News for hackers. Stuff that matters.

A comment on Comparing Google’s TPUv2 against Nvidia’s V100 on ResNet-50. Parent comment

Re:Comparing Google’s TPUv2 against Nvidia’s V100 on ResNet-50 by MrBuddyCasino on Thursday April 26, 2018 @10:20AM (#16932099)
I found it interesting that they are so close together in performance - I mean what are the odds that they end up within 2% of each other?
[ Parent | on HN ]
Re:Comparing Google’s TPUv2 against Nvidia’s V100 on ResNet-50 by jacksmith21006 on Thursday April 26, 2018 @10:48AM (#16932325)
The TPUs are doing almost 2x the images for the same cost.

That is not all that close is it?

[ Parent | on HN ]
Re:Comparing Google’s TPUv2 against Nvidia’s V100 on ResNet-50 by elmarhaussmann on Thursday April 26, 2018 @10:28AM (#16932147)
Yeah, pretty big coincidence. However, this may change with the next TensorFlow versions, which supposedly has further speed improvements for the TPUv2.

Note also, that the ~2% performance difference is only on one model (ResNet-50) and cannot be generalized to all workloads/all of deep learning (at least not without further proof).

[ Parent | on HN ]