Yes but no. Asics are very inflexible and have a hard time dealing with new functions. GPU’s are general purpose and therefore new types of convolutions are much easier to implement and have the driver support for it.
There are frequent claims of random companies making NPU’s TPU’s etc but they always have software issues for training new models with the latest methods. If not Nvidia wouldn’t be having such massive market share anymore.
If deepseek bought those chips only for inference, they wouldn’t be able to keep up in model development and wouldn’t need inference anymore.
As Meituan uses these chips for training, Deepseek will also use these chips for training. You claiming the chips are somehow defective won’t make Deepseek not use them.
Different chips can be used for training but as I said it’s driver hell and much easier to just use Nvidia. As time goes on different vendors are catching up to Nvidia but I still think they’re having a hard time with it. Chinese companies aren’t still buying Nvidia because they love paying ludicrous rates for a tiny bit of extra VRAM on a GPU chip that’s basically the same as the gaming one.
Yes but no. Asics are very inflexible and have a hard time dealing with new functions. GPU’s are general purpose and therefore new types of convolutions are much easier to implement and have the driver support for it.
There are frequent claims of random companies making NPU’s TPU’s etc but they always have software issues for training new models with the latest methods. If not Nvidia wouldn’t be having such massive market share anymore.
If deepseek bought those chips only for inference, they wouldn’t be able to keep up in model development and wouldn’t need inference anymore.
As Meituan uses these chips for training, Deepseek will also use these chips for training. You claiming the chips are somehow defective won’t make Deepseek not use them.
Different chips can be used for training but as I said it’s driver hell and much easier to just use Nvidia. As time goes on different vendors are catching up to Nvidia but I still think they’re having a hard time with it. Chinese companies aren’t still buying Nvidia because they love paying ludicrous rates for a tiny bit of extra VRAM on a GPU chip that’s basically the same as the gaming one.