Page 1 of 1

Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 4:58 am
by Welch
So this just happened..

https://wccftech.com/nvidia-geforce-eul ... lowed/amp/

Thoughts? I realize they want to recoup their R&D but this seems like the **** possible way to do it. Perhaps an additional licensing fee would have been a better route. Not even surprised with Nvidia anymore.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 5:09 am
by Ryu Connor
Good for gamers. Should help make sure that the average consumer isn't competing with datacenters to get a GPU.

Good for shareholders. Should help the bottom line.

Good for the company. Volta V100 was supposedly a $3b chip. I don't see those kind of costs getting lower in the future.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 5:48 am
by the
Not really enforceable. More realistically it'll be an excuse for nVidia to provide support for data centers that have tried to snatch up Titan Vs for server usage.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 6:00 am
by NovusBogus
Agreed that legal thuggery is a terrible approach to market segmentation, but it should be noted that Quadro/Tesla basically *is* the 'business license fee' that you suggest. The blockchain exception is interesting, but I suppose they realize the hilariously low probability of actually making something like this stick against a Chinese mining cluster.

In any case, it should scare off the corporate drones but actually trying to enforce it beyond voiding warranties would lead to some quality entertainment given the nature of those who run very large GPU clusters.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 7:28 am
by Chrispy_
NovusBogus wrote:
the hilariously low probability of actually making something like this stick


And there lies the rub. Even if they do try to enforce it, the Quadros and Teslas have much of the shared driver and architecture as Geforces, so all a company with a technical issue needs to do is prove that the problem exists with ONE Quadro/Tesla.

I keep a Quadro and a Firepro on tap, just so that I can confirm any issues I'm seeing with Radeons and Geforces in the office fleet aren't driver-related. They never are, but on the very rare occasions I've had to invoke pro-level support from either AMD or Nvidia I have a valid serial number and hardware to liaise with them on the issue. I've even loaned another firm my WX9100 at the moment just so that they can prove to AMD that the issue happens on a Radeon Pro and that the issue is something that AMD and Autodesk need to resolve across their entire driver stack.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Tue Jan 02, 2018 7:33 am
by Krogoth
They just don't want their customer-tier Volta SKUs cannibalizing their Telsa/Quadro sales. It is more evidence that Volta microarchitecture family is much more focused around general compute performance this time around.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 1:00 am
by Kougar
So does this apply to the Titan V's or not? I recall reading the Titan V still uses the Geforce driver line.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 1:11 am
by DancinJack
Kougar wrote:
So does this apply to the Titan V's or not? I recall reading the Titan V still uses the Geforce driver line.


I believe so. While it doesn't bear the Geforce name, it still uses the Geforce suite of driver rather than the Quadro stuff.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 1:33 am
by Duct Tape Dude
This feels like first-amendment violations of a computer processor. There's no way this will be enforceable.

To skirt around this limitation, what's stopping someone from bootstrapping their existing code with a single call to git or some other blockchain tech?
Why not instead offer premium support/licensing options for GeForces in the datacenter?

Ugh. Stupid nvidia.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 3:48 am
by Topinio
Duct Tape Dude wrote:
Why not instead offer premium support/licensing options for GeForces in the datacenter?

The Tesla driver is validated sufficiently to enable an offer of premium support, and this is baked into the TCO.

Or, another way to look at it is that the very existence of the Tesla range of boards at its higher price point is exactly how NVIDIA offers "premium support/licensing options for [its product] in the datacenter".

Use of GeForce products in the datacentre circumvents NVIDIA's market segmentation to reduce its margins and undercut its own profitability. Refusing to support this is about all it can do in response.

There's also only a low chance that NVIDIA doesn't know who has (or is considering/planning) thousands of its high-end GeForces in their datacentres, therefore this is also viewable as a possible shot across the bows or a PFO :wink:

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 4:20 am
by just brew it!
Duct Tape Dude wrote:
This feels like first-amendment violations of a computer processor. There's no way this will be enforceable.

It's about as enforceable as any other arbitrary EULA limitation that isn't backed up with some form of DRM. IOW, yeah, not really. It'll scare some potential users off, while others will ignore the EULA restriction and use them anyway.

Duct Tape Dude wrote:
To skirt around this limitation, what's stopping someone from bootstrapping their existing code with a single call to git or some other blockchain tech?

Not even sure what you're suggesting here. I read the words, yet they seem to have no meaning.

Duct Tape Dude wrote:
Why not instead offer premium support/licensing options for GeForces in the datacenter?

As NovusBogus already pointed out, that's essentially what Quadro/Tesla are.

Duct Tape Dude wrote:
Ugh. Stupid nvidia.

From a business perspective it may actually be a pretty smart move.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 11:29 am
by Waco
While I don't really like legal enforcement like this, it makes a lot of sense. It'll keep Geforce prices lower for gamers and force datacenter deployments to use the intended cards and software stack.

I have no doubts that Nvidia would go after a large customer circumventing the EULA with legal force.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 7:03 pm
by Welch
I'm not against them having some form of market segmentation, that is just business. I'm just against them being heavy handed with that segmentation so far as to threaten legal action.

Simply don't support GeForce cards in a data center environment (this would need a specific definition). JBI, you're dead on... Without DRM of some form then you are getting back into Microsoft's old licensing ways for CALS... An honor system. I can't even begin to tell you the number of accounts I took over where previous IT guys didn't give a damn about CAL licenses because the OS didn't enforce anything.

Of course this would probably be a waste of cash for Nvidia to try and bake a hardware form of DRM into their chips, not to mention R&D and staying on top of it.. possible issues it could create for a legitimate user base too. It's also not lost on me that a software solution would be just as pricey, be an on going effort and most likely be quickly circumvented by people who choose to use customer drivers or the like.

So yeah, I can see Nvidia being pissed that there is this market segmentation that they have little to no power over controlling. At the same time, is there a less forceful way for them to accomplish their goal that isn't an artificial steering of the market? It just seems shady to push an artificial "Fix" that only serves you the company and in turn screws over a large customer base where you could have came up with an actual solution.

I could be misreading this situation all together, I'm not 100% on the up-and-up for this particular GPU computing market stuff.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 7:12 pm
by bthylafh
Duct Tape Dude wrote:
This feels like first-amendment violations of a computer processor.


:roll: :roll: :roll:

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 8:11 pm
by just brew it!
bthylafh wrote:
Duct Tape Dude wrote:
This feels like first-amendment violations of a computer processor.

:roll: :roll: :roll:

"Compute cycles just want to be free!"

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 8:49 pm
by bthylafh

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Wed Jan 03, 2018 9:58 pm
by Krogoth
Nvidia never need to do this before because their previous customer-tier micro-architecture SKUs had general compete functions and performance gimped at the hardware level. They are easing this up with their customer-tier Volta SKUs in light of the crypto-currency craze. Notice how the EULA exempts crypto-currency mining? Not a coincidence.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Thu Jan 04, 2018 1:49 pm
by Kougar
I don't have a problem with NVIDIA wishing to enforce some measure of product segmentation. Processor and hard drive vendors already do it anyway.

Call me selfish, but as a consumer I'm for this simply because it helps protect me from price inflation, supply shortages, and whatever else massive companies might cause as a result from massive consumer graphics cards orders/deployments. Particularly if such deployments began making use of the consumer warranty process for their bulk orders.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Thu Jan 04, 2018 1:54 pm
by Topinio
Kougar wrote:
Call me selfish, but as a consumer I'm for this simply because it helps protect me from price inflation, supply shortages, and whatever else massive companies might cause as a result from massive consumer graphics cards orders/deployments.

That's unsupported and contrary to evidence. The only thing that protects you from NVIDIA increasing prices on its consumer parts is competition from AMD, and that isn't happening as people are not buying AMD's products.

You might argue that a reason for that is people with data centres (in China) buying up AMD's consumer GPU products, ofc...

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Thu Jan 04, 2018 2:34 pm
by Kougar
Topinio wrote:
That's unsupported and contrary to evidence. The only thing that protects you from NVIDIA increasing prices on its consumer parts is competition from AMD, and that isn't happening as people are not buying AMD's products.


It's simple supply and demand. The repeated drastic price inflation of the entire new (& secondary used) GPU markets caused from multiple, different cryptocurrency bubbles shows the potential still exists for there to be an effect on GPU markets.

Nvidia can price it's products at whatever it wants. But if for whatever reason NVIDIA did away with product segmentation and data centers and server farms began buying up consumer hardware instead of Tesla/Quadro parts then I could only imagine NVIDIA raising consumer prices as a result. So once again product segmentation provides some protection to consumers.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Thu Jan 04, 2018 4:06 pm
by Duct Tape Dude
just brew it! wrote:
Duct Tape Dude wrote:
This feels like first-amendment violations of a computer processor. There's no way this will be enforceable.

It's about as enforceable as any other arbitrary EULA limitation that isn't backed up with some form of DRM. IOW, yeah, not really. It'll scare some potential users off, while others will ignore the EULA restriction and use them anyway.
Yeah, fair point. Maybe that's enough of a reason to do it.

just brew it! wrote:
Duct Tape Dude wrote:
To skirt around this limitation, what's stopping someone from bootstrapping their existing code with a single call to git or some other blockchain tech?

Not even sure what you're suggesting here. I read the words, yet they seem to have no meaning.
If nvidia allows blockchain processing, it seems easy to sneak in a use of a blockchain (like say, git or an iteration of any hashing algorithm) into any code and claim it's not against the EULA. It's trivial to play the "WELL TECKNIKULLY *pushes up glasses*" card.

just brew it! wrote:
Duct Tape Dude wrote:
Why not instead offer premium support/licensing options for GeForces in the datacenter?

As NovusBogus already pointed out, that's essentially what Quadro/Tesla are.
Duct Tape Dude wrote:
Ugh. Stupid nvidia.

From a business perspective it may actually be a pretty smart move.
Right, I get what Quadro/Tesla are for, and it'd make sense to just block GeForce altogether from the datacenter, or you know, refuse to support GeForce in compute applications altogether. Instead they want to block specific workloads on GeForce from the datacenter. It'd seem more apt if they just left out the blockchain part.

I'm quite happy to avoid fighting datacenters for my next GPU. This seems a silly way to do it.

Re: Nvidia forbids GeForce use in Data Centers...

Posted: Thu Jan 04, 2018 4:09 pm
by chuckula
This is a semi-related thread about the use of consumer GeForce cards in server cases. It looks like Nvidia is moving more strongly to prevent this from happening beyond the phyiscal setup of its consumer cards.