Arm Co-Founder Doesn't Think NVIDIA Owning the Company Would be in Its Best Interests

[Raevenlord]

Arm co-founder Hermann Hauser recently gave an interview to BBC where he expressed some concerns regarding the prospective buy acquisition from NVIDIA, which has been in talks with Arm owner Softbank towards the IP-designer's acquisition. As Hauser puts it, "It's one of the fundamental assumptions of the ARM business model that it can sell to everybody," Hauser told BCC, "The one saving grace about Softbank was that it wasn't a chip company, and retained ARM neutrality. If it becomes part of Nvidia, most of the licensees are competitors of Nvidia, and will of course then look for an alternative to ARM."

Hauser doesn't think the NVIDIA deal will follow through due to these aspects of the chip design ecosystem, with many Arm clients - such as Intel, Apple, Qualcomm, TSMC, Samsung, among others - being direct or otherwise indirect competitors to NVIDIA. Hauser thinks that Arm would be much better served through a British government intervention in bringing the company back towards the British fold: "The great opportunity that the cash needs of Softbank presents is to bring ARM back home and take it public, with the support of the British government." The Softbank acquisition occurred back in 2016 and cost the company some $24 billion; however, recent estimates from New Street Research LLP placed Arm's valuation at USD $44 billion if its IPO took off in 2021, and as much as $68 billion by 2025.



View at TechPowerUp Main Site

 

[john_]

Everyone loves Nvidia. I wonder why.....

 

[Fluffmeister]

Well Nvidia is about as neutral as it gets compared to that lot. Intel hold the x86 license and are happy with their duopoly with AMD whilst also trying to push further into GPU land. Apple compete directly with Samsung and Qualcomm in the mobile market so neither should hold any stake in Arm because the other parties would go into meltdown.

All in all, good to see the UK shake up the tired x86 market.

 

[Recus]

First they sold ARM and now thinking to bring it back in post Brexit Britain?

Intel, Apple, Qualcomm, TSMC, Samsung

They don't use ARM for PC/HPC. More like deep state conglomerate is afraid of Nvidia's domination. /s?

 

[Legacy-ZA]

"Arm Co-Founder Doesn't Think NVIDIA Owning the Company Would be in Its Best Interests"


I could have told you that.

 

[midnightoil]

They don't use ARM for PC/HPC. More like deep state conglomerate is afraid of Nvidia's domination. /s?

Yes they do. In both.

The uncore in all Intel x86 chips has been based on ARM IP since the early days of the Core chips. Many of their iGPUs are based on ARM designs. Intel's newer iGPUs which are in house still contain ARM IP. I'd be very surprised if their new graphics line up don't have significant ARM licensing.

Most of AMD's security stuff on their x86 and graphics cards goes through ARM trust designs.

There are increasing numbers of full ARM (on CPU side) supercomputers. Including exascale designs. Fujitsu's new design is the most powerful / efficient in the world for its workload.

There are very few major commercial compute, graphics, communications or controller chips that don't have ARM IP in them.

They're the most pervasive hardware design IP in the world, and arguably the most influential.

 

[PowerPC]

This is just a dance. ARM is hot right now and NVIDIA has given it even more wind. Naturally ARM (Softbank) wants to capitalize on the situation. "Yea, this is not such a good idea" just means "NVIDIA doesn't want to pay us enough money anymore".

But it's probably just British seller's remorse on this one. The Dude has got a point though, that it should just probably remain a pure licensee. That could really be the reason why ARM is gaining this momentum over x86.

Imagine a world with only ARM-compatible chips where it would be easy for everything to work cross-platform. And anybody could make their own ARM-compatible designer CPUs just like now (but people would actually do it). This parallel universe would never exist if any proprietary chip-company ever buys ARM, especially NVIDIA. But any other company that is interested in selling their own produces would be just as detrimental for that.

 

[Fluffmeister]

But it's probably just British seller's remorse on this one.

It's what we do, we create iconic brands, iconic sports, hell... we create iconic countries, and then send them off on their own and watch everyone fight over them.

 

[Steevo]

This is just a dance. ARM is hot right now and NVIDIA has given it even more wind. Naturally ARM (Softbank) wants to capitalize on the situation. "Yea, this is not such a good idea" just means "NVIDIA doesn't want to pay us enough money anymore".

But it's probably just British seller's remorse on this one. The Dude has got a point though, that it should just probably remain a pure licensee. That could really be the reason why ARM is gaining this momentum over x86.

Imagine a world with only ARM-compatible chips where it would be easy for everything to work cross-platform. And anybody could make their own ARM-compatible designer CPUs just like now (but people would actually do it). This parallel universe would never exist if any proprietary chip-company ever buys ARM, especially NVIDIA. But any other company that is interested in selling their own produces would be just as detrimental for that.

ARM isn't getting faster at general computing or efficiently, they are RISC designed CPUs that excel at specific tasks VS X86 CISC designs that are good at everything, and also happen to be getting "Acceleration" in the form of RISC hardware.

Designing a ARM cpu for most is simply checking boxes on features. Its ability to power gate and small add in design with small caches makes it power efficient, not some magic that comes with the ARM logo.

That result was from one high performance linpack run using the SVE vector extension hardware that is only used in that chip. By the same thought process the fastest X should also be great at Y, but its not the case.

 

[PowerPC]

That result was from one high performance linpack run using the SVE vector extension hardware that is only used in that chip. By the same thought process the fastest X should also be great at Y, but its not the case.

I'm not a chip designer, so this tells me very little. But your argument seems to be similar to a lot of people who say ARM isn't going to replace x86 and I agree with you up to a certain point. Theoretically, it won't replace it because yes, it still has its strengths. But this argument doesn't answer what the actual market will do with ARM. Most normal user applications today are already fitting for RISC architecture. Checking boxes like 4K video and audio streaming, compiling software, video editing etc. If those things are checked, most people will benefit from faster and cooler devices with ARM.

So, it may not replace x86, but how much space will it take from x86 where x86 doesn't make sense anymore? For the average consumer or someone who does certain specific tasks for their work, ARM would be the way to go. And these are most people right there.

Look at how phones have gotten more and more features over the years and more of those boxes checked. At this point, the line between phones and PCs are getting blurrier and blurrier. Nobody is asking for a CISC phone and it has made perfect sense for years to use RISC. It has definitely been already battle-tested and it'll only make more and more sense to switch to ARM as time and development on the right checkboxes progress. If you buy what you need and it checks your personal boxes better than x86, I'd strongly argue most people will buy that and not care that x86 has an advantage at being good for everything. People have specific needs and that's exactly how ARM may make x86 irrelevant or at least relegate it to the ones who really need that advantage of x86 you describe.

 

[xrror]

Whhaaaaahhhh.... gotta love it when a company goes public and essentially "sells themselves out" to get that fat capitol. Then whine when later shareholders dump them / do whatever to max profit vs. some idealogical ideal they former owners had. That's the trade-off, unless the original owners have some way to keep 51% control somehow that's how the game is played sadly.

That said I'm not sure on the original conditions where when Arm sold to SoftBank, but without that majority shareholding you don't control jack anymore.

 

[TheoneandonlyMrK]

I'm not a chip designer, so this tells me very little. But your argument seems to be similar to a lot of people who say ARM isn't going to replace x86 and I agree with you up to a certain point. Theoretically, it won't replace it because yes, it still has its strengths. But this argument doesn't answer what the actual market will do with ARM. Most normal user applications today are already fitting for RISC architecture. Checking boxes like 4K video and audio streaming, compiling software, video editing etc. If those things are checked, most people will benefit from faster and cooler devices with ARM.

So, it may not replace x86, but how much space will it take from x86 where x86 doesn't make sense anymore? For the average consumer or someone who does certain specific tasks for their work, ARM would be the way to go. And these are most people right there.

Look at how phones have gotten more and more features over the years and more of those boxes checked. At this point, the line between phones and PCs are getting blurrier and blurrier. Nobody is asking for a CISC phone and it has made perfect sense for years to use RISC. It has definitely been already battle-tested and it'll only make more and more sense to switch to ARM as time and development on the right checkboxes progress. If you buy what you need and it checks your personal boxes better than x86, I'd strongly argue most people will buy that and not care that x86 has an advantage at being good for everything. People have specific needs and that's exactly how ARM may overtake x86..

There's a slight bias implied here, arm can improve performance yes but, X86 is also improving on efficiency and performance, I don't think arm can catch it for raw general purpose compute personally giving x86 a shoe in for years yet.

 

[PowerPC]

There's a slight bias implied here, arm can improve performance yes but, X86 is also improving on efficiency and performance, I don't think arm can catch it for raw general purpose compute personally giving x86 a shoe in for years yet.

This is the big question, but people at Apple seem to think it can and if I'd had to bet on it today, I'm willing to believe them. It could theoretically happen this year when Apple releases its first Mac with ARM. From the rumors I've heard, it will do very well, if not even overcome it and still be cooler and more efficient while doing it. I'm very surprised how people still don't believe this can happen at this point with all the leaks from their developer kits. But people also never believed AMD could overcome Intel and it just took Zen and a few short years to do it.

 

[semantics]

ARM isn't getting faster at general computing or efficiently, they are RISC designed CPUs that excel at specific tasks VS X86 CISC designs that are good at everything, and also happen to be getting "Acceleration" in the form of RISC hardware.

Designing a ARM cpu for most is simply checking boxes on features. Its ability to power gate and small add in design with small caches makes it power efficient, not some magic that comes with the ARM logo.

That result was from one high performance linpack run using the SVE vector extension hardware that is only used in that chip. By the same thought process the fastest X should also be great at Y, but its not the case.

PowerPC, MIPS, ARM all have pretty wide applications throughout their lifetime. It's just a matter if you can make it work, hardware and software. I would say the biggest thing having x86 as the popular pc processor is software, legacy crap is the reason why windows is popular, it's also what makes windows bad.

 

[TheoneandonlyMrK]

This is the big question, but people at Apple seem to think it can and if I'd had to bet on it today, I'm willing to believe them. It could theoretically happen this year when Apple releases its first Mac with ARM. From the rumors I've heard, it will do very well, if not even overcome it and still be cooler and more efficient while doing it. I'm very surprised how people still don't believe this can happen at this point with all the leaks from their developer kits. But people also never believed AMD could overcome Intel and it just took Zen and a few short years to do it.

Certainly in some accelerated applications they will shine, they always have, in use some find the limits before others.
Some don't see the limits, like a friend of mine, sold on Apple ,she is truly content with what she's already using.
Apple will just dilute the Mac brand into big performance phones, why have two OS after all.

Some software innovation sound like a dream come true but don't pan out, like steam Os.

 

[moneyman8]

why have two OS after all

Yup, developers can create one application and deploy for both Mac and iOS with Apple Silicon in both. Imagine many developers will develop both versions of apps with the same codebase. Then, the OS will become more similar than ever.

 

[Crackong]

Rogue China division problem solved ?

 

[watzupken]

Well Nvidia is about as neutral as it gets compared to that lot. Intel hold the x86 license and are happy with their duopoly with AMD whilst also trying to push further into GPU land. Apple compete directly with Samsung and Qualcomm in the mobile market so neither should hold any stake in Arm because the other parties would go into meltdown.

All in all, good to see the UK shake up the tired x86 market.

When I think of neutral, it means that if ARM gets acquired by Nvidia there should not be any concerns about anti competition or restrictions of IP to licensee. Now looking at Nvidia's historical pattern, they are certainly not a neutral party by any means. Between Nvidia, Samsung, Apple and Qualcomm, if any of these are to acquire ARM, I feel Samsung may be the better candidate to acquire ARM since they are also supplier of many other components to other companies.

 

[PowerPC]

Certainly in some accelerated applications they will shine, they always have, in use some find the limits before others.
Some don't see the limits, like a friend of mine, sold on Apple ,she is truly content with what she's already using.
Apple will just dilute the Mac brand into big performance phones, why have two OS after all.

I don't think Apple is so bad. Not sure where you're coming from. To me OSX / MacOS >>>> Windows.

It's unix based, so already much more user friendly in my opinion than what Windows has. But I'm generally a unix / linux guy.

Apple could make something really good with this move, if it pans out. Improve their own products and image and also elevate the industry by showing how to do ARM well in the PC / Laptop sector. So if it really turns out to work better (which Apple is betting on), I think the adoption by others will be very quick.

 

[watzupken]

I don't think Apple is so bad. Not sure where you're coming from. To me OSX / MacOS >>>> Windows.

It's unix based, so already much more user friendly in my opinion than what Windows has. But I'm generally a unix / linux guy.

Apple could make something really good with this move, if it pans out. Improve their own products and image and also elevate the industry by showing how to do ARM well in the PC / Laptop sector. So if it really turns out to work better (which Apple is betting on), I think the adoption by others will be very quick.

The problem is that Apple only cares about their own ecosystem. As you can tell, they have never sold their A series SOC to anyone and will not be selling their ARM based processors for laptop as well. So it makes little sense for them to acquire ARM (at least not as a majority shareholder). They are likely contented being a licensee as long as there is no immediate threat to them. I am pretty sure Apple would have factored in the uphill challenges that awaits Nvidia or any potential buyer by the respective regulators.

 

[Vya Domus]

Yes they do. In both.

The uncore in all Intel x86 chips has been based on ARM IP since the early days of the Core chips. Many of their iGPUs are based on ARM designs. Intel's newer iGPUs which are in house still contain ARM IP. I'd be very surprised if their new graphics line up don't have significant ARM licensing.

Most of AMD's security stuff on their x86 and graphics cards goes through ARM trust designs.

There are increasing numbers of full ARM (on CPU side) supercomputers. Including exascale designs. Fujitsu's new design is the most powerful / efficient in the world for its workload.

There are very few major commercial compute, graphics, communications or controller chips that don't have ARM IP in them.

They're the most pervasive hardware design IP in the world, and arguably the most influential.

You are dramatically overestimating the use and importance of ARM IP in the x86 space and even then AMD and Intel seem to be capable of doing what ARM can't with IP of their own.

 

[Xaled]

Wow! TPU changed the title from "Disaster" to "not best interest"

I really wonder what interests with Nvidia you have because your choice of word against other brands is so harsh and click-bait while your are being so soft when it is
nvidia.

 

[midnightoil]

You are dramatically overestimating the use and importance of ARM IP in the x86 space and even then AMD and Intel seem to be capable of doing what ARM can't with IP of their own.

The x86 fanboy speaks. I haven't overestimated anything. I stated what their designs and IP is used for.

Also, you should take a look at Amazon's latest in-house ARM server chips ... they give Intel an absolute hiding, and beat EPYC first gen in many metrics.

 

[Vya Domus]

The x86 fanboy speaks. I haven't overestimated anything. I stated what their designs and IP is used for.

Also, you should take a look at Amazon's latest in-house ARM server chips ... they give Intel an absolute hiding, and beat EPYC first gen in many metrics.

And the ARM fanboy tried his best but failed horribly by comparing AMD's first gen 32 core chips with the 64 core Gravitation. That's with twice as many cores, twice, it better beat first generation EPYC otherwise it would have been absolute garbage.

 

[Steevo]

I'm not a chip designer, so this tells me very little. But your argument seems to be similar to a lot of people who say ARM isn't going to replace x86 and I agree with you up to a certain point. Theoretically, it won't replace it because yes, it still has its strengths. But this argument doesn't answer what the actual market will do with ARM. Most normal user applications today are already fitting for RISC architecture. Checking boxes like 4K video and audio streaming, compiling software, video editing etc. If those things are checked, most people will benefit from faster and cooler devices with ARM.

So, it may not replace x86, but how much space will it take from x86 where x86 doesn't make sense anymore? For the average consumer or someone who does certain specific tasks for their work, ARM would be the way to go. And these are most people right there.

Look at how phones have gotten more and more features over the years and more of those boxes checked. At this point, the line between phones and PCs are getting blurrier and blurrier. Nobody is asking for a CISC phone and it has made perfect sense for years to use RISC. It has definitely been already battle-tested and it'll only make more and more sense to switch to ARM as time and development on the right checkboxes progress. If you buy what you need and it checks your personal boxes better than x86, I'd strongly argue most people will buy that and not care that x86 has an advantage at being good for everything. People have specific needs and that's exactly how ARM may make x86 irrelevant or at least relegate it to the ones who really need that advantage of x86 you describe.

I think you are missing the point.


ARM, if it doesn't have the hardware to perform X or Y function built in its performance is horrible, or it simply will not run. How many original Arm CPU phones wouldn't run flash despite being used extensively at the time? It run just fine on almost every computer from old to new?

Jazelle - Wikipedia

en.m.wikipedia.org

Java was/is another one. Supported on essentially every X86 platform, but the specific hardware had to be added to ARM cpus.

Planned obsolescence by virtue of evolving software. You want to buy a computer that is only good for 2 to 3 years before games, streaming services, applications, etc stop working?