Announcement

Collapse
No announcement yet.

NVIDIA May Be Trying To Prevent GeForce GPUs From Being Used In Data Centers

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by kpedersen View Post
    Easy DRM solution.
    The driver keeps contacting NVIDIA's servers. If too many connections come from one address, it means it is from a "data center". Then the server can tell the driver to disable itself!
    With this you just caused the GPUs of many many many innocent consumers to disable themselves.
    Many ISPs bunch together a large amount of people unless they require a public IP (and pay for it) and place them all behind the same IP through a beefy router doing NAT and port-forwardings.

    So the traffic will come from the same public IP, even if it's different people living all over a city or region.

    I think they should do it. Consumers are quite happy to put up with Steam and I personally believe this will give them the best consumer consumer experience.

    Of course, this DRM scheme won't be easy to enforce with their Linux or FreeBSD drivers
    No they can easily have their linux/BSD drivers phone home too if they wanted. It's not hard. It's just pointless.

    Comment


    • #12
      Originally posted by GunpowaderGuy View Post
      this is a blatant violation of the rights the consumers have of using their products in the way they see fit * ( with exceptions , but not like this one )
      This is here to discourage companies residing in the US from doing that.

      NVIDIA has no way of enforcing this on consumers, nor on companies outside USA.

      Comment


      • #13
        Originally posted by sireangelus View Post
        Nvidia can be translated from the italian " invida", meaning "envy". That always explained a lot.
        They are also green, as in the "green with envy" idiom (also used in italian).

        They can't get more obvious than this.

        Comment


        • #14
          I don't see this as too big of a deal for 2 reasons: 1) Organizations that have need for GPU powered data centers could potentially just hire someone to create a custom driver based on the open source frame work currently available and 2) consumer video cards are probably poor choices for data centers anyway because of the limited frame buffers they have, big data requires big ram which is why both NVIDIA and AMD have so much ram on their pro caliber cards.

          Come to think of it, that may be the practical reason NVIDIA is making this change, namely support calls, which can be quite expensive for a company. If organizations are using consumer grade video cards for large data centers, it's likely they are experiencing issues due to the smaller frame buffers of these cards, resulting in crashes. Also if I remember correctly, the high end Quadro feature ECC VRAM, which a must have in environments like data centers.

          I just don't think this is motivated by greed.

          Comment


          • #15
            Originally posted by starshipeleven View Post
            Many ISPs bunch together a large amount of people unless they require a public IP (and pay for it) and place them all behind the same IP through a beefy router doing NAT and port-forwardings.
            Oh right! In the UK I have never heard of that. Nice, then no problem. on the consumer NVIDIA GeForce box it will say "Internet access with a public IP required". "Or, you must live in the following city / region to use this product"

            Originally posted by starshipeleven View Post
            No they can easily have their linux/BSD drivers phone home too if they wanted. It's not hard. It's just pointless.
            I suppose, but because we relink these things, we could link against a dummy socket() call. Yes, I suppose we could do this with MS Winsock too.

            Pointless? Yes DRM is pointless but since so many companies are doing it, they obviously don't think so. I am still disappointed that so many consumers put up with Steam. I was personally devastated when this kind of thing started to become the norm. Worse still, these days I have to slap down so many colleagues when they try to bring this crap into our workspace or development pipelines. A great example is consumer sh*t like Unity.

            Comment


            • #16
              Originally posted by Spooktra View Post
              I just don't think this is motivated by greed.
              Sure, it is just to ensure that we get the best possible consumer experience. Or to make sure that companies get the best possible solution for their needs.

              Almost like Sony removing the PS3 "Other OS" feature was to "improve our choice as consumers".

              Comment


              • #17
                The typical monopoly play. Companies who make their customers suffer, will get their revenge. sooner or later.

                Comment


                • #18
                  Originally posted by kpedersen View Post
                  Oh right! In the UK I have never heard of that.
                  The UK isn't different. There are not enough public IPs to give to people, which is the reason most people is kept behind some kind of NAT by the ISP, so they can keep IPs free for websites. https://en.wikipedia.org/wiki/IPv4_address_exhaustion

                  If you want something like that to work, you need to force people to have ipv6 first. If that's the case, I fully approve this move. Would be one of the few times NVIDIA did something useful for the community.

                  I suppose, but because we relink these things, we could link against a dummy socket() call. Yes, I suppose we could do this with MS Winsock too.
                  Good luck doing that with a closed source driver shipped pre-compiled, the same as the Windows one.

                  Besides, if all it takes is blocking the connection then even Windows Firewall or linux's onboard application firewall (controlled by gufw for example) can do that fine.

                  Pointless?
                  Yes, pointless. The whole idea of using IPs is silly because networking does not allow what you need, but that's tangential. The whole idea of using DRM here is pointless or even counterproductive.

                  If they were really serious about this they can just make a ME/PSP module do that for them (gathering the number of devices using a NVIDIA GPU in the LOCAL network, checking that there are no high-speed fiber or copper interconnects as used in computing nodes, monitoring the load level of the NVIDIA GPUs in the local machine and so on and so forth) and just report back to a central server regardless of IP or whatever.
                  And of course it will disable PCIe slots if it detects a situation where it might be a computing node AND it has no way of connecting to NVIDIA to report this.

                  Since this must be available everywhere it would be integrated by default in all boards, just like the other Intel crap.

                  And then NVIDIA is fucked because everyone gives them the finger and switches to AMD for the sake of not getting into this madness. Which makes the DRM pointless, or even counterproductive. A less-powerful DRM does less damage, but still does damage.

                  Any strong DRM is applied only by dim-witted individuals, (aka media industry), NVIDIA isn't stupid, with this license they leave the door open to throwing lawsuits to large offenders (so any large company will be discouraged from going cheap), small offenders are not an issue for them (as it's unlikely they would have found the $$$ to buy the same amount of Quadro cards instead).

                  Comment


                  • #19
                    Originally posted by Spooktra View Post
                    I don't see this as too big of a deal for 2 reasons: 1) Organizations that have need for GPU powered data centers could potentially just hire someone to create a custom driver based on the open source frame work currently available
                    I doubt this has any kind of cost-effectiveness.

                    Come to think of it, that may be the practical reason NVIDIA is making this change, namely support calls, which can be quite expensive for a company. If organizations are using consumer grade video cards for large data centers, it's likely they are experiencing issues due to the smaller frame buffers of these cards, resulting in crashes.
                    I somehow doubt that the GeForce line of cards has any kind of company support for datacenter usage, if at all, by itself. I always thought company support contracts on these kinds of things were arranged on a case-by-case basis.

                    I just don't think this is motivated by greed.
                    Choices of a for-profit company are usually motivated by greed, one way or another. Even avoiding costs associated with support calls is eventually greed.

                    Comment


                    • #20
                      I have a simple solution for this: Sell your Nvidia on Ebay and buy a AMD Card instead!

                      it is really that simple. Nvidia can only do this to people who are "slaves"
                      Phantom circuit Sequence Reducer Dyslexia

                      Comment

                      Working...
                      X