RISC-V Gaining Traction

[erek]

"Handschuh: RISC-V is really not so different here. It can allow you to do any type of implementation for any type of market. That’s because of the possibility to make extensions to the ISA, and so it can adapt to what you need. You can do root-of-trust products on RISC-V and use it at the edge to secure the device, and help you protect the data and do secure inferencing and things like that. I would actually argue it gives you a bit more latitude to operate in new areas where new requirements are showing up and coming coming around because you can innovate around it very easily.

Talukdar: And it’s not just China and AI. In India, there is an indigenous processor development project going on at the Indian Institute of Technology. There has been a bid drive in India to do these kinds of things indigenously, but it hasn’t really gone on at scale yet. RISC-V opens up loads of opportunities for development that’s happening today in India. It would not be a surprise to see India coming out with a brand-new, from-scratch SoC based on RISC-V.

Part 2 coming soon …."

https://semiengineering.com/risc-v-gaining-traction/

 

[erek]

 

[Zarathustra[H]]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalty-free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

 

[Lakados]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalties free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

I agree with this to a point, it depends on the nature of the open source license.

 

[Factum]

Headline: "Gaining traction"..
Article: Gives NO solid numbers.

ZZZzzzzZZZZzzzZZZ....

 

[sunruh]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalty-free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

how do you get competition if everyone has access to the same design?
without roshambo (ford/chevy/dodge) you dont get any.
if someone makes a design change then everyone gets it. there is NO competition at all.

 

[Dan_D]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalty-free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

Except, this isn't really how things have worked in this industry. The primary driver of most innovation has been proprietary technologies and licensed technologies.

 

[Zarathustra[H]]

how do you get competition if everyone has access to the same design?
without roshambo (ford/chevy/dodge) you dont get any.
if someone makes a design change then everyone gets it. there is NO competition at all.

They would compete on architecture. The "how you get there". Not on the instruction set.

A car analogy here would be, you compete on how the engine is designed, how the suspension is tuned, what features are included etc. without having to pay a license for it to be able to run on roads.

 

[Zarathustra[H]]

Except, this isn't really how things have worked in this industry. The primary driver of most innovation has been proprietary technologies and licensed technologies.

Not always. And where this has been the case it has done more harm than good. See Intel suing everyone and anyone who tried to compete with them on x86.

Now that ARM holdings is reportedly up for sale there is a very real risk the same wi happen with ARM, with a buyer coming in like Oracle did when they bought Sun, and undoing all the good they have done, setting industry back significantly.

The only reason the PC platform has survived the test of time where so many others have failed is the relative openness and compatibility allowing many players to come in and compete for the best solution.

Every time an Intel (x86) or an Nvidia (G-Sync, etc.) come in and try to make things more proprietary, lock customers in and lock others out, it is an existential threat to everything that makes the platform good.

One of the times of greatest advancement in PC computing came in the late 90's and early 2000's when Intel and AMD were competing with eachother and there were 5 different GPU makers competing on the marketplace.

In the doldrums that followed we see how uninspiring things get when one company has hegemony and tries to maintain that position by locking others out. It is a disaster.

 

[[Spectre]]

x86 is dead again? Well crap. Stupid zombie architecture.

 

[pendragon1]

 

[whateverer]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalty-free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

But an architecture that doesn't specify Vectors in it's default configuration sounds like you're going to needs layers of cruft just to make a serviceable desktop processor

https://sourceware.org/pipermail/binutils/2020-July/112570.html

It will take design wins in embedded, but once you step-up to desktop-level core complexity, the "free" advantage will be lost over already established ARM...not to mention the standardized Scalable Vector Extensions they defined.

 

[Zarathustra[H]]

But an architecture that doesn't specify FP in it's default configuration sounds like you're going to needs layers of cruft just to make a serviceable desktop processor
It will take design wins in embedded, but once you step-up to desktop-level core complexity, the "free" advantage will be lost over already established ARM.

The thing as I see it is that ARM although requiring a license is a fairly easy license which is one of the reasons it has flourished over the last couple of decades.

With ARM Holdings potentially up for sale and Nvidia reportedly interested and in advanced talks, that could change for the much worse in a hurry

ARM Holdings right now is just interested in driving revenues from designs and licensed. A bigger player with an agenda, and a history of trying to leverage questionable lock-ins, lock-outs and proprietary designs to maximize their advantage could seriously change that.

It has happened before. Just look at what happened when evil Oracle bought nice guy Sun Microsystems...

This is why a open source royalty/license free instruction set is so important. Not because the ARM license costs too much today, but because of the real risks of control in the future.

We don't need another Intel suing everyone and everything out of existence.

 

[Zarathustra[H]]

This article in Forbes takes out time to explain the basics to non-tech readers, but otherwise does a good job of illustrating the problem:

If Nvidia Buys Chipmaker Arm Holdings For $32 Billion From SoftBank, What Will Happen To Its Model Of Neutrality?

 

[serpretetsky]

View attachment 266082

I always think this quote is a little funny because people often quote it to demonstrate another rediculous things said in Hackers. Pentium is CISC not RISC! And today we are running in an X86 CISC world!

But correct me I'm wrong, the P6 was the pentium pro. And as far as i know, this was when intel decoupled the actual instruction set (CISC) from the micro ops that the core actually runs (RISC). Every intel cpu after that followed in the same footsteps (well.. i'm not sure about netburst... but let's forget that ever existed )

Furthermore, these days ARM is everywhere. And since ARM originated as purely a RISC design, it's fair to say that RISC did change everything, and that this quote from Hackers is 100% accurate.....


.... ofcourse it's still really cheesy

 

[DanNeely]

I always think this quote is a little funny because people often quote it to demonstrate another rediculous things said in Hackers. Pentium is CISC not RISC! And today we are running in an X86 CISC world!

But correct me I'm wrong, the P6 was the pentium pro. And as far as i know, this was when intel decoupled the actual instruction set (CISC) from the micro ops that the core actually runs (RISC). Every intel cpu after that followed in the same footsteps (well.. i'm not sure about netburst... but let's forget that ever existed )

Furthermore, these days ARM is everywhere. And since ARM originated as purely a RISC design, it's fair to say that RISC did change everything, and that this quote from Hackers is 100% accurate.....


.... ofcourse it's still really cheesy

More or less. x86 does have one inescapable bit of CISC baggage. Its external instruction set is a complicated mess, and decoding it into uOps is much harder than arm/riscv/etc. For larger/higher powered chips this hasn't mattered for close to 20 years. But it's a real drag for small/low power designs; having hurt early atom on the laptop and being an important factor in Intel's mobile aspirations failing a few years back. If the future of servers becomes chips hundreds of sub-watt cores; it may again become a problem for both Intel and AMD.

 

[whateverer]

The thing as I see it is that ARM although requiring a license is a fairly easy license which is one of the reasons it has flourished over the last couple of decades.

With ARM Holdings potentially up for sale and Nvidia reportedly interested and in advanced talks, that could change for the much worse in a hurry

ARM Holdings right now is just interested in driving revenues from designs and licensed. A bigger player with an agenda, and a history of trying to leverage questionable lock-ins, lock-outs and proprietary designs to maximize their advantage could seriously change that.

It has happened before. Just look at what happened when evil Oracle bought nice guy Sun Microsystems...

This is why a open source royalty/license free instruction set is so important. Not because the ARM license costs too much today, but because of the real risks of control in the future.

We don't need another Intel suing everyone and everything out of existence.

If you think that making something open-source is somehow gojng to make it free to produce , you've got another thing coming. Competitive chip designs have become so complicated ARM Corporation has consolidated all chip design outside of Apple, and even with the added income of semi-custom cores, they're still not making enough money off licenses to pay for new chip designs.

Oracle bought Sparc when the platform was already dead., so not really a good comparison. Since Apple has an eternal ARM license, you can bet ARM Corporation is going to be kept honest just trying to keep-up with Apple's massive chip design advantage. But at the same time, they will have ti raise prices to keep up (just like Apple had almost doubled the cost of new phones over the last few years). It's not even close to when Oracle bought Sun; unlike Sparc, there is actualy a thriiving competitive ARM engineering talent base (with no signs of stopping.) As long as there is demand for more complex cores, the engineering competition will continue.

Sun only created the Sparc because the 68000 was shit, but they stopped being a driving factor in the future of that architecture by the late 1990s (essentiall handed it off to Fujitsu), so of course Oracle hasd no choice but to kill it.

When you don't have the luxury of hiding increased design costs in hundreds of millions of $1000 handsets, then you have to increase the licensing costs. Switching to RISC-V is not going to magically get you around this. If ARM Corp suddenly imploded, then you can just go back to "build your own core," and slow down chip development for everyone outside Apple. But ARM architecture is not going to get replaced by RISCV on the high-end market.

 

[Zarathustra[H]]

If you think that making something open-source is somehow gojng to make it free to produce , you've got another thing coming. Competitive chip designs have become so complicated ARM Corporation has consolidated all chip design outside of Apple, and even with the added income of semi-custom cores, they're still not making enough money off licenses to pay for new chip designs.

Oracle bought Sparc when the platform was already dead., so not really a good comparison. Since Apple has an eternal ARM license, you can bet ARM Corporation is going to be kept honest just trying to keep-up with Apple's massive chip design advantage. But at the same time, they will have ti raise prices to keep up (just like Apple had almost doubled the cost of new phones over the last few years). It's not even close to when Oracle bought Sun; unlike Sparc, there is actualy a thriiving competitive ARM engineering talent base (with no signs of stopping.) As long as there is demand for more complex cores, the engineering competition will continue.

When you don't have the luxury of hiding increased design costs in hundreds of millions of $1000 handsets, then you have to increase the licensing costs. Switching to RISC-V is not going to magically get you around this.,

I'm not disagreeing with you.

It is expensive to design CPU architectures.

If the instruction set is open source - however - it removes the risk of platform lockouts because whoever developed or owns the instruction set doesn't want to have to compete with you.

I could easily see an Nvidia owned ARM Holdings dangle a refusal to renew their license or something like that, as a negotiation tool to gain more out of the likes of Apple or Qualcomm.

So, as I stated before, it is not about the relatively minor licensing fees. It's about the control over the instruction set, and what a less scrupulous company like Nvidia does with it when push comes to shove and they have money to be gained by playing hardball.

 

[Zarathustra[H]]

But an architecture that doesn't specify Vectors in it's default configuration sounds like you're going to needs layers of cruft just to make a serviceable desktop processor

https://sourceware.org/pipermail/binutils/2020-July/112570.html

It will take design wins in embedded, but once you step-up to desktop-level core complexity, the "free" advantage will be lost over already established ARM...not to mention the standardized Scalable Vector Extensions they defined.

That work is already underway.

It is purposely a modular ISA. Enough parts to make it usable have been frozen, but there are many more being worked on that will come online. An extension for vector operations is among them:

edit: Whoops, that image was MUCH larger than I expected...

 

[Revenant_Knight]

First, and most importantly, TURN DOWN THE BASS FILTER. Good lord...

Back to topic. I've always hoped we could see more competition on the CPU market, and I'm all for an open standard that any company could design off of. It's a bit ridiculous that we have only 2 companies left designing X86 (China excluded).

 

[sleepeeg3]

The thing as I see it is that ARM although requiring a license is a fairly easy license which is one of the reasons it has flourished over the last couple of decades.

With ARM Holdings potentially up for sale and Nvidia reportedly interested and in advanced talks, that could change for the much worse in a hurry

ARM Holdings right now is just interested in driving revenues from designs and licensed. A bigger player with an agenda, and a history of trying to leverage questionable lock-ins, lock-outs and proprietary designs to maximize their advantage could seriously change that.

It has happened before. Just look at what happened when evil Oracle bought nice guy Sun Microsystems...

This is why a open source royalty/license free instruction set is so important. Not because the ARM license costs too much today, but because of the real risks of control in the future.

We don't need another Intel suing everyone and everything out of existence.

Android was open-source... and now it is a mix - at best.

If I understand this correctly, what's to stop the big players from developing proprietary ISA's, just like Google does with Google-based Android apps?

 

[1337Goat]

Android was open-source... and now it is a mix - at best.

If I understand this correctly, what's to stop the big players from developing proprietary ISA's, just like Google does with Google-based Android apps?

It's up to we-the-consumers to never buy or support them if they use an unethical license.
Don't buy from them. Ethics > Performance.

 

[ChadD]

I support RISC-V becoming the next desktop architecture rather than ARM.

Royalty-free open source all the way. It's the only way we can get a truly competitive field without various players being excluded on the basis of licenses.

RISC-V becoming the desktop ISA to replace x86 would actually be terrible.

Here is why.

ARM is open to licence to anyone. They also have 2 different types of licences. You can be a Apple, Samsung or huawei and licence the ISA and design your own core. Or more importantly you can alternatively licence a fully designed and proven core compatible with the fab process of the day.

ARM has done a good job so far of making sure they have stock arm core designs that re competitive. Yes the Apple core may be faster... the qualcom core may be more power efficient, the huawei one may be Chinese made which makes their owners happy. However you can also have a mediatek who just takes the stock cores and builds good quality low cost parts cause they haven't had to spend billions on design teams and extensive fabrication tests.

RISC-V would end up excluding all but the largest tech companies from creating product. There is no RISC-V design team consisting of a couple hundred high level chip designers sitting around all day building RISC-V cores. Yes Apple AMD Samsung a bunch of companies COULD decide to all spontaneously spend the 1-5 billion min it would cost to design a high performance RISC-V core. But none of them would play sharesies. Its also a 100% open source ISA... so if that came to pass. AMD would include its own custom instructions, Samsung would Apple would. Everyone would and guess what no ones RISC-V would be interchangeable. It would be a nightmare of multiple chips that where ALMOST compataibile. That would destroy one benchmark on software optimized for their added extensions... but run like a 1990 IBM on any other RISC-V.

Like it or not the chip business has always needed a steward. Someone to control the ISA... let everyone play. Give no real advantage. And most importantly as I say let everyone that has a desire to play into the game. I want smaller 100 employee start ups to be able to licence a working our of order A75 core and build a desktop chip. I want the Teslas and Fords of the world to be able to hire a 50 person team licence a ARM designed core and build a mil spec indestructible in house auto chip. By leveraging all of that... we get the bets Desktop chips... cause its in ARMs interest to continue designing better and better cores... and cores for all fabrication processes.

RISC-V will never have a multi billion dollar server or desktop class design for those reasons. For say AMD to build a RISC-V core anything special they do will be unique to them... they will not have to share extensions ect and why would they. MMX and SSE and all those things got shared due to pre existing patent deals. If x86 had been a clean ISA... and no one owned it requiring no one to cross licence. Intel would have added MMX and SSE and patented them and that would have been the end. Sure the runner up would come up their 3D now type implementation but it wouldn't be interchangeable software wise, and would be a legal night mare for years.

I am a pretty strong neck beard... but I can see the practical issues with a open source ISA. It has its place for things like internal controllers... which Nvidia and WD are using RISCV for. But as a full on general purpose compute ISA. Na it won't ever work. ARM is the closest we will get.... unless and its a big unless. The people behind Risc-V where to start a foundation as Linux has... and convince the entire industry to fund the creation of ready made cores, and have everyone agree to give them complete control over extension additions. This sort of works for Linux... but for hardware. I don't know the scale in terms of required $ would be much larger, and good luck convincing companies like Apple that donating a couple hundred million a year to do their part to keep the RISC-V foundation designing cores for their low ball competition and the likes of Tesla ect ect seems really implausible.

RISV-V good on paper... a controller chip / iot player at most.

 

[1_rick]

ut none of them would play sharesies. Its also a 100% open source ISA... so if that came to pass. AMD would include its own custom instructions, Samsung would Apple would. Everyone would and guess what no ones RISC-V would be interchangeable.

A hypothetical Windows RISC-V would probably use the lowest common denominator of functionality so there wouldn't have to be 5 RISC-V Windows flavors. Something like that would actually tend to decrease the likelihood of the chip makers rolling their own incompatible extensions, because they would likely see relatively limited use.

Edit: beyond that I'm not disagreeing with your larger point. It would be possible to imagine, say, Samsung having their own customized Linux kernel for Android or Tizen, for example.

 

[Nobu]

RISC-V becoming the desktop ISA to replace x86 would actually be terrible.

Here is why.

ARM is open to licence to anyone. They also have 2 different types of licences. You can be a Apple, Samsung or huawei and licence the ISA and design your own core. Or more importantly you can alternatively licence a fully designed and proven core compatible with the fab process of the day.

ARM has done a good job so far of making sure they have stock arm core designs that re competitive. Yes the Apple core may be faster... the qualcom core may be more power efficient, the huawei one may be Chinese made which makes their owners happy. However you can also have a mediatek who just takes the stock cores and builds good quality low cost parts cause they haven't had to spend billions on design teams and extensive fabrication tests.

RISC-V would end up excluding all but the largest tech companies from creating product. There is no RISC-V design team consisting of a couple hundred high level chip designers sitting around all day building RISC-V cores. Yes Apple AMD Samsung a bunch of companies COULD decide to all spontaneously spend the 1-5 billion min it would cost to design a high performance RISC-V core. But none of them would play sharesies. Its also a 100% open source ISA... so if that came to pass. AMD would include its own custom instructions, Samsung would Apple would. Everyone would and guess what no ones RISC-V would be interchangeable. It would be a nightmare of multiple chips that where ALMOST compataibile. That would destroy one benchmark on software optimized for their added extensions... but run like a 1990 IBM on any other RISC-V.

Like it or not the chip business has always needed a steward. Someone to control the ISA... let everyone play. Give no real advantage. And most importantly as I say let everyone that has a desire to play into the game. I want smaller 100 employee start ups to be able to licence a working our of order A75 core and build a desktop chip. I want the Teslas and Fords of the world to be able to hire a 50 person team licence a ARM designed core and build a mil spec indestructible in house auto chip. By leveraging all of that... we get the bets Desktop chips... cause its in ARMs interest to continue designing better and better cores... and cores for all fabrication processes.

RISC-V will never have a multi billion dollar server or desktop class design for those reasons. For say AMD to build a RISC-V core anything special they do will be unique to them... they will not have to share extensions ect and why would they. MMX and SSE and all those things got shared due to pre existing patent deals. If x86 had been a clean ISA... and no one owned it requiring no one to cross licence. Intel would have added MMX and SSE and patented them and that would have been the end. Sure the runner up would come up their 3D now type implementation but it wouldn't be interchangeable software wise, and would be a legal night mare for years.

I am a pretty strong neck beard... but I can see the practical issues with a open source ISA. It has its place for things like internal controllers... which Nvidia and WD are using RISCV for. But as a full on general purpose compute ISA. Na it won't ever work. ARM is the closest we will get.... unless and its a big unless. The people behind Risc-V where to start a foundation as Linux has... and convince the entire industry to fund the creation of ready made cores, and have everyone agree to give them complete control over extension additions. This sort of works for Linux... but for hardware. I don't know the scale in terms of required $ would be much larger, and good luck convincing companies like Apple that donating a couple hundred million a year to do their part to keep the RISC-V foundation designing cores for their low ball competition and the likes of Tesla ect ect seems really implausible.

RISV-V good on paper... a controller chip / iot player at most.

Eeh, no reason someone like ARM Holdings (or ARM Holdings themselves) couldn't do something like ARM Holdings is doing, but with RISC-V. They might have to make a spin-off to make the investors happy, but it could be done.

 

[ChadD]

A hypothetical Windows RISC-V would probably use the lowest common denominator of functionality so there wouldn't have to be 5 RISC-V Windows flavors. Something like that would actually tend to decrease the likelihood of the chip makers rolling their own incompatible extensions, because they would likely see relatively limited use.

Well RISC-V is very risc... I would suggest in order to provide any real performance for things like Photoshop ect. You would either bits tacked onto a SOC (as Apple is about to do with ARM), or you would need to add some form of MMX/SSE type extensions which no one is going to share with their competition willingly.

Microsoft isn't going to step up and design a RISC-V core for everyone to use. And the big players like AMD aren't going to spend that money then say here Mediatek you wanna compete with us ?

RISC-V being viable for higher power devices would have to have a setup like Linux... where a foundation that takes money from everyone in the industry to fund itself. Controlled the ISA and ALL extensions. But of course being open source people could still just do there own thing at any time. So ya 10 years down the road in a scenario where eveyrone goes RISC-V. Any company that gets enough market share could decide to fork it and say see ya suckers.

As much as it sucks. Someone need to own the ISA... and licence it out. Which is exactly what ARM has been doing. I would love to see ARM on desktops for that reason. Software will be cross compatible.. and as long as ARM stays strong and continues to never be more then one cycle behind performance wise with their stock cores. Things should remain pretty open. ARM fosters a lot of competition... cause even if your Apple or Samsung and you throw billions at your own core design. In a year mediatek or some other small player can licence the latest ARM designed core and build a competitive chip. RISC-V would stifle competition as odd as that sounds.

 

[ChadD]

Eeh, no reason someone like ARM Holdings (or ARM Holdings themselves) couldn't do something like ARM Holdings is doing, but with RISC-V. They might have to make a spin-off to make the investors happy, but it could be done.

They could but would they bring any real advantage over ARM ? They would also always be in a position where some other player could come along a few years down the road take all their work fork it and steal all their customers. Yes shareholders would love the idea of having zero protection of their investment.

If say Nvidia decided they where going to take RISC-V and build their own open cores. If they managed to make them worth while to licence (which means they would have to be very custom... its a OPEN source ISA they can't copy write the ISA just potential extensions). Once the software ecosystem was in place. Nothing stops say Google from saying... ok good work NV. Here is our version... who wants it ?

The problem is the ISA being open source.

 

[Nobu]

They could but would they bring any real advantage over ARM ? They would also always be in a position where some other player could come along a few years down the road take all their work fork it and steal all their customers. Yes shareholders would love the idea of having zero protection of their investment.

If say Nvidia decided they where going to take RISC-V and build their own open cores. If they managed to make them worth while to licence (which means they would have to be very custom... its a OPEN source ISA they can't copy write the ISA just potential extensions). Once the software ecosystem was in place. Nothing stops say Google from saying... ok good work NV. Here is our version... who wants it ?

The problem is the ISA being open source.

The code is 100% copyright protected. Open source does not remove copyright protections...

 

[1_rick]

Microsoft isn't going to step up and design a RISC-V core for everyone to use.

No, of course not. What I meant is that if (say) 3 different companies produced RISC-V chips with differing extensions, and MS ported Windows to RISC-V, they would likely only use the common instruction set. They already historically have done something like that with the various other CPU archs they've supported, even if it was just to the extent of not using all the addressing modes a CPU might support.

 

[Nobu]

They could but would they bring any real advantage over ARM ? They would also always be in a position where some other player could come along a few years down the road take all their work fork it and steal all their customers. Yes shareholders would love the idea of having zero protection of their investment.

If say Nvidia decided they where going to take RISC-V and build their own open cores. If they managed to make them worth while to licence (which means they would have to be very custom... its a OPEN source ISA they can't copy write the ISA just potential extensions). Once the software ecosystem was in place. Nothing stops say Google from saying... ok good work NV. Here is our version... who wants it ?

The problem is the ISA being open source.

And to answer your hypothetical, no they can't. The ISA is open source, but users of the ISA aren't required to release their modifications if they don't want to.

Even if they did, someone who wanted to "steal" this work would still have to bring up their own manufacturing or contract with someone who can, and be competent in the venture. If they have buggy hardware or drivers, I doubt there would be many who would buy their copycat product. The one with the better implementation and support would win.

 

[ChadD]

The code is 100% copyright protected. Open source does not remove copyright protections...

Not it just allows for forking. You can always fork it do your own thing and change the licence of your design. That is the problem. If anyone company was to become THE RV desktop chip supplier cause their product was superior marketing or what have you... what incentive do they have to stay loyal to a RISC-V foundation... and continue paying dues and playing nice helping everyone else catch up ? If Intel or AMD was to switch to RISC-V and the market took it up and in 10 years any one company had say 70% of the market... why would they play nice to help the other guys catch them. Zero incentive to stay and play along.

We are not talking software like a Linux kernel... the RISC-V foundation guys are trying to make it a Linux kernel situation where everyone contributes and helps develop the RISC-V in the same way. Problem is Linux is JUST a kernel... its not a OS, or any sort of working system. If Google had to share major parts of android they would just switch to their own kernel. So all the RV foundation can do is try and get everyone to agree on ISA changes and perhaps keeping extensions open. (but nothing requires that).

In order for them to do like ARM and actually design a stock core... they would have to get all the major players to agree to bank roll their competition directly. That will never ever ever happen. In ARMS case they are for profit they design a core and licence it to the smaller fish so they can yes compete with the big boys. Those licences cost more... and that + the licence fees from the big boys pays for all that. I would rather have Apple paying a small licence fee to ARM so that arm can design Cortex cores for the Mediateks of the world to buy. Without that licencing money the big chip makers just get bigger and the little guys never happen. No small upstart with a couple million in investment is ever going to make something to compete with a Samsung, Apple, AMD or Intel RISC-V design if they go that way. Who would have the capital to take the RISC-V white papers and design their own cutting edge chip. No one RISC-V would not foster competition, it would stifle it.

On the other hand ARM creates competition by having a dual licencing format... and the money those fees brings in to design cores. Would a smaller player like mediatek really have the money to design a new core for every single fab iteration. Of course Apple and Samsung can afford that... but the smaller players can not. They don't have to... ARM builds core designs for every major fab process. Even the Qcoms of the world only really tweak those stock ARM cores most cycles. If the world went RISC-V only the largest would be able to continue building out high quality cores every fab cycle. We would trade Intel and AMD... for Apple and Samsung... and perhaps Intel and AMD if they played along. No one else would be able to compete without type of design assistance they currently buy from ARM.

 

[KarsusTG]

The thing as I see it is that ARM although requiring a license is a fairly easy license which is one of the reasons it has flourished over the last couple of decades.

With ARM Holdings potentially up for sale and Nvidia reportedly interested and in advanced talks, that could change for the much worse in a hurry

ARM Holdings right now is just interested in driving revenues from designs and licensed. A bigger player with an agenda, and a history of trying to leverage questionable lock-ins, lock-outs and proprietary designs to maximize their advantage could seriously change that.

It has happened before. Just look at what happened when evil Oracle bought nice guy Sun Microsystems...

This is why a open source royalty/license free instruction set is so important. Not because the ARM license costs too much today, but because of the real risks of control in the future.

We don't need another Intel suing everyone and everything out of existence.

I was going to post this exact thing. I really like ARM, but if this happens it could be a huge blow to the community.

 

[ChadD]

No, of course not. What I meant is that if (say) 3 different companies produced RISC-V chips with differing extensions, and MS ported Windows to RISC-V, they would likely only use the common instruction set. They already historically have done something like that with the various other CPU archs they've supported, even if it was just to the extent of not using all the addressing modes a CPU might support.

Exactly MS would support the official ISA... and official RISC-V foundation extensions. But if the companies start adding their own bits and extensions there is nothing stopping that.

Would AMD be around if they didn't have MMX ... SSE and AVX. I wouldn't think so. They only have access to those things cause they agreed years back with Intel to cross licence each others stuff forever as a Legal compromise. How could that work for RISC-V where no one has any legal toes to step on... there are no cross licence deals, and no one would sign one.

 

[Zarathustra[H]]

RISC-V becoming the desktop ISA to replace x86 would actually be terrible.

Here is why.

ARM is open to licence to anyone. They also have 2 different types of licences. You can be a Apple, Samsung or huawei and licence the ISA and design your own core. Or more importantly you can alternatively licence a fully designed and proven core compatible with the fab process of the day.

ARM has done a good job so far of making sure they have stock arm core designs that re competitive. Yes the Apple core may be faster... the qualcom core may be more power efficient, the huawei one may be Chinese made which makes their owners happy. However you can also have a mediatek who just takes the stock cores and builds good quality low cost parts cause they haven't had to spend billions on design teams and extensive fabrication tests.

RISC-V would end up excluding all but the largest tech companies from creating product. There is no RISC-V design team consisting of a couple hundred high level chip designers sitting around all day building RISC-V cores. Yes Apple AMD Samsung a bunch of companies COULD decide to all spontaneously spend the 1-5 billion min it would cost to design a high performance RISC-V core. But none of them would play sharesies. Its also a 100% open source ISA... so if that came to pass. AMD would include its own custom instructions, Samsung would Apple would. Everyone would and guess what no ones RISC-V would be interchangeable. It would be a nightmare of multiple chips that where ALMOST compataibile. That would destroy one benchmark on software optimized for their added extensions... but run like a 1990 IBM on any other RISC-V.

Like it or not the chip business has always needed a steward. Someone to control the ISA... let everyone play. Give no real advantage. And most importantly as I say let everyone that has a desire to play into the game. I want smaller 100 employee start ups to be able to licence a working our of order A75 core and build a desktop chip. I want the Teslas and Fords of the world to be able to hire a 50 person team licence a ARM designed core and build a mil spec indestructible in house auto chip. By leveraging all of that... we get the bets Desktop chips... cause its in ARMs interest to continue designing better and better cores... and cores for all fabrication processes.

RISC-V will never have a multi billion dollar server or desktop class design for those reasons. For say AMD to build a RISC-V core anything special they do will be unique to them... they will not have to share extensions ect and why would they. MMX and SSE and all those things got shared due to pre existing patent deals. If x86 had been a clean ISA... and no one owned it requiring no one to cross licence. Intel would have added MMX and SSE and patented them and that would have been the end. Sure the runner up would come up their 3D now type implementation but it wouldn't be interchangeable software wise, and would be a legal night mare for years.

I am a pretty strong neck beard... but I can see the practical issues with a open source ISA. It has its place for things like internal controllers... which Nvidia and WD are using RISCV for. But as a full on general purpose compute ISA. Na it won't ever work. ARM is the closest we will get.... unless and its a big unless. The people behind Risc-V where to start a foundation as Linux has... and convince the entire industry to fund the creation of ready made cores, and have everyone agree to give them complete control over extension additions. This sort of works for Linux... but for hardware. I don't know the scale in terms of required $ would be much larger, and good luck convincing companies like Apple that donating a couple hundred million a year to do their part to keep the RISC-V foundation designing cores for their low ball competition and the likes of Tesla ect ect seems really implausible.

RISV-V good on paper... a controller chip / iot player at most.

Assuming an ARM Holdings that continues to be independent of other chipmakers and continues to be neutral and sell licenses to anyone who want them without giving anyone an unfair advantage, you are right.

The problem is that as long as there is a proprietary ISA there is always a risk that someone takes control of it and uses unfair business practices to restrict the industry and drive it to their own advantage.

That is a very real risk if - as we are reading - Nvidia buys ARM Holdings from SoftBank.

Heck, I would be fine with it if they open source the ISA itself, but continue to license and sell their architecture to others. The risk of an Nvidia, or an Intel, or an Apple taking control of the ISA and using that to suppress competition is absolutely huge as long as he ISA remains proprietary.

 

[ChadD]

Assuming an ARM Holdings that continues to be independent of other chipmakers and continues to be neutral and sell licenses to anyone who want them without giving anyone an unfair advantage, you are right.

The problem is that as long as there is a proprietary ISA there is always a risk that someone takes control of it and uses unfair business practices to restrict the industry and drive it to their own advantage.

That is a very real risk if - as we are reading - Nvidia buys ARM Holdings from SoftBank.

Heck, I would be fine with it if they open source the ISA itself, but continue to license and sell their architecture to others. The risk of an Nvidia, or an Intel, or an Apple taking control of the ISA and using that to suppress competition is absolutely huge as long as he ISA remains proprietary.

Agreed... I doubt there is much to worry about though. Frankly Nvidia can't afford ARM. lol I don't believe any of the recent rumors that NV is in serious late stage talks to buy ARM for 32 Billion. Softbank paid 31 billion 4 years ago... that isn't a nice enough profit even if Nvidia could raise that type of cash. ARM has grown in value the last 4 years by a lot more then that. NV does have a market valuation of 230 billion or so. But I really don't see them getting ARM for less then mid 40s at best. I also doubt they would pass regulatory approval.

I suspect softbank will be holding on to ARM... and as ARMs ceo has talked about in the next year or two they will be offering a ARM IPO.

Anyway I would put more stock in the rumors if they where even well thought out fabrications. Softbank may be looking for cash... but selling off one of their jewels for a pittance seems like obvious BS. I wouldn't doubt NV looked into it... but I doubt they got a reasonable 32 billion dollar quote. Unless the deal was more like... 32 billion in cash and another 10-20 billion in NV stock to softbank or some such thing.

 

[whateverer]

Exactly MS would support the official ISA... and official RISC-V foundation extensions. But if the companies start adding their own bits and extensions there is nothing stopping that.

Would AMD be around if they didn't have MMX ... SSE and AVX. I wouldn't think so. They only have access to those things cause they agreed years back with Intel to cross licence each others stuff forever as a Legal compromise. How could that work for RISC-V where no one has any legal toes to step on... there are no cross licence deals, and no one would sign one.

Right, when you have a free-for-all, you end up with wasted effort. Just look at 3D-NOW, a long-dead vector extension.

If you build a processor without any REQUIRED rules, you encourage TONs of wasted attempts to rebuild a the same portion...like say, a vector unit.. Just HOW MANY DIFFERENT INCOMPATIBLE PACKAGE MANAGEMENT SYSTEMS does Linux have, for one example?

Offerings with a much more complete open core and free-to-use architecture already exist:

https://www.mips.com/mipsopen/

Really, RiiscV was just an attempt to make a replacement for MIPS, so they could sell you new Computer Architecture books. The incredible simplicity of the base architecture is intended for a student to build this thing on an FPGA within a semester, not power advanced architectures competitive with ARM.

You think I'm kidding? here's their latest revision:

https://www.amazon.com/Computer-Organization-Design-RISC-V-Architecture/dp/0128122757

 

[Lakados]

Softbank is hurting, yet another loss write off this one like 20 Billion from their tech holdings. SoftBank is looking to offload things to make money back and they are seeing who’s biting. NVidia May be the poster boy on this but I would bet Huawei would be interested, they can’t get any more Arm chips and they have started working on their own RISC-V stuff, they are touting it as the best thing since the communist party invented sliced bread. I’m less worried about NVidia buying it, they would have so many regulators up their behind that it would force them honest out of the vast teams of lawyers that would be watching their every bathroom break. A China owned one would just be bad news.

 

[1337Goat]

If Nvidia bought ARM, the real result would be both boring datacenter stuff, but also: ~~~New Console Producer~~~!

Xbox vs PlayStation vs Nintendo vs Nvidia.

Would PC still be the top cat?

 

[juanrga]

how do you get competition if everyone has access to the same design?
without roshambo (ford/chevy/dodge) you dont get any.
if someone makes a design change then everyone gets it. there is NO competition at all.

In the same way you have actually competition between AMD and Intel, with both accessing to the same x86 ISA or the same way you have competition between Apple, Qualcomm, Samsung, or between Ampere, Amazon, Marwell, with all them accessing to the same ARM ISA.

The difference is that third companies cannot acquire a x86 license, whereas anyone can pay to get a ARM license, whereas RISC-V licenses are free, so you can spend the money in the design.

 

[juanrga]

View attachment 266082

Well, there is no longer pure CISC chips in the x86 world. Any modern x86 core is a hybrid with a CISC front-end and a RISC engine. x86 instructions are translated into uops, which are stored (uop cache), renamed, reordered, issued, executed, and retired.

Also x86 is in clear decline



RISC finally won!

More or less. x86 does have one inescapable bit of CISC baggage. Its external instruction set is a complicated mess, and decoding it into uOps is much harder than arm/riscv/etc. For larger/higher powered chips this hasn't mattered for close to 20 years. But it's a real drag for small/low power designs; having hurt early atom on the laptop and being an important factor in Intel's mobile aspirations failing a few years back. If the future of servers becomes chips hundreds of sub-watt cores; it may again become a problem for both Intel and AMD.

The "CISC baggage" matters for larger chips as well. It is the reason why Intel and AMD have included a uop cache in Sandy and Zen.

The x86 tax has two parts: a fixed and a variable. The fixed part affects more to small cores. The x86 tax not only affected Intel mobile desings, it is also part of the fiasco behing Larrabe/Phi or stuff as Quark.