Nvidia EULA limits GeForce Data Center Usage *fork*

Is that in any way legally enforceable? You can't control what someone does with a product they buy. They could try and limit the scope of their drivers through a software license, but that'd be like MS trying to control Office and you're not allowed to use Word to write books.
 
New Nvidia EULA allegedly bans use of Geforce cards in datacenters:
If this isn't the arrogance of a corp with too much market power, I don't know what would be...
In a way it reminds me of those companies with free software licenses granting you the privilege of using it free if certain conditions are met:
such as only being installed on one physical/virtual computer,
non-commercial usage,
restricted to companies with less than xx employees, etc...
 
that'd be like MS trying to control Office and you're not allowed to use Word to write books.
Like MS banning using Home version of office for commercial purposes (like writing books); demanding you pay for full version...

Just spitballing here, but if you ask NV themselves, I suspect they'd say it's not at all to make themselves more money, but to prevent companies from vacuuming the market of gaming cards so regular people can't find any to buy... ;)
 
One wonders what the (accepted or otherwise) definition of "datacenter" is which they're using...
 
It will be interesting to see what the pricing on the V100 Quadro series will be though should be similar to the P100 Quadro.
 
Point of Titan Volta is to seed scientists and developers so they can have volta on desktop and develop locally. Titan is not meant for datacenters.

It could be argued the Titan X was designed for small-medium size labs/some datacenters as it was the 1st promoted card by Nvidia for accelerated Int8 Inferencing to support the P100.
I appreciate a bit later there was the Tesla GV102, but before then the Titan X was promoted in this context.

On the Nvidia statement.
I guess though as well some of this comes down to spec/semantics on the definition/size for data center/cluster-node size, and possibly this is more directed at 'assisting' the Elite type service providors who build quite a lot of the datacenters involving Nvidia contracts and would be under pressure by some clients, along with what has been mentioned by others.
 
Last edited:
If this isn't the arrogance of a corp with too much market power, I don't know what would be...
NVIDIA isn't clamping down on individual customers or university students with this regulation. They are targeting companies that buy regular Geforce cards and then lease them in a virtual environment to other customers, effectively selling time on their servers. This is what the Japanese company "Sakura Internet Inc" is essentially doing. And NVIDIA isn't happy about it, so they are they requiring them to use Tesla/Quadro grade cards instead of Geforce.
 
Except how is that legal? That's like selling someone a minibus and requiring them to only use it for old people, or selling someone a TV and restricting its use to factual programming. If nVidia were renting the cards, they could limit usage, but a sale is an outright thing. The other company takes ownership and can do what they want, whether gaming, bitcoin mining, or turning the GPUs into fashionable broaches, or selling them on to someone else to do whatever they want with them.

If they are selling through a proprietary contract for bulk-purchasing, the EULA is immaterial and it's the contract that needs to specify use limits.
 
Except how is that legal? That's like selling someone a minibus and requiring them to only use it for old people, or selling someone a TV and restricting its use to factual programming. If nVidia were renting the cards, they could limit usage, but a sale is an outright thing. The other company takes ownership and can do what they want, whether gaming, bitcoin mining, or turning the GPUs into fashionable broaches, or selling them on to someone else to do whatever they want with them.

If they are selling through a proprietary contract for bulk-purchasing, the EULA is immaterial and it's the contract that needs to specify use limits.
I'm guessing in the same way warranty is often void for products for commercial use or if the product is "tampered" with.
 
Warranty will either be statutory, in which case the company can't control it, or extended by the company, in which case all they can do is refuse to service defective parts. That is, warranty is a post-sales service that the company chooses to supply and is therefore free to chose not to. eg. Apple sells a phone saying you have to use a certain service provider. You are free to jailbreak the phone and use any provider you want - all Apple can do is not service your phone if it goes wrong. They cannot legally bind you to any use whatsoever for the hardware you own. That's a fundamental principle of ownership, and the only legal way to control what people do with your hardware is to lease it rather than sell it, or supply through a negotiated contract.
 
And NVIDIA isn't happy about it
Well yeah, duh. That's pretty much what I said in my post. :p

Anyhow, it's not any of NV's business how their customers choose to use their own cards; they don't get to decide that. Car analogy time, because car analogies are the best analogies, but if Ford sells you a van and then dictate you can't use it for your plumbing business; you have to buy their "quadro" line of vans which are 5x more expensive despite being the exact same van as the one you purchased, would you think that's OK? Especially if Ford owned basically the entire professional market for vehicles?
 
Cars have warranty for miles or years whichever fills first. If gpus were cars datacenter use case would be out of warranty fast due to hours.
 
GPUs don't have odometers though. There's no way of telling how much they've been running, and if you sell a GPU with the implied suggestion it shouldn't be used or else go out of warranty, well, then you're out to scam your buyers and shouldn't be allowed to be in business of selling anything.
 
Are those car warranties statutory or corporate? On a GPU, warranty is statutory (certainly here in UK - I've never seen am extended warranty for a GPU advertised!) so it'll be 1 or 2 years whatever you do with it. HDDs with extended 5 year warranties have caveats. Again, all nVidia could do is have the warranty voided, meaning the data-centres couldn't get a repair or replacement in case of failure. The savings will easily offset that.
 
Back
Top