@mikiobraun: "I did a couple simple benchmarks for tensorflow on a M1 Max...."

Back to 2021

Mikio Braun

@mikiobraun

I did a couple simple benchmarks for tensorflow on a M1 Max. TL;DR: M1 GPU is only 60% slower than a GTX1660S, but the CPU is 4x faster than the AMD. I took this simple benchmark that trains a ResNet5 on the cifar data set. medium.com/analytics-vidh…

medium.com · Access denied

Dec 21, 2021 · 18:16

2 Retweets

8 Likes

Full thread (6 tweets) →

I did a couple simple benchmarks for tensorflow on a M1 Max. TL;DR: M1 GPU is only 60% slower than a...

Dec 21, 2021 · 18:16

For comparison, I used a Linux box with an 8 core AMD CPU and a seasoned GTX 1660S. One epoch took 6...

Dec 21, 2021 · 18:16

I followed these docs to install tensorflow with M1 support from Apple https://t.co/mF5FWV552P Resul...

Dec 21, 2021 · 18:16

So in summary, the M1 Max GPU is about 57% slower than the GTX, but the CPU is 4x faster than the AM...

Dec 21, 2021 · 18:16

I also ran it on Google‘s colab. I got a K80 and one epoch took 162s. Pretty nice.

Dec 21, 2021 · 18:16

View all 6 tweets in thread