3

I'm currently looking for the correct number of 2018 budget for the AWS instances in preparation of more deep neural net training, but now a bit confused because of this chart below:

AWS EC2 Cost

p2.16xlarge costs 3 times more than g3.16xlarge with the same number of processors but slightly less memory. Is this diff just because of the memory difference? or is there any other key difference between p2 and g3 instances? If I'm missing anything here, please share your wisdom!

Thanks,

YJ Kim
  • 39
  • 1
  • 3
  • Did you research the question and look up the complete P2 and G3 instance descriptions on the AWS web site? Do they feature the same GPUs, GPU cores and GPU memory? – mjul Oct 30 '17 at 17:28
  • I actually couldn't locate the detailed description on AWS website. Can you shoot me a link if you have it? – YJ Kim Oct 30 '17 at 18:56

1 Answers1

3
  • p2.16xlarge has 16 Tesla K80 GPUs
  • g3.16xlarge has 4 Tesla M60 GPUs
  • p3.16xlarge has 8 Tesla V100 GPUs.

I believe the number is a reference to the vCPUs (they all have 64). Comparing the 1 GPU instance of each type, p3 is the most expensive, then g3, then p2, as expected.

Stephen Rauch
  • 1,783
  • 11
  • 21
  • 34
Micah Price
  • 31
  • 3