NVIDIA Cracks Down on CUDA Translation Layers, Changes Licensing Terms

[john_]

People will always figure out a way to circumvent a closed ecosystem. The only way to prevent this is by making those circumventions illegal via your EULA, aka protectionism, which inherently decreases the appeal of your ecosystem as fewer people can use it. And eventually, you'll step too far, and someone will challenge your EULA in court, and you'll lose. This is exactly what has happened to Apple - after being allowed to dictate exactly how people use its products for far too long, the EU has finally stepped up to enforce consumers' rights. Court cases are bad for share prices, and especially bad if the company loses.

Therefore, closed ecosystems always ultimately fail, which means they are ultimately stupid. Expending massive amounts of effort in an attempt to protect them will ultimately fail, so this is also stupid.

The smart thing to do would be to start negotiating with other GPU vendors to come up with a shared open standard that just so happens to look a lot like CUDA, and just so happens to require very minimal translation from CUDA. Boom, you've immediately got a rung up on the competition because they have to expend way more effort to make their cards work with your not-CUDA. And because not-CUDA is so similar to CUDA and so easy to port over, users who are already using CUDA won't be unduly affected. As a bonus, because this standard is based on CUDA, NVIDIA would wield a lot of control over it - which is exactly what they want anyway.

Unfortunately, corporations rarely make logical decisions.

Depends. Depends on how much you gain from such an attitude until someone challenges you in court, fine you a couple of billions, or decides -as a consumer- to ditch you for a competitor. Depends on how much you have become a necessity. And right now, Apple and Nvidia can enforce restrictions, put whatever prices they like, ignore competition and fines, or those 2-3 consumers who decide to go to the competitor, while keep making billions and expanding their influence and their image of being a necessity and the only option.
This move from Nvidia will only favor Nvidia, because it wouldn't hurt CUDA in any way. CUDA has become something like the only option for many. Think Windows and Microsoft. Microsoft has done it's mistakes the last 25 years (like Vista or Win8) and in every mistake, people where expecting it will lose market share, with Linux being the winner. Well, in desktops and laptops it's only losing against another monopoly, Apple. Linux is still at low single digit. This move from Nvidia comes just a few days after that post from Herkelman saying that Nvidia is a cartel, showing that Nvidia doesn't really cares if people realize that it acts like a cartel. On the contrary.

 

[AnotherReader]

X86 emulation is allowed. Emulation of X86 ISA extensions that are still under patent protection is allowed if you're Microsoft or Apple. Emulation of Arm ISA, for example - I don't know but Arm ISA is a major product they make money on, by selling it to architectural licence holders. QEMU is inefficient enough and Arm doesn't see it as a threat.

Even Transmeta was allowed to release processors that ran x86 code. Wine is a better example, and while Microsoft hampered it, they never sued anyone. When you are exceeding peak Microsoft levels of asshattery, you have to stop and think if what you're doing is right in the long run.

 

[R-T-B]

Such move is probably illegal anyway. There is even a precedent in US law, when IBM's hdds got reverse engineered in 196x.

Why reference such an ancient case when we have Oracle v Google and it decided that yes, these things are legal, at the level of the Supreme Court no less?

No I don't agree with it either but it seems to be law right now.

Even Transmeta was allowed to release processors that ran x86 code. Wine is a better example, and while Microsoft hampered it, they never sued anyone. When you are exceeding peak Microsoft levels of asshattery, you have to stop and think if what you're doing is right in the long run.

All those are really in a legal grey area since Oracle v Google. Wine basically exists because Microsoft doesn't want the bad PR of suing it. Transmeta was made before the Oracle v Google case so they didn't know what hot water it was getting into.

 

[AnotherReader]

Why reference such an ancient case when we have Oracle v Google and it decided that yes, these things are legal, at the level of the Supreme Court no less?

No I don't agree with it either but it seems to be law right now.


All those are really in a legal grey area since Oracle v Google. Wine basically exists because Microsoft doesn't want the bad PR of suing it. Transmeta was made before the Oracle v Google case so they didn't know what hot water it was getting into.

Didn't Oracle vs Google find in Google's favour, but sidestep the issue of copyrighting APIs?

 

[R-T-B]

Didn't Oracle vs Google find in Google's favour, but sidestep the issue of copyrighting APIs?

Reading the detailed wiki article, it seems you are right. They basically allow it in a very narrow "fair use" context.

 

[Minus Infinity]

Didn't Oracle vs Google find in Google's favour, but sidestep the issue of copyrighting APIs?

I'd like Nvidia to prove to the Supreme Court using CUDA on other hardware is not fair use. Google won the case 6-2 so Nvidia would have a hell of time getting them to overturn their own decision.

Nvidia can write what they want, but they have no legal standing. No one is reverse engineering their API.

Google vs Oracle

 

[trsttte]

They basically allow it in a very narrow "fair use" context

Not really that narrow, the four pillars of fair use have been well described, there's loads of room for interpretation but clear guidelines exist for what can constitute fair use.

I'd like Nvidia to prove to the Supreme Court using CUDA on other hardware is not fair use. Google won the case 6-2 so Nvidia would have a hell of time getting them to overturn their own decision.

Nvidia can write what they want, but they have no legal standing. No one is reverse engineering their API.

Google vs Oracle

A saving grace for google was how little of java api's they've used and how many different things they did with it. A cuda wrapper would need to cover almost everything and try to do as close to the same as cuda as possible so the case would lead to different discussions

 

[Wirko]

CUDA, Wine, Java, x86 - we have mentioned several [legal] cases that have little in common between them to make any predictions on this latest one. The case of x86 is probably the most similar.

What exactly is Nvidia trying to protect here? Not the ISA - it remains secure by being obscure enough, as of now. Not the CUDA API - at least Tom's Hardware notes that "Recompiling existing CUDA programs remains perfectly legal. To simplify this, both AMD and Intel have tools to port CUDA programs to their ROCm and OpenAPI platforms, respectively."

So they are legally protecting their compiled binaries, or rather the IP contained within. The big precious bunch of optimisations and whatever else. And how much does this matter? Full compilation is necessary anyway if one wants the best performance, and we don't buy accelerators for mediocre performance, right?

 

[john_]

Nvidia can write what they want, but they have no legal standing.

Do they care if they can have any legal standing?
For example. Let's say that I go to my neighbor and I say "Look, we are good neighbors but if you keep watering your garden at 10 pm I will call the police and sue you". Even if my neighbor knows that the police will do nothing and I have no legal base in what I say, he still might stop or consider stopping watering his garden at 10 pm just to avoid ending up with a hostile neighbor. And in my opinion, this is what Nvidia tries to do. Scare people.

 

[RUSerious]

Why stupid? I remember well when/how it started - it was Cg or C for graphics on GeForce3 and much time and effort went into producing the tool CUDA is today, so no wonder nvidia defends it. AMD and intel surely didn't chip in for development... Make sense.

Yeah, started out with researchers using shader engines for more general compute. CUDA debut was 2007 and the GF 8800 series really started to get the ball rolling with better compute capabilities. I think it would be a case of irresponsible fiduciary execution if the company didn't protect their investors by maximizing profits. Nvidia could be sued for that. So, they are not really being mean, or even greedy - they are acting in the best interests of it's investors and their employees.

 

[Wirko]

So I went looking up Wiktionary to maybe find zluda in Croatian/Serbian, and found this in Polish instead. That guy knew how to pick a name so closely related to AI.