NVIDIA Announces the DGX-2 System - 16x Tesla V100 GPUs, 30 TB NVMe Memory for $400K

[Patriot]

One should take a better look at these things before drawing such a conclusion , Nvidia is very specific with their wording. Those 2 PFlops come with the help of Tensor Cores, it should go without saying that this sort of performance is not fully comparable with traditional heterogeneous computing.

Do not live under the false impression that Nvidia has some sort of magic sauce that no one else can conjure up. It's just a lot of dedicated silicon designed for a specific set of tasks. I can guarantee you that in the majority of cases AMD's traditional system is faster and more cost effective while Nvidia's only truly crushes it under very , very specific scenarios. Notice how Jensen talks about this stuff pretty much exclusively within the context of CNNs and that sort of stuff because that's really the only area they've focused on.

Vega 7nm later this year should have tensor cores as well. But yes, they are well over a year behind on the tensor compute. I have been keeping tags on the ROCm dev for the past few years and it has been making major strides forward in making a competitive ecosystem. Clearly still not there but it is hard to close a gap that has been 5+ years in the making. Nvidia and Intel are both not part of GenZ and are basically showing themselves to be taking on the rest of the consortium.

 

[Jism]

That would be a very real problem if nVidia were to ever abandon the gaming market for higher margin segments.

They wont. Their market-share into gaming is still 80% global. You dont abandon that.

 

[lexluthermiester]

"But can it run Crysis"

Easily, and if my math is right, it could run 52 fully separate instances of Crysis @120fps.

 

[mobiuus]

CAN IT RUN CRYSIS? heheh