3 ms·

They compete more than ever. Google, Amazon, FB, Apple are all developing AI, CPU and GPU chips that compete directly with Nvidia using ARM You know why? Becau

They compete more than ever. Google, Amazon, FB, Apple are all developing AI, CPU and GPU chips that compete directly with Nvidia using ARM You know why? Because working with Nvidia has become such a pain in the ass. That it's better just to do it yourself. They charge you 10x for a gpu that's a little better than their consumer GPU but won't sell you the consumer GPU because they want you to pay 10x. Now they will definitely do the same with ARM licensing.

jjeaff 5y ago
How do they stop companies from buying consumer GPUs? I have used some minor firmware tweaks to unlock full commercial capabilities out of Nvidia GPUs in the past. Perhaps that is against the tos and a large company couldn't get away with it like an individual can?
- cmeacham98 5y ago
  If you're Google/Apple/etc, you don't want to be relying on some unofficial firmware tweak where the next driver update could completely brick your business and force you to never update the driver again.
- mroche 5y ago
  > Perhaps that is against the tos and a large company couldn't get away with it like an individual can? A while back they made a change to the terms of the GeForce driver, barring it from deployment in datacenters except for blockchain processing. https://www.datacenterdynamics.com/en/news/nvidia-updates-geforce-eula-to-prohibit-data-center-use/ https://www.datacenterdynamics.com/en/news/nvidia-updates-ge... https://www.nvidia.com/en-us/drivers/geforce-license/ https://www.nvidia.com/en-us/drivers/geforce-license/
  - Gene_Parmesan 5y ago
    Wasn't the ostensible justification here making the cards more available for the enthusiast gaming market they were (to some extent) intended for? That's an honest question, I can't fully recall.
    - impulser_ 5y ago
      That might be what they labeled it as, but it really was to prevent companies from buying the cheaper GPU for AI training. They need their datacenter GPU numbers to look good at the end of the Q, they can't do that when all they are selling are 3090 that are perfectly fine to train the majority of AI models.
- neilparikh 5y ago
  Nvidia doesn't allow you to use consumer CPUs in servers/data centers.

jjeaff 5y ago

How do they stop companies from buying consumer GPUs? I have used some minor firmware tweaks to unlock full commercial capabilities out of Nvidia GPUs in the past. Perhaps that is against the tos and a large company couldn't get away with it like an individual can?

cmeacham98 5y ago
If you're Google/Apple/etc, you don't want to be relying on some unofficial firmware tweak where the next driver update could completely brick your business and force you to never update the driver again.
mroche 5y ago
> Perhaps that is against the tos and a large company couldn't get away with it like an individual can? A while back they made a change to the terms of the GeForce driver, barring it from deployment in datacenters except for blockchain processing. https://www.datacenterdynamics.com/en/news/nvidia-updates-geforce-eula-to-prohibit-data-center-use/ https://www.datacenterdynamics.com/en/news/nvidia-updates-ge... https://www.nvidia.com/en-us/drivers/geforce-license/ https://www.nvidia.com/en-us/drivers/geforce-license/
- Gene_Parmesan 5y ago
  Wasn't the ostensible justification here making the cards more available for the enthusiast gaming market they were (to some extent) intended for? That's an honest question, I can't fully recall.
  - impulser_ 5y ago
    That might be what they labeled it as, but it really was to prevent companies from buying the cheaper GPU for AI training. They need their datacenter GPU numbers to look good at the end of the Q, they can't do that when all they are selling are 3090 that are perfectly fine to train the majority of AI models.
neilparikh 5y ago
Nvidia doesn't allow you to use consumer CPUs in servers/data centers.