future, and likely always will be.
Ahem, the Power (formerly PowerPC) architecture is still here 😂
There are few companies designing high performance arm cores. Since there are so few designers, it is warrented to ask if this will change with risc-v?
Anyone can design a RISC-V core without asking for permission, so probably, yes.
Those same designers are also going to be working on RISC-V chips, yeah. Sooner, rather than later, most will be.
Anyone see the booth lady in white dress? ❤
When you look back on your life trying to sort out why you are such a looser. I present to you... your comments like this.
Hard to miss her, as the cameraman kept her in-frame 99% of the time.
@@Look_What_You_Did Don't be jealous.
@@Look_What_You_Did I love when people call other people losers and can't even spell it right.
to be fair, we have proton on linux translating windows games to work on linux/steamdeckOS, so could something happen to desktop computing where we move to RISK and have a translation layer that flawlessly runs windows x86-64 applications? because if that happens we might see a switch away from x86-64 pretty fast.
riscV, riscV, riscV
I think it's way more low level than Wine/Proton so we need emulators to do it (like Rosetta on Mac) but the resource cost is quite high. Hardware translation layers are possible to just do the "micro-ops" interpreter work but Qualcomm got legal issues when trying to implement that.
"Flawlessly", really? I've never seen such a thing in the computing world.
@@rursus8354 Since x86 designs are not pure CISC internally for decades, this can be done flawlessly by using a RiscV core inside a x86 with the micro-ops interpreter being an optional path triggered by a special instruction. Though it requires a Windows flavor for RiscV (a Linux running Wine won't be flawless)
And of course like for ARM, Intel and AMD won't do such a chip as they will still continue to protect their juicy duopoly.
RISC-V emulators are already a thing.
The instructions set is whatever GCC and LLVM output. Chips will need to support that, and compilers are going to support the chips. People will only add new instructions for very specialised systems.
My prediction: expect RISC-V to appear in Google mobile phones soonish.
Qualcomm already has a RISC-V SoC coming, and google put out stuff for readiness for android for RISC-V
Yeah, there will be a bunch of RISC-V phones in the next couple years. I don't think people realize how much enthusiasm for RISC-V there is among all these tech companies.
Why no link to SilverStone?
I am eager to have a Horse Creek development workstation
As someone who owns both a HiFive Unleashed and a HiFive Unmatched, I have a lot less enthusiasm for HiFive Pro / Horse Creek than I did a year ago. Per core I'm expecting it to be maybe 50% faster than the C910 cores in the Lichee Pi 4A, which is great, but 1) it's not going to cost $119 ... probably more like $600 to $1000 like its predecessors, and 2) it's only got 4 cores, same as LPi4A, and no vector ISA. But the killer is the Sophgo SG2042 chip with SIXTY FOUR C910 cores. I've been using a prototype board with it since the end of March, and it rocks. Horse Creek's four cores are probably equal to about six of the cores in the SG2042. Unless you want to do vector stuff, which the SG2042 has (draft 0.7.1, not 1.0, admittedly, but still very useful), and which Horse Creek doesn't have at all.
Flash & RAM is generous on the RISC-V's
when will we have a risc-v 3.x ghz CPU. ?
The Sifive P870 is what you're looking for.
i like Ian. he knows his stuff.
I'm ready for RISC-VI
i think if risc-v was licensed under GPT it would help a lot with the fragmentation issue
There are no fragmentation issues with RISC-V.
The existence of LLVM compiler for RISC-V means hardware designers create unnecessary work for themselves if they modify the instructions in an incompatible way.
Extra instructions are not a problem. They are usually only useful to accelerate a specific operation and those can be hand coded in libraries without affecting the rest of the code
Same as x86 things like avx vs mmx
@@ppokorny99 The only risk is the rise of some proprietary instructions that will be used by some big software editors to reduce the usability of the competition. This already happened with OSes, productivity software and x86.
@@PainterVierax
_"Embrace, extend, and extinguish"_
Yep, we've known that one for years.
My dear is that China's CCP will do something dumb like trying to extend RISC-V to _control_ it.
Try to get the world to use their extensions so that they end up in control of the standard but they accidentally end up extinguishing it.
In my humble, the importance of RISC-V cannot be overstated.
Absolutely agree. And if this quote from an Ars Technica article from January this year is to be believed, we could see RISC-V hit user-facing devices sooner rather than later: "Lars Bergstrom, Android's director of engineering, wants RISC-V to be seen as a "tier-1 platform" in Android, which would put it on par with Arm." That would certainly be an enormous boost for the project.
@@TheDoubleBee
Do you have a link to that article?
@@TheDoubleBee though the real question is how the manufacturers that are already struggling keeping their ARM binary drivers up-to-date will do with another ISA to support?
ARM on Linux (and by extension Android) can be a nightmare for system maintainers and without the myriad of opensource projects many 2+ years old devices are just too unsafe to be decently usable with internet services.
You have an actual humble?
@@esaedvik someone stole her opinion
good report. mb
good observation Gordon "freedom always sounds great until it turns into chaos" very succinct. mb
what will happen to amd x xilinx, it seems no news / leak about it
Well I say there's 2 instructions that should be required no matter what (making them up btw):
TAILOR: give 0 if following a global standard like x86 is, an address to a string of instructions if a custom standard
STANDARD: 0 if custom standard, an address for a string for what global standard following along with the version
As long as those 2 are supplied any installer that supports the relevant standards they indicate can produce the software binary/binaries needed
EASY
Kudos to the cameraman who kept subtly panning to include the booth babe in the white dress in the shot.
That's the content we want.
Betteridge's Law.
RISC-V is to ISAs what Linux is to operating systems: the lack of licensing royalties and restrictions alone adds considerable value. Even if it's not that good *now*, it'll get better soon.
There is no value in Linux. There is only cost to fix all the things that are broken in the standard release. Unless you have hundreds of millions in budget for that Linux is useless.
@@lepidoptera9337 Thanks for telling us you have never personally used Linux but did read a couple articles about it in the 90s.
It's an open source instruction set..not an open source CPU architecture. There's a big difference. There's a big misconception about RISC-Vs "openness"
So what does that mean? the architecture means basically transistors and the arrangement of them to bigger logic groups and registers etc..
is the architecture not open? and how does the instruction set come into play here
The ISA is just the instructions, the code of 1s and 0s that tell a chip to do things. Various companies, academics, and individuals can design any chip they like using the open RISC-V ISA. Those chips and their design can then be subject to whatever intellectual property rights the designer wishes to keep for themselves.
RISC-V chips are not necessarily open source. But for an open source chip to exist, it requires first that the ISA be open. So RISC-V is an important first step towards a truly open hardware ecosystem. Naturally big companies will invest a lot of money into designing RISC-V chips, and will want to keep the rights to those designs, in order to make money either selling chips or selling licenses to those chips. It's also possible for companies to extend the ISA with their own proprietary extensions. You can expect certain software will rely on those extensions, and thus require end users to use certain hardware featuring certain proprietary extensions. But you can also bet that baseline software like the Linux kernel will never require any proprietary extensions, and there will be a lot of effort in the open source community to create other software that doesn't require those extensions. Basically those extensions will be used in things like big data centers running enterprise apps, and various manufacturers may license each other's extensions. Things will standardize around certain extensions for those use cases, you'll see cloud providers offer certain chips for certain workloads just as they offer certain GPUs. But general purpose computing will coalesce around the core ISA, and probably there will be open extensions that are widely adopted, too.
So the exciting thing about RISC-V is not the chips themselves, but the fact that anyone anywhere can make new RISC-V chips. This is a big difference from x86 and ARM.
👍 The doctor's in the house
Doctor of _chemistry_ ..He is not like Lisa Su.who has got an actual doctorate in semiconductors.
@@FindecanorNotGmail It's material science. I can categorically say that the skills I took help me understand a lot more complex intracacies in any IEEE conference. Etching in Si/SiGe layers for GAA-FETs? It's an interesting area of research for the last few years.
RISC-V would be the platform for opensource hardware like networking, display controllers etc....
Ideally, for everything eventually, including workstations, gaming or servers in datacenters.
RISC-V follows the same adoption as Linux, invisible but in plain sight.
Yeah, how many people know or care that their phone is running Linux? How many people know or care that virtually every website or internet-connected app they visit is running on a Linux server?
RISC-V first takes up the microcontroller market. Right now it is already risky not to use either ARM or RISC-V there.
Then there are more chips sold using RISC-V rather than anything else.
Finally standards and community economies of scale start kicking in, Once they come up with a successful kind of Pi alternative in that chip or a dev board.
Then it reings in GPUs and other coprocessors, finally Intel and ARM will start to worry as performant and efficient CPUs start cropping up in RISC-V.
Or at least that is a plausible future.
Intel isn’t worried. They are working with SiFive to make a RISC-V board for developers. AMD will also design RISC-V chips when the market is big enough.
Yes.
SBC as "Small Business Computer", that's a new one. His doctorate is in Chemical Engineering.
Oh, has he got a PhD? Wow. How can you tell?
ha ha ha
I was talking about Bluetooth back in 2001 with manufacturer reps while working at a B2B IT products provider and it really only has grown into it's own in more recent years. Lots of great stuff out there but adoption and maturity of tech is very slow. Look how long we had to accept downgraded and lower resolutions of 1080p before industry would or could move on.
The whole thing about licensing with ARM is a beat up. ARM charges peanuts for it. Consider the scale of ARM usage, and the paltry value of the company. What you are paying for is a patron, someone ensuring that standards are adhered to. RISC-V is at risk of fragmenting, everyone will just implement their own standards. The groups pushing for RISC-V adoption are more likely those that are affected by things like technology export controls.
Those peanuts turn into real money when you're shipping a lot of product. The bigger companies are most affected, and the bigger companies are also the same ones who will have the biggest impact when they inevitably switch.
I believe that RISC-V will become huge! Even surpassing both x86 and ARM within a decade. Because it's open source and with a well-defined simple instruction set set in stone. ARM and x86 are much messier and proprietary.
Agree RISC-V is the future.
Gordon is the only reason PCWorld is still relevant
I dare say if he left pcworld the audience would follow him. Like Chris and Jordon leaving dpreview
@@Autotrope Where did they go?
@@fakecubedPetaPixel
RISC-V is not really that much different from ARM when it first began. It is all about industry adoption, a x8086 works as well as a ARM chip in a PC once it is adapted for the purpose.
So I would disagree with Ian on 10-15 years, if the China choose to adopt it for their market it will happen in 3. It doesn't take very long to extend the instruction set and a little longer to design the circuits to implement them. It is a lot easier now with the established execution sets.
I have worked with at least 6 CPU families, some at the assembler level. You can design a CPU around an old execution set like Sun SPARC or MIPS and phones will still work on it. It is all about the adoption and funding, it is not all that difficult. If Nvidia buys ARM, RISC-V will take off. Same with OS, Harmony OS will take off if the Chinese government funds it and decrees it.
The question is just: What is doing Risc-V better than ARM? Is it more powerful, cheaper and more power saving? The answer is that no, it is not better in any area that matters than ARM, so we just keep using ARM.
@@nigratruo The difference is the politics and the cost. RISC-5 is free from western sanctions and free from royalty mostly. The main attraction is exactly that, it is just as good or at least good enough. For China, the politics and sanction free attribute is everything. For now the Chinese government is essentially staying out of it and allowing market forces to decide
@@nigratruo It's more free, both free as in beer and free as in speech. Once people have figured out how best to use it, it should have no meaningful advantage over ARM, and vice versa. They should be equally as performant and efficient, if given equal amounts of development. But the financial and political incentives all favor RISC-V so it will win. Not immediately, since ARM has a head start on that development. But ARM paved the way with showing companies that adopting a new architecture was possible, and now companies have a lot of experience with switching architectures, and with using RISC architectures in particular (ARM is RISC). ARM went through the door first, took the arrows, and now RISC-V will come in and conquer. And unless something really weird and unexpected happens, there's nothing really to bump RISC-V off its pedestal once it takes it. CISC certainly isn't going to beat it. Another RISC-based architecture won't have any advantage either, unless it's somehow even more open than RISC-V, which isn't really possible. Fully open source chip designs would almost certainly be based on the RISC-V ISA, for the same reason nobody is really making new POSIX-based operating systems that aren't based on the Linux kernel. So while some RISC-V hardware manufacturers might go out of business, they'll only ever really be beaten by other RISC-V hardware manufacturers.
System76 is a surprisingly bad example for a potential user of RISC-V. Everything they do is premium x86. I think he meant companies like PINE64.
well, I think the comparison is good for preinstalled OS vendors. People/companies who seek for light netbooks or mini-stations will get the OS they want and not care about the hardware, just like Android or Apple users.
It's a niche market but like the Chromebook example, targeting people who only cares about UI and few apps is the way to popularize RV chips or Linux to the general public.
@@PainterVierax But System76 hardware is not light at all, and uses the heavyweight (well, I mean relatively) PopOS. They haven't even touched ARM, and would probably be some of the last of these sized vendors to make a low powered RISC machine, ARM or RISC-V or otherwise.
@@FlergerBergitydersh again, I think the comparison is good for preinstalled OS vendors. I agree with what you mean but you are just thinking too deep on it.
Pine64 hasn't the same popularity though. And honestly like almost every devboard/ARM-type manufacturers they are quite lacking on software support, expecting the FOSS community to do the dirty work.
@@PainterVierax queue the posts over the brief and dramatic exchange between pine64 picking favorites with linux distros and developers
@@Lol10000000000000001 I don't understand your sentence. I probably miss some insight.
When I saw the thumbnail I thought that was Ru Paul
A 20 second scan of the comments and it confirms Gordon was spot on!
If they're going to post ignorant and wrong information then what do they expect? For example, there are at least four companies that are currently designing Apple M1-class CPU cores. With actual ex-Apple engineers in many cases. Three of them have formally "announced", which means they are now ready for customers who want to make SoCs to license those cores. The hardware pipeline from that point to having a mass-production chip, and boards with them, in shops, is 4-5 years. NOT 10-15 years. Big difference. And that time period is pretty much entirely predictable. It's the same for e.g. A53 (2012) or A72 (2015) announcement to Pi 3 (2016) and Pi 4 (2019) respectively. Or in the RISC-V world from U74 (October 2018) to VisionFive 2 and Star64 (H1 2023), or C910 (July 2019) to Sipeed LPi4A (quad core TH1520, shipped May 2023) and Milk-V Pioneer (64 core SG2042, shipping in July I think, some people already have prototypes). RISC-V computers comparable to Apple M1 are going to be in the shops at prices the average consumer can afford in 2026-2028. This is just not even in doubt. Sure, the M1 will be a few years behind the leading edge by then, but it will still make a perfectly reasonable PC for most people.
@@BruceHoult Yeah, his timeline estimates were pretty insane, makes him look really out-of-touch. Corrections are warranted. Anyone who's been following ARM and RISC-V all these years knows that RISC-V has been adopted at a much faster rate. It went from a few white papers to enterprise chips in just a few years. There's SBCs all over the place. Adoption on desktop computers may take a long time, but ARM has only barely just gotten there themselves. The first guy through the door takes all the arrows, and ARM has done that now. RISC-V is coming through next and it has so many advantages that ARM never had.
Speaking of bad jokes, I thought of this one the other day: "Elixir, she loves it."
We'll see if RISC is good. Yes, that was a Hackers reference.
Risc-V in the consumer market will be huge, but that's probably 10 years or more into the future.
Sooner.
@@fakecubed Hope so.
Arm is over twenty five years old now. Fifteen years for RISC V to reach Arm in terms of maturity and penetration is a realistic endorsement. If the market gets constrained through monopoly or politics then Risc V might get accelerated development and use.
That's cool, kid, but I was an ARM user a few years after the first silicon came out. RISC-V... not so much. I don't see any of the companies that have successful ARM product lines switch anytime soon.
You seem to be lacking a historical understanding of ARM. It was in use straight out of the gate. It was a matter of there not being a mass market for it be used in. Most technologies were far behind it at the time. We're talking when more than 64k RAM was still a rarity. Within only a couple of years it was already in portable devices and things like set-top boxes.
RISC-V adoption is making ARM look slow. And ARM was fast.
Pc world still exists?
We're still here having fun!
-Adam
@@pcworld I'm glad. I hope y'all are making big profits. *Pats chest*
Not sure about that :P
-Adam
Ian, where is my potato chip?!
WebAssembly makes the standards now. ARM, x86 and RISC-V implement that standard.
The whole RISC-V fandom seems to be based on primarily a misunderstanding of computing efficiency. People really think CISC is complex (I guess because it's in the name) and therefore inefficient, while in reality if you go look at an average pipeline pretty much everything is very similar at similar levels of performance. You just can't go without transaction lookahead or instruction reordering or any other 'big-on-silicon' features if you want to go to modern computing loads. Simplifying processors to the degree that 'people who think RISC is actually a simple instruction set' say only works if you're making a coin cell operated microcontroller. Anything remotely more complex is going to benefit from many of the more complex features of modern processors and the instruction set becomes basically meaningless at that point.
Lower power consumption and fewer gates has already been proven with ARM processors. RISC-V has potential to do even better. X86 instruction decoding is complex.
Intel CPUs are already mostly risc chips with their microcode pretending to be a cisc chip
@@devrim-oguzNo, they are not.
@@NoX-512 RISC-V should be basically the same as ARM in terms of performance and efficiency. They aren't already neck-and-neck simply because ARM is older and chip designers have more experience with it. RISC-V is quickly catching up.
our interest is in potential PERFORMANCE gain of RISC V, given that x86 barely has any development.
BUT, yeah someone needs to shout National Security Threat before they go 64bit.
I am sure someone is already looking to drive that move.
RISC-V performance will be essentially identical to ARM. It just hasn't had as much development time in actual physical chips.
11:33 The US in a nutshell
The llvm compiler has made fragmentation a small issue.
Competition is good.
Can you guys add subtitles please?
it would be enough if they turned on the automatically generated subtitles... (may be any reason for that?)
@@aguijon6 yea, wished we at least had that option
Did you miss the risc-v processors for anroid wearables.
Like Linux odds are without centralized standards that are enforced it's gonna get the "fork it until it dies" treatment. Super fragmented development projects that don't play nice with one another that ends up so convoluted that it's not really suitable for general consumer use.
The only key word missing from this was Celeron 😆
I don't get why there isn't an Apple Silicon competitor for laptops still.
There is and they sell many more laptops than Apple: it is called Intel and AMD.
@@nigratruo what a smart comment
A reduced instruction set, makes computations faster. The "blocks of data" people say this is better. But pre-described blocks do not take into account, that most files are small, and rarely other files are big. Having a data blocks to suit sectors on a hard drive are no longer relevant, because of Solid State Drives that operate at 14,000 Mega Bytes a second. Sector size is no longer relevant.
Chip with ARM architecture is RISC too right ?
Not RISC V but RISC.
What is the issue of having “Domain Specific” functionality in a CPU? Is this not the way Apple has been able to differentiate their product? I thought the whole purpose of embedded is just that, a sole purpose solution. Arm does this already “ again Apple as a reference“ get DSf in their chip even for their PC. Why would you not want to have the best HW if you are building a sole purpose solution, even those that could be running an OS, it happens in Intel, they license their tools and give you different level of IP access to best leverage the CPU… RISC-V enable a baseline from which to build and develop you application closer to the HW, which is really the trend that is happening *look at GPU and MPU solutions.
"will never ... for now"
What’s stopping someone from making a RISC-V core that accepts ARM instructions?
ARM instruction set is copyrighted by ARM. You cannot use ARM instruction sets without ARM license.
It costs money. There is no point to RISC-V except being free of any patents other architectures are burdened with. If you make something that speaks more instruction sets, it is just another (rather bad and primitive) instruction set.
It wouldn't be a RISC-V core then.
2.40 woman in white checks if her breath smells XD
Idk lad, if it delivers the same performance as of the rpi 4 while demanding four times as less energy compared, then I'd call it a win as is.
If it doesn't deliver ten times the performance of RPI, then it's dead on arrival, which it is. The game is now entirely in the GPUs for those applications. You need to pull your mind out of the sewer of 2007.
You're both wrong.
The longest PA I think I've ever heard
As a consumer point of view. I am sick.of ARM, most if not all ARM devices are trash, as in throw away. You cant really run the latest Linux, Android or Windows. You have to be very lucky and have some major community support to get ports of anything let alone unlock the bootloader and get binary drivers working. Most OEM drop support and that makes me weary about SOC on x86. X86 is still the most open hardware from a user, you can run anything. Risc-V would be a nice addition, however. I expect this same companies will continue the ARM/Snapdragon/M* restritive gimpin games, but with arm. Once get good risc routers the gates will open.
ARM has the problem with proprietary driver blobs. Also, I want chips which can be sanded down to look for hardware backdoors.
RISC-V is not free of that either. Nothing in it prevents proprietary drivers or firmware. If you want open hardware insist on only buying hardware with open drivers and firmware.
@@Carewolf An open ISA is the first step to having an open chip.
This will end up like the MSX standard in the '80s. Fragmented and with software that cannot leverage special capabilities of the hardware. Since software was/will be developed for the lower common denominator.