Driver EULA update: No Datacenter Deployment

GeForce and Titan driver EULA seems to be updated (http://www.nvidia.com/content/DriverDownload-March2009/licence.php?lang=us&type=GeForce) with this statement:

What is a datacenter?

What if blockchain processing in a datacenter is permitted? Can I use SOFTWARE (GeForce/Titan driver) for non-blockchain processing purposes in such datacenters?

As in all such cases, only a lawyer can give you suitable advice on how to interpret such contractual language, including whether these provisions are enforceable in your jurisdiction.

I am not a lawyer. What I am (speculatively!) reading between the lines is: We (NVIDIA) spent a lot of money developing deep learning software, and we can’t get a decent return on that investment in the consumer market where product prices have to be set according to competitive gaming performance, so we want you to use our more expensive professional products for industrial scale deep learning uses.

Apparently (again, speculation!) NVIDIA feels that large-scale crypto mining on consumer products is fine because they spent no money to penetrate that market, so it provides a nice incremental boost to GeForce sales at no cost to them.

I will further speculate that it will only be a question of time until some clever people create a deep learning framework based on blockchain :-) [In case it’s not clear: my attempt at humor]

I fully agree with this speculation, but also have a feeling that lawyer can’t give us answer about this ambiguity of the word ‘datacenter’ without any specific guidance from Nvidia.

What if a company let its employees, each of them, have their own computing workstation and do some research? Does that count as ‘Datacenter Deployment’? If not, this will be quite easy proxy to this new EULA, which is against Nvidia’s intent.

This should be much more specific than current text.

As I already stated in a different recent thread on EULA terminology, I doubt that anybody from NVIDIA’s legal department is following these forums. As a consequence, these forums would not be a suitable feedback path for such issues, nor should we expect to see clarifications of the EULA terms to be posted here by NVIDIA representatives. I think contact information for NVIDIA’s legal department is provided in the EULA.

These forums are intended as an information-exchange platform for the CUDA developer community where users can assist other users with technical issues.

I suspect that if you were with an IT specialist of a major enterprise, and he/she said “let me show you our datacenter” and you walked into a room full of server racks, none of you would be surprised.

I suspect that if the same IT specialist said “let me show you our datacenter”, and you walked into a cubicle farm with a workstation under each desk, you’d be surprised. Huh?

I suspect, in fact, most people who know what GPU computing is, know pretty well what a datacenter is.

I also suspect, that the vast majority of folks asking questions have never actually been concerned about whether they were adhering to the EULA, and have no actual intention of doing so now. I also suspect, that those who wish to follow the EULA can make pretty good sense of what is written there now, and discern its intent.

Although the signal-to-noise ratio is pretty low, it’s evident from reading other forums that there are some level-headed people out there who are reasonable in their expectations, and have done a pretty good job of discerning the intent:

[url]https://www.reddit.com/r/nvidia/comments/7m0gtn/new_nvidia_eula_prohibits_deep_learning_on/[/url]

Yes, you do have to pick out the signal from the noise. I think its not hard to do.

The fact that NVIDIA does not expect GeForce products to be used en-masse in server farms is not a secret or a mystery, and has been part of their public positioning for a long time. This merely brings the EULA into consistent alignment with the rest of their positioning.

https://devblogs.nvidia.com/parallelforall/12-things-tesla-accelerated-computing-platform/

From that last article, Microway seems to have understood the positioning, a long time ago:

[b]Warranty
NVIDIA’s warranty on GeForce GPU products explicitly states that the GeForce products are not designed for installation in servers. Running GeForce GPUs in a server system will void the warranty. From NVIDIA’s manufacturer warranty website:

Warranted Product is intended for consumer end user purposes only, and is not intended for datacenter use and/or GPU cluster commercial deployments (“Enterprise Use”). Any use of Warranted Product for Enterprise Use shall void this warranty.[/b]

But now we suddenly have no idea what a datacenter is? Or what NVIDIA could possibly mean with this incredibly bizarre and arcane wording?

I don’t find the linked Reddit thread to be particular enlightening, in fact I find it useless. I would tend to agree that those organizations which NVIDIA likely meant to address with the added language about data centers will know exactly that this is directed at them and will have no problems parsing the EULA’s legalese (and have a sizeable legal department of their own to parse it for them, should there be any doubts).

But I can understand if the added language raises questions in some people’s mind. If an organization has a server room with just one rack full of GPU-accelerated servers, is that already considered a data center for the purposes of this EULA? What about two such racks? Five? Ten?

Warranties are a different kettle of fish from the EULA provisions. From comments I have gotten in these forums before, organizations that use large amounts of consumer-grade GPUs for large-scale processing seem to operate under a model of “we do not care for hardware warranties; if the hardware breaks we’ll just toss it; still much cheaper than deploying Teslas”.

I don’t want to discuss NVidia motivation :-) I just want to understand EULA. Because without understanding I can’t agree with it, and so - I can’t install driver. On EULA page I can find only “contact us” in footer - info@nvidia.com. I sent email and will post response here. And this is why I think that discussion on this forum is viable - I hoped that somebody already clarified this statement. Seems that this is not true, so I will post here response and anybody who can’t install driver because of non-clear EULA will be able to google clarification.

Of course I googled a lot - and reddit contains a lot of interesting comments, but there is no useful information.

While big companies with big legal departments are able to do research for understanding these statements - smaller companies which tries to do their best to not violate any patents or license terms (and private individuals/universities research labs/etc.) - are not able.

P.S. russian version of EULA differs a lot (http://www.nvidia.ru/content/DriverDownload-March2009/licence.php?lang=ru&type=GeForce):

This can be translated as:

In #4 I explained why I think that this is likely an unrealistic assumption / expectation.

Seems that I was unclear - I mean that probably somebody already contacted info@nvidia.com (for example) and already got the clarification.

I see I misunderstood. In general, that could be somewhat likely. But considering that this issue popped into public awareness only 48 hours ago and that we are currently between Christmas and New Years in the US, still quite unlikely.

OK, if you have 1000 racks, that’s clearly a datacenter. I think the real question is:

  1. What if you have just 1 server room in your building? Is it still a datacenter?
  2. What if you have just 1 rack (even no special "server room")?
  3. Just 4 3-unit servers mounted in a small wheeled rack?
  4. Just 4 tower servers on your table, 4 GPUs each?

In any case, just a lousy business decision. I can see voiding the warranty of geforce cards used in data centers, but to actually prohibit their use is ridiculous. Come on NVIDIA, you are better than this! Do you really think this is the best thing to do to establish longterm business relationships? Are you really so confident that other vendors aren’t capable of competing at the data center level at some point? Why open the door so wide? Short term greed?

Great news if you own a mining datacenter. Otherwise, not so much.

There were some news over the internet with such statements of NVidia spokesperson (source - CNBC):

But NVidia explicitly allows to use them for blockchain processing in data centers in the same EULA sentence. While blockchain processing in many cases much more close to 100% 24/7 GPU utilization than typical GPGPU application.

Moreover NVidia uses GeForce GTX 1060 and 1080 for its own GeForce Now (source - nvidia.com):

Isn’t it a twice doublethinking?

Moreover, from the same NVidia spokesperson statements (source - CNBC):

Okey, fine, how customer can contact NVidia? I tried via this topic and via info@nvidia.com and got only response “Thank you for your inquiry. Although we aren’t always able to respond to all inquiries, we do appreciate your interest. …”.

This is ridiculous.

NVidia changed EULA - ok, I just want to know what exactly datacenter is and what if blockchain processing in a datacenter is permitted.