Google Cloud Platform cuts the price of GPUs by up to 36 percent

techcrunch.com

121 points·spacemanspiffy··9 comments

A few months ago I did an analysis of the cost various Google Cloud Platform system configurations for deep learning, and found that due to the price savings offered by premptable VMs, it may actually be more cost-effective to use CPUs instead of GPUs. (HN thread: https://news.ycombinator.com/item?id=14727673)

A 36% price cut, however, might be enough to tip the scales back in favor of GPUs. I’ll rerun the experiment with updated VMs soon and see what happens.

really curious to see your updated analysis :)

Here too, please.

In some batch use cases, time is not an issue (i.e, I could spend 10x as long without anyone noticing) as much as cost.

But the GPUs are faster overall right? So you get a faster turn around time.

Correct, which matters in some use cases like production environments, but not most ad hoc experiments.

> using the somewhat older K80 GPUs will now cost $0.45 per hour while using the newer and more powerful P100 machines will cost $1.46 per minute

This is hopefully supposed to say $1.46 per hour.

Yep, beat me to it. I was shocked when I read that at first.

Here's Google's official pricing.

https://cloud.google.com/compute/pricing

$1.46 USD per hour per GPU

We recently started to build out GPU infrastructure onsite. Our GPU needs are low bandwidth and we keep a few in the cloud in case we lose our onsite. So far so good. We have predictable consistent load, so our use case is different than many DNN users.

Onsite == onoffice ?

Reply on news.ycombinator.com