Nvidia EULA limits GeForce Data Center Usage *fork*

You can control how they are sold. You can't control how they are used after they have been sold. Ford cannot create a line of cars for Uber use and refuse use of people's private motors for Uber. They could at best tell their suppliers to check what it's being used for and enter into contracts with their car purchasers.

Anyone still arguing that post-consumer use of a bought product can be controlled needs to present some evidence at this point. It's agreed that licensing (contracts) can be used to control use, and support services can be withheld. No issues with nVidia doing that. But the notion that anyone can control what happens once a product is bought (not licensed) with that product is to my mind impossible through our laws and I'd need some actual evidence to the contrary. Every argument thus far has looked at licensing and service providing.

And going back to the original article, nVidia are perfectly in their rights now I've looked at the thing () because it's a term in the T&Cs regards the use of the software. Sakura can continue using their old drivers. Edit: But as DmitryKo points out, enforceability of such EULA's is an unknown.

 

EULA exists to limit the liability of hardware/software provider in case of failures and errors - provisions that restrict user activities like modifying, reverse engineering, redistributing etc. so far have not been successfully enforceable without some intricate hardware/software protection system, even though these provisions are based on protections provided by copyright and intellectual property laws.

This new restriction OTOH is based on some arbitrary usage pattern and not an established legal framework or case law, so it would be largely ignored by anyone - unless NVidia tries to enforce it somehow in their driver code, at which point it would be brought in a court with a good chances of being overruled.

 

Car makers sell hardware, software and support... Just because car software is a non-issue doesn't make it disappear.

You see, what you are defending is absolutely anti-consumer and the same logic would allow a company to prevent me from driving on... "Tuesdays" or if I have "ice cream on the trunk". It's so subjective it's astounding.

 

Like a software company that prohibits consumers the right to decide what updates they eventually end up with on their devices, for better or worse auto-updating will decide. Consumers are not knowledgeable enough to know what is best for their particular situation.

 

That's a nonsensical comparison.

 

It's all related to EULA's excluding any fantastical mention of disallowing driving on "Tuesdays" or "having ice cream on the trunk." Businesses are regarded differently than consumers, and heavy contractual penalties may come into play for companies like Sakura if found breaching the EULA.

As @Shifty Geezer mentioned above,

 

Then it appears that Nvidia has refused to give permission for Sakura Internet to copy its Geforce drivers for the purposes of using them in a data center. Unlike the drivers in the box, future downloads would require agreeing to another EULA, and clicking yes and then not following the agreement has more dire consequences for an enterprise.

I am not sure there's that much of a distinction for function/application. How the software is used, or what it's used for is often a differentiating factor for license types, and authors can refuse to license their work to organizations precisely because they know what applications the work might be used for.
For copyright, it's somewhat moot, since it's a monopoly that gives the right to withhold permission for almost any arbitrary reason.
At least it's frequently in the case in the US. I'm not up on whether this is a case of Nvidia's being headquartered in the US, or if Japan's copyright regime is similarly strong.

Policy makers are not unsure as much as they've been donated or lobbied for a mostly rights-holder friendly position.

That's literally what Nvidia did. They used their role as the provider for a new version of their Geforce drivers to deny Sakura Internet any legal future updates for those cards.
Perhaps Sakura Internet blindly checked yes and didn't save a copy of the old drivers, they cannot reasonably offer their service on old drivers, or maybe in private Nvidia is making an even stronger claim than what is in the news story.
Otherwise, it looks like once it became more of an investment for Sakura Internet to run their service on a Geforce, they took Nvidia's suggestion that it wasn't a worthwhile endeavor.

If you're a consumer, there's a body of (weakening) law that protects you. The laws generally take a dim view on bigger fish doing the same thing for significant amounts of money.

 

Yeah, once I read the article I saw that. So my discussion has been mostly hypothetical.

 

It's really not. We are talking about a prohibition enforced only by licensing. Your example is a separate routine for update reminders while your reasoning was "costumers are ignorant". For the comparison to work the geforce drivers would need to lock out on "datacenter deployment" because "corporate customers are ignorant".

 

HW and firmware still has strict guidelines when it comes to international tech companies creating products for retail and also large business deployments, Nvidia are not the only ones that need to cater to different sales-support channels and can be storage manufacturers to network manufacturers.

So where does that leave those who have games they can no longer use due to GFWL or similar services?
Where does it leave those with games using pre Denuvo protection that no longer works due to XP/Win7 upgraded to Windows 10 and not supported?
Where does it leave Denuvo games that probably will not work with next OS?

Anyway I am surprised no-one sees the issue of bypassing the correct sales and support channels and the right partners who sell datacenter products.
But when you buy a HW product it does not include equal rights to the software/firmware, albeit it would be a grey area about using existing software/firmware provided with the HW in a way that was never intended from a business/service model perspective.
This has started to become more firm when one looks at digital games and applications, and does apply to firmware as well that has similar levels of Manufacturers' ownership/protection.
I need to read the EULA for Geforce but considering it was never intended for large businesses maybe it had some loopholes.

As I mentioned earlier, one is not meant to be able to buy a datacenter amount of Pascal Titans and so questions should be asked how Satora bypassed the mechanism in place to limit what they managed to do.
The downside is that Nvidia will now restrict Titans even more sigh, and especially the Titan V.
You may blame Nvidia but Sakora has a lot to answer for as well; what they attempted to do not only affects Nvidia but also their elite/preferred partners that deal with many of the larger professional requirements.

Your 1st point though is skewed, you cannot pop down to local PCWorld and buy a datacenter worth of Pascal Titans.
Try buying even a cluster for one project without involving Nvidia or a partner officially supported by Nvidia, Nvidia deliberately limited its availability or scope if buying a node and applies to Titan V and Pascal.
In essence Sakora broke Nvidia's rule of only 2 per customer that is clear on Nvidia/Geforce website, sure we could argue this is another grey area but shows Sakora has gone out of its way to bypass what Nvidia has put in place differentiating between Geforce/Quadro/Tesla from business and support-service channel model.
You can get more than 2 but requires specific use such as I mentioned earlier with DL (and specific cases).
But my point and context is more than just the software EULA and also comes back to agreements from a business/purchase/project entity perspective.

Most international tech companies would be unimpressed with what Sakura has done; those companies that sell to retail and also has a sales&support partner channel designed for multi-million pound large scale projects/deployments.
That said Nvidia and their elite partners can make life difficult for Sakura when it comes to warranty/support and going forward.

And like I said AMD is placing differentiators that impacts many buying the Vega FE as a prosumer/small professional enterprise GPU rather than the WX9100 or MI25.
Maybe I am biased because I have spent my life working in such tech companies that have a similar structure to that of Nvidia when it comes to retail/professional/large projects and the relationship between the business (client)/sales&support channel-partners/manufacturer, but I see the need for such differentiation if one is to have a healthy ecosystem and continued product growth and evolution (R&D).
However my context is not small businesses/individuals but ones such as Sakora and larger, which is what kicked off Nvidia's response.
This is not about consumers but business entities such as Sakora pushing the boundaries.
If this went on unchecked consider the impact down the line when such datacenters start to pop up in China/some other countries and cutting cost/channels that support the broad ecosystem worldwide....
Although the other impact is probably even stricter control of Titan sales sigh (and I blame Sakura more for this than Nvidia).

Edit:
Bolded a bit to emphasise the differentiation and intended scope beyond buying 2 per customer.

 

May have been due to Sakura Internet's involvement with Blockchain initiatives.
https://www.sakura.ad.jp/en/press/pdf/20170228block_chain.pdf

 

That makes sense, oh boy if Sakura broke their intended agreement for project/implementation clause I am sure Nvidia and the partner they bought/project from are rather irritated and pretty shortsighted by Sakura due the fallout from it.
Basically Nvidia is killing any support for the project-deployment in latest context, which will probably require specialised agreed support contract going forward; similar to what Microsoft offers to business entities at a higher cost.

 

Since there is this fixation with car analogies, here is a perfect one:

Tesla introduces new ‘Supercharger Fair Use’ policy to focus on long distance travel and deter commercial use.

You can buy the hardware (Car or video card) to your heart's content. If you want to use the service (charging network or drivers) for commercial purposes, you are out of luck.

 

Yes, but they did explicitly state that this only applies to vehicles purchased from that point onward. Previously purchased vehicles can still be used commercially and charge from Superchargers.

OTOH, if CSI's guess is true and Sakura did mislead Nvidia, well, then they just plainly mislead their partners and deserve what happened.

 

All tesla cars can be used commercially, it's the supercharger service (free charging?) that got changed. Tesla could end the program by tomorrow if it wants to, but they can't prevent commercial usage of the cars.

 

OK a more suitable example.
There is plenty of software out there that has a different charge price between retail and commercial/business entities, in fact there is shareware that is free for home owners but as a business you must pay to use said software.
Or what about how OS/application licensing differs between consumer with multiple cores and business entity servers with multiple cores where costs are substantial in comparison?

In all cases they are differentiating from an EULA between retail and the server wide/business entity even though in essense it is the sameor similar product.
But the Titan situation goes beyond EULA because as I said earlier you cannot buy these in bulk generally and doing so usually requires to be in conjunction with Nvidia/Partner for a specific referenced project.
Meaning they have to be purchased in agreement with Nvidia on use; I doubt a manufacturer would discuss said project requirements and then expect to find within 12 months a totally different business entity commercial use like Sakura did.
Go to the Geforce or Nvidia website and it will clearly state trying to order via the website a customer is limited to 2 Titans, you are meant to go through the partner/Nvidia if you need more for your project that would be part of the discussion.

That said there is a further clarification TheRegister reported:

Nvidia has 2 options; either make it much more restrictive to buy Titans with more hoops to go through or stop anyone trying this again by enforcing there is zero support/software limitation for the datacenter unless they pay some kind of penalty level support price (other companies do this for special-custom service support).
I still think they may make it harder now to purchase Titans, but not as bad as it could be if they can get this to work with regards to the Geforce EULA and time will see which side of perspective is right on the legality of this.
And going from the link earlier from Pharma, Sakura approached this initially as a Blockchain solution (which is still allowed) and not what Sakura are also doing now that is sort of misleading if they engaged with official partners.
Meaning they could had purchased under misleading circumstances to begin with.
The solution and definition of Data center is still a bit vague but Nvidia do mention multiple nodes/clusters to many users, then again this only kicked off after what Sakura implemented, not what others have done in the scientific world/universities/etc.

 

That's licensing, again.

 

So some wish to complain about Nvidia EULA change and legality (which is a consumer vs business entity setup differentiation) but ignore those examples/situations that are same context of retail-consumer to business entity use differentiation?
And like I keep saying is also about project purchase-support agreement to get that many Titans so quickly (meaning it was not a consumer context/focused purchase), along with a few other points I mention.
Edit:
Anyway Nvidia is now clarifying the breaking of its use as multiple nodes as a service to many users in context of data center, just repeating as was not touched upon from my previous post.
So the change will not affect those wanting to do alternative node to DGX-1, and able to expand beyond that in specific circumstances.

 

No. The point was raised about products being sold and controlled (nVidia bans use of product for purpose), which wasn't accurate to the situation. This spawned a subject matter on the legality of controlling product uses after sales, for which examples were requested, and the answer is 'there's no way to stop anyone using a product they buy however they want.' But returning to the specifics of this nVidia EULA, the reality is 'nVidia refuses licensing of drivers to certain users' and nVidia are perfectly within their rights to decide who gets to use their freely provided software (excepting the convoluted intricacies of EULAs themselves, such that this may not be enforceable in places like Germany).