TOP KEK
NvidiaCucks won again
>Pascal 10x faster than Maxwell
>1,0TBs bandwidth
>Volta on some Alien level
>mfw you actually bought 980Ti or Titan X
>Release Q2 2016
https://www.youtube.com/watch?v=xQhb3C2hQoE
>>52632841
>deep learning
god will he ever let it go?
>>52632869
very very deep learning right here
>>52632841
They will cost what? The price of an Oculus rift?
>52632841
>Pascal 10x faster than Maxwell
In applications that utilize FP64, which was dropped starting from Maxwell. This is on the assumption that the code requires high precision only 1/10th of the time, as it does in machine learning.
>1,0TBs bandwidth
Meaningless without benchmarks to back those numbers, HBM1 flopped and didn't perform better than regular GDDR5.
>Volta on some Alien level
Alien as in no one has ever seen it and it's some near-mythical thing or Ayy Lmao?
>mfw you actually bought 980Ti or Titan X
Released almost a year ago now. Amazing how people who bought those GPUs back then still have the highest-end hardware today.
>Release Q2 2016
Some cut-down versions running on GDDR5X, HBM2 and the full chips won't be ready till much later. There's no way nvidia will release a Titan or Ti anywhere near release. Expect another half-year wait minimum for those.