News Arm Pioneer: Nvidia's Grace CPU Is Proof That It Will 'Compete Unfairly'

[Admin]

Arm co-founder Hermann Hauser said that Nvidia is "clearly showing it will compete unfairly" if it acquires the company.

Arm Pioneer: Nvidia's Grace CPU Is Proof That It Will 'Compete Unfairly' : Read more

 

[ezst036]

A lot of visionaries believe that ARM is finally the one who will displace x86.

Nvidia is the biggest threat to that new future. What good is breaking from x86 when you replace it with the likes of Nvidia? Meet the new boss. Same as the old boss.

 

[InvalidError]

Any company can license ARM's IP and then put together semi-custom silicon with whatever proprietary extra stuff they want tacked on or even license just the ISA with essential supporting IP then build their own CPU from scratch with any proprietary extensions they want for their specific application. Nvidia putting together their own CPU with its own proprietary 900GB/s interface doesn't take away other ARM licensees' ability to do the same.

 

[hotaru.hino]

As much as I don't like NVIDIA buying ARM, I also don't really buy the argument that they shouldn't get it simply because they have a "competitive advantage" with NVLink. There's nothing stopping any other ARM developer from incorporating some other high speed bus or making their own. Example: AMD developed HyperTransport in response to whatever Intel was doing.

 

[Gomez Addams]

This is a ridiculous argument. I am not particularly a fan of Nvidia but I fail to see how utilizing a proprietary interface is unfair in any way. No company is prohibited from also doing that so how is it unfair? ARM licensees do not all use the same pin layouts on their chips, nor are they required to, so I fail to see what the issue is.

 

[RandomGuy2]

Totally agree that this is a ridiculous argument! There might be many reasons to dislike NVidia owning ARM, but this is rubbish. Making NVLink available only to their ARM SoC is something NVidia could do with or without owning ARM, as NVLink is not ARM IP.

 

[thuckabay]

After the monopolitic <Mod Edit> that Amazon and Google pulled toward Parler, I can only say this: may nVidia clean their clocks! $crew Amazon and Google, as they have destroyed the goodwill they formerly enjoyed. I now recommend to everyone against those companies that have so fraudulently, dishonestly and dishonorably behaved toward conservatives; may they be broken-up and go out of business as the evil monopolies they have actually become.

 

[Dantte]

Boohoo, its so "unfair" that NVIDIA has better tech then everyone else, so everyone wants to use them... 'We have to bring NVIDIA down because were too incompetent to bring ourselves up.'

 

[waltc3]

Back in the 90's, the scuttlebutt was that RISC CPUs would "soon" be displacing CISC x86. Didn't happen then, when "x86" was actually something real. There are no "x86" CPUs today--neither Intel nor AMD makes CPUs on the original x86 ISA. Current "x86" CPUs today are RISC-CISC hybrids that carry a bit of the x86 instruction set for backwards compatibility purposes only, but which otherwise bear no resemblance to "x86" of the 80s/90s. The common misunderstanding is that current "x86" CPUs by AMD and Intel are static--sort of frozen and unchanging--but the reality is the exact opposite.

I think the UK would be well advised to nix the nVidia deal--over the last 25 years the company has shown a marked proclivity for cheating in various commercial and competitive memes and seems to have no qualms in using unfair and dishonest trade practices of all kinds. Microsoft found nVidia so disagreeable to work with that it dropped nVidia after the original xBox contract--which Microsoft had awarded nVidia. After the first xBox, Microsoft awarded the xBox contracts to ATi, now AMD, and left nVidia in its rear-view permanently. After 3dfx pioneered the first actually playable 3D GPU solutions in the late 90s, I watched nVidia level cheat after cheat against 3dfx in order to try and compete--as always, the cheating involved benchmarks almost exclusivity. Doesn't seem to have changed much, imo. The stories I could tell...!

 

[bkuhl]

A lot of visionaries believe that ARM is finally the one who will displace x86.

Nvidia is the biggest threat to that new future. What good is breaking from x86 when you replace it with the likes of Nvidia? Meet the new boss. Same as the old boss.

Who do you propose takes ownership of ARM then? Rumor already has it that Softbank was pressuring ARM to ratchet up their licensing fees:

SoftBank: Oi, we paid $32bn for you, when are you going to strong-Arm some more money out of your customers?

Failing mega-group discovers again it didn't buy the right biz

www.theregister.com

Who is going to be a "proper" owner in your eyes? It seems that Softbank no longer wants to deal with the slow and steady licensing free they get from ARM so they appear to want to unload it...

 

[Umfriend]

Totally agree that this is a ridiculous argument! There might be many reasons to dislike NVidia owning ARM, but this is rubbish. Making NVLink available only to their ARM SoC is something NVidia could do with or without owning ARM, as NVLink is not ARM IP.

But I wonder whether ownership of ARM would not incentivize nVidia to direct R&D of ARM towards proprietary stuff, i.e. further develop ARM such that it would only or mostly benefit nVidias proprietary IP?

 

[InvalidError]

But I wonder whether ownership of ARM would not incentivize nVidia to direct R&D of ARM towards proprietary stuff, i.e. further develop ARM such that it would only or mostly benefit nVidias proprietary IP?

I would expect that to be one of the core concerns most governments ARM needs approval from would have about this deal - Nvidia reserving most improvements it makes to ARM for itself, leaving its licensees who were relying on ARM's IP-core development for new CPUs, GPUs and other components stuck several months or even years behind.

 

[FakeMike]

This is a ridiculous argument. I am not particularly a fan of Nvidia but I fail to see how utilizing a proprietary interface is unfair in any way. No company is prohibited from also doing that so how is it unfair? ARM licensees do not all use the same pin layouts on their chips, nor are they required to, so I fail to see what the issue is.

The issue is that ARM is not producing it's chips whereas Nvidia is so they would be motivated to cripple competition's products (Qualcom, Samsung,Mediatek, Amazon). NVLink is only an example here.

 

[hotaru.hino]

The issue is that ARM is not producing it's chips whereas Nvidia is so they would be motivated to cripple competition's products (Qualcom, Samsung,Mediatek, Amazon). NVLink is only an example here.

MediaTek's the only one who use ARM's IP cores. The rest that's mentioned make their own.

EDIT: Had another look, Amazon also relies on ARM's IP cores, but I wouldn't be surprised if they're looking to branch from that.

 

[-Fran-]

I have to say, I agree with the general sentiment that this feels more like a weak rant than an actual argument towards preventing the merger.

I'm no fan of nVidia and its antics, but (unfortunately) this particular argument is weak sauce (to say the least). That is, unless we're missing something this fella knows about Grace. I would imagine the nVLink used in it has to be a mix of something from ARM (like an interconnect they also sell) they could have re-branded as nVLink? Remember ARM licenses things, but as soon as you change it slightly, marketing takes over and makes it a new shiny thing like ever before; case in point: rounded rectangles, SAM, nVLink (it's from 3DFX originally), etc.

Back in the 90's, the scuttlebutt was that RISC CPUs would "soon" be displacing CISC x86. Didn't happen then, when "x86" was actually something real. There are no "x86" CPUs today--neither Intel nor AMD makes CPUs on the original x86 ISA. Current "x86" CPUs today are RISC-CISC hybrids that carry a bit of the x86 instruction set for backwards compatibility purposes only, but which otherwise bear no resemblance to "x86" of the 80s/90s. The common misunderstanding is that current "x86" CPUs by AMD and Intel are static--sort of frozen and unchanging--but the reality is the exact opposite.

I think the UK would be well advised to nix the nVidia deal--over the last 25 years the company has shown a marked proclivity for cheating in various commercial and competitive memes and seems to have no qualms in using unfair and dishonest trade practices of all kinds. Microsoft found nVidia so disagreeable to work with that it dropped nVidia after the original xBox contract--which Microsoft had awarded nVidia. After the first xBox, Microsoft awarded the xBox contracts to ATi, now AMD, and left nVidia in its rear-view permanently. After 3dfx pioneered the first actually playable 3D GPU solutions in the late 90s, I watched nVidia level cheat after cheat against 3dfx in order to try and compete--as always, the cheating involved benchmarks almost exclusivity. Doesn't seem to have changed much, imo. The stories I could tell...!

One nitpick: don't confuse ISA with the underlying silicon architecture that implements the ISA. They're two different things. That is also why AMD got away with calling Bulldozer an 8 "core" CPU instead of "4FPU+8INT module weirdness" or "4F8IMW", heh. The way you define a CPU "core" is really ambiguous, based on early X86 implementations of the ISA that used (undertandably) integer register and units exclusively until the FPU made it's debut as the "math co-processor" (I think?) in 486DX and it was never part of the CPU "core". Just like we call the IMC the un-core, the FPUs, strictly speaking, are also un-core! Weird, isn't it?

Cheers!

 

[InvalidError]

The issue is that ARM is not producing it's chips whereas Nvidia is so they would be motivated to cripple competition's products (Qualcom, Samsung,Mediatek, Amazon). NVLink is only an example here.

NVLink isn't something anybody needs to put together an ARM-based CPU, it is just a proprietary interface for Nvidia's own stuff. Any ARM licensees can have 100 proprietary interfaces integrated into their semi-custom ARM-based SoCs if they want to put in the design effort for it.

What would be a problem is if Nvidia decided to stop Mali development and refuse to license any substitute tech, which would force ARM licensees to go elsewhere for updated graphics. Another thing that would hurt is if Nvidia introduced new instruction set extensions but refused to license those to licensees, implement it in their own updated ARM cores that they don't make available to licensees either, then use those Nvidia-exclusive new instructions to gain market share in markets previously served by ARM licensees. Nvidia could also keep its designs for future memory controllers to itself, forcing ARM licensees to design their own. Rinse and repeat with PCIe, USB and every single other IP block in ARM's portfolio that smaller ARM licensees rely on to put their semi-custom SoCs together.

 

[InvalidError]

The way you define a CPU "core" is really ambiguous, based on early X86 implementations of the ISA that used (undertandably) integer register and units exclusively until the FPU made it's debut as the "math co-processor" (I think?) in 486DX and it was never part of the CPU "core".

Math co-processors on x86 have been a thing since the 8086/8088 which paired with the (optional) 8087 for FP up to the 386+387 pair. The 486DX is the first Intel CPU with FP integrated on the same die (if we exclude the 487 which is basically a 486DX that disables the original CPU when installed) but otherwise mostly independent until the Core i-series where Intel decided to mix instruction classes across ALU execution ports and use a single shared scheduler for everything. AMD's Ryzen still uses separate scheduler and execution ports for FP/SIMD.

 

[Chung Leong]

That's a really dumb argument. Guy's getting it completely backward. The danger of Nvidia acquiring ARM is that the company will unfairly compel ARM licensees to adopt technologies like NVLink. Proprietary tech staying proprietary isn't a problem--that's what happens normally.

 

[c2p2]

This is pure clickbait. The argument is so weak that it is laughable at the author's intelligence.

 

[digitalgriffin]

Arm co-founder Hermann Hauser said that Nvidia is "clearly showing it will compete unfairly" if it acquires the company.

Arm Pioneer: Nvidia's Grace CPU Is Proof That It Will 'Compete Unfairly' : Read more

The problem becomes "How long until NVIDIA inserts proprietary technology which they hold the sole patents on (ie: NVLInk) and makes it mandatory?" And then uses that patent to charge outrageous sums?

NVIDIA has shown time and time and time again that is has ZERO qualms about screwing people over when they can get away with it. From bad solder on motherboards, to defective nForce chipsets, to dropping support for PhysX on cards AFTER they acquired PhysX. Remember that time a 3rd party developer enabled support and NVIDIA threatened to sue? And you have to buy RTX because only RTX support AI audio (bull). GSync compatible labeling requirements (to divert away it was an AMD technology) NVIDIA partner programs. The list just keeps going on and on and on.

The professional business market is HIGHLY lucrative. And like all highly lucrative payouts (ie: Gov't) they are less concerned about price. So NVIDIA can really bend some people over if they want to. Given their past behavior, I have little faith in NVIDIA to not get greedy with the corporate world. They couldn't even play nice with Apple and Microsoft. And those are two giants you don't want to tick off.

NVIDIA wants arm for a reason. And with all things NVIDIA, follow the cash cow and how to get there. Nothing is done in benevolence for the general good.

 

[Gomez Addams]

The issue is that ARM is not producing it's chips whereas Nvidia is so they would be motivated to cripple competition's products (Qualcom, Samsung,Mediatek, Amazon). NVLink is only an example here.

Except Nvidia does not produce chips that compete with those companies. They don't make SoCs, modems, or CPUs and they do not have a foundary either. They would not be motivated to cripple anyone because that is a) illegal and b) bad for business since some of their license agreements involve royalties.

 

[-Fran-]

Except Nvidia does not produce chips that compete with those companies. They don't make SoCs, modems, or CPUs and they do not have a foundary either. They would not be motivated to cripple anyone because that is a) illegal and b) bad for business since some of their license agreements involve royalties.

CUDA. That's all you need to know about nVidia's "willingness" to cooperate with others.

It cracked me when in the GTC Jen said "what makes ARM great is its openness". Like... Holy cow, how can you say that with a straight face, dude... No shame at all.

Cheers!

 

[hotaru.hino]

Except Nvidia does not produce chips that compete with those companies. They don't make SoCs, modems, or CPUs and they do not have a foundary either. They would not be motivated to cripple anyone because that is a) illegal and b) bad for business since some of their license agreements involve royalties.

NVIDIA makes Tegra, they have their own ARM implementation in Project Denver, and they're exploring a RISC-V implementation (though they've made a custom RISC CPU named FALCON), so NVIDIA does indeed make SoCs and CPUs. They don't make modems, sure, but neither does most of the other ARM developers (the only that makes their own ARM CPU core and a modem is Qualcomm).

So it would definitely be in NVIDIA's interest to buy ARM. And I would have the same position as I do against NVIDIA if Apple, Samsung, Qualcomm, MediaTek, or any of ARM's other licensees wanted to buy them.

 

[TCA_ChinChin]

Don't really think this is a good argument against Nvidia, although I understand the reasoning behind many UK objections to the acquirement. At a time where every sovereign nation is trying to secure semi-conductor self-sustenance, I can sympathize with the UK for wanting to keep tech company in country. Same reason why the US wants fabs built in the US or why China and Russia keep trying to develop their own CPU architectures and/or fabs.

 

[zodiacfml]

yeah this makes Nvidia more powerful but makes compete better to Intel and AMD in the server space, and leave the Arm server vendors behind