NVIDIA forbids deployment of GeForce drivers in data centers...

[leadeater]

1 hour ago, Theguywhobea said:

It's not like this line in the ToC says Nvidia will send in a SWAT team if they see you using a Geforce card in your datacenter, it's just because they must be tired of getting support calls from uneducated owners who thought they were saving a buck. This way they can deny these people the support they don't deserve. The price premium on Quadro cards is justified as I stated before. These cards are validated for data center use, the drivers are validated for data center use, and you get support for data center applications, none of those things are free, so they front load the cost onto the card. This is something that is repeated in almost every market. I work for a distributor, our prices in a lot of cases don't beat our competitors, but our customers are more than willing to pay more for their parts because they know we have knowledgle and experienced tech support staff who will provide them support for "free"

Only a very select few cards are specifically validated for data center use yet the Quadro cards that are not, which is nearly all of them, are still being allowed to be used. Driver certification that you are talking about is application specific for things like SolidWorks and the drivers support specific API calls which are mostly considered legacy and taken out of the GeForce drivers, which is the core reason for why specific features in applications perform faster on Quadro drivers and exactly where the performance gain came from for the Pascal based Titan cards after the Vega FE was released, Nvidia just implemented those legacy API support.

 

It would have been better to word this as Nvidia will no longer support GeForce cards being used in server workloads, many people are fine with deploying unsupported configurations and most of the people who have being using GeForce cards in server already assumed they were unsupported. This is more Nvidia explicitly stating so but using some rather brute force wording.

[XenosTech]

3 minutes ago, NumLock21 said:

It's not like the drivers will refuse to install. Your on your own, when it comes to tech support.

Most people would take that risk over paying for products that cost more and offer less performance.

2 minutes ago, Zodiark1593 said:

Well, a workstation isn't necessarily a datacenter. Heck, I use Blender a lot with my own GeForce card, and it seems as happy as can be, minus the 2GB RAM bit.

Well depends on the Os you're running on said workstation. If you're running windows server Os then technically it's a server and nvidia wouldn't be able to tell if you're a small data center or just some bloke with a lot of money to splash on that kinda stuff

[leadeater]

5 minutes ago, NumLock21 said:

It's not like the drivers will refuse to install. Your on your own, when it comes to tech support.

Potentially they could, I've not checked since I don't have any current Nvidia GPUs but miners don't use server operating systems. Even if it's not blocked now don't be surprised when it does get blocked.

[AkatsukiKun]

2 hours ago, Princess Cadence said:

Wasn't this that prosumer card jack of all trades? I know lots of people who invest on one of those above any Quadro solution because it actually was good value there, I personally see this shutting down even further the TITAN X segment...

 

I have always felt that nVidia and pretty much all the others have always charged too much for the "certifications" for real if you do a product that should work doing something why the hell you get to overprice for the certainty it will work? it just feels like milking extra on the work they should do to all.

To answer your question/thoughts. The Original Titan was the prosumer jack of all trades back then with its precision floating point capabilities until the next iterations of the Titan lineup and their x80 Ti counterparts. Which pretty much explains why non consumer grade hardware/software are almost "always" charged a higher premium. Of course unless you are one of those adventurous people who modifies their hardware by flashing the BIOS, and some even adds or removes a transistor or two.

Quote

"What differentiates Quadro from GeForce is that GeForce usually has its dual precision floating point performance severely limited, e.g. to 1/4 or 1/8 of that of the Quadro/Tesla GPUs. This limitation is purely artificial and imposed on solely to differentiate the gamer/enthusiast segment from the professional segment. Lower DP performance makes GeForce boards bad candidates for stuff like scientific or engineering computing and those are markets where money streams from. Also Quadros (arguably) have more display channels and faster RAMDACs which allows them to drive more and higher resolution screens, a sort of setup perceived as professional for CAD/CAM work."

Original quote is from article: https://stackoverflow.com/questions/10532978/difference-between-nvidia-quadro-and-geforce-cards

[leadeater]

2 minutes ago, AkatsukiKun said:

Original quote is from article: https://stackoverflow.com/questions/10532978/difference-between-nvidia-quadro-and-geforce-cards

The modern Quadros also had their DP performance limited too just like GeForce, only cards that weren't used the largest Gx100 based die so basically the Quadro P100 and Teslas.

[System Error Message]

5 hours ago, kerradeph said:

That part bugs me because they are basically saying "Yeah, feel free to keep running your mining farms which are utterly fucking over the prices".

blockchain isnt about mining, its not limited to currency. All blockchain does is create a hash trail of something which is a secure way of logging.

For instance you could use block chain with your phone's GPS history and know when someone has altered the data.

 

I have my own GPGPU cluster and use geforce. Mainly because im not making a profit so i cant afford to use tesla/quadro and on linux, tesla/quadro dont offer any benefit over geforce in my use case.

[RagnarokDel]

5 hours ago, Dabombinable said:

Can they even enforce the EULA and dictate what drivers the owners can and can't use?

No. All they can do is not provide support.

[Nicnac]

There are so many services offering 1080 ti servers for deep learning or rendering! Do they all have to pull these offers now?

[Bit_Guardian]

4 hours ago, sazrocks said:

What even qualifies as a datacenter anyways?

I believe that actually has a strict definition of more than 1 rack/tower in the same building being used by the same firm.

[RagnarokDel]

1 minute ago, Bit_Guardian said:

I believe that actually has a strict definition of more than 1 rack/tower in the same building being used by the same firm.

Well, I have a data center I guess.

[Bit_Guardian]

1 minute ago, RagnarokDel said:

Well, I have a data center I guess.

Do you seriously have a fully loaded 40U tower and then some?

[Bit_Guardian]

4 hours ago, XenosTech said:

Most people would take that risk over paying for products that cost more and offer less performance.

Well depends on the Os you're running on said workstation. If you're running windows server Os then technically it's a server and nvidia wouldn't be able to tell if you're a small data center or just some bloke with a lot of money to splash on that kinda stuff

Furthermore, it's not like said server OS would have internet access.

[straight_stewie]

6 hours ago, Raxzzer said:

I find it disturbing that NVIDIA feels like they have to prohibit usage in data centers just because they fail to offer enough value to their customers through their insanely expensive Quadro and Tesla cards. Are NVIDIA admitting here that their enterprise solutions aren't good enough/offer enough value compared to their "mainstream" alternatives?

It could be more complicated than that. Is there additional wording that you left out?

One of the reasons could be that they haven't tested the software to see if it provides the reliability necessary for use in such a high reliance application. They even made an allowance for data centers whose purpose is crypto-currency mining, so this would lead me to believe that they are just waving liability.

Because, you know, they will support GeForce for bitcoin mining but not for whatever kind of very application specific GPU acceleration people are doing these days. In other words, it is on the users shoulders to make sure it works, not Nvidia's.

[RagnarokDel]

5 minutes ago, Bit_Guardian said:

Do you seriously have a fully loaded 40U tower and then some?

I mean, not a proper server rack but yes, motherboards and a dozen GPUs. Basically every GPUs I've bought since the 500 series + some really good deals on used GPUs

[XenosTech]

6 minutes ago, Bit_Guardian said:

Furthermore, it's not like said server OS would have internet access.

In some instances it will. I mean if I'm using it as a regular pc just to do work, I may have a browser or two on there to do a little web surfing here and there and read a couple emails.

[leadeater]

8 minutes ago, Bit_Guardian said:

Do you seriously have a fully loaded 40U tower and then some?

I do, 42U actually. I even have servers in remote locations to replicate to. Yes this my own personal stuff not work.

[Bit_Guardian]

21 minutes ago, leadeater said:

I do, 42U actually. I even have servers in remote locations to replicate to. Yes this my own personal stuff not work.

Sheesh...

 

And here I am advising my company, that has a ton of Java web apps, to seriously look at Cavium or Qualcomm-based servers to run them when it comes time to retire the current hardware.

[leadeater]

Just now, Bit_Guardian said:

Sheesh...

 

And here I am advising my company, that has a ton of Java web apps, to seriously look at Cavium or Qualcomm-based servers to run them when it comes time to retire the current hardware.

I have a server hardware addiction, I need help .

[XenosTech]

Just now, leadeater said:

I have a server hardware addiction, I need help .

I can help you if you send some of it my way 

[leadeater]

4 minutes ago, XenosTech said:

I can help you if you send some of it my way 

[XenosTech]

1 minute ago, leadeater said:

You'll still have it. It'll just be extra remote

[Bit_Guardian]

2 minutes ago, XenosTech said:

You'll still have it. It'll just be extra remote

He needs guaranteed ssh and SFTP access.

[XenosTech]

1 minute ago, Bit_Guardian said:

He needs guaranteed ssh and SFTP access.

shhh don't give him ideas lol

[FPSwithaWacomTablet]

I want VNC access and/or Xauthority

[KhandakeF]

Wait I thought this was good because then NVidia isn't responsible if something happens when using drivers that aren't meant to be used for that workload?