GCP A100 Shortage

avivsham · 09-20-2023 06:50 AM

Hi All,

I am unable to start the instance with my A100 GPU card, even after attempting to launch it in a different regions (US, EU and Asia). Trying to find a solution I encountered this question.

We would appreciate if you could help us with the following questions:

Is there a convenient way to search for regions with free A100 GPUs? Just manually going over all regions trying to find available card is really frustrating and inconvenient.
If I am lucky enough to find the A100 GPU card, is it possible to reserve it for XXX hours? These VMs can be quite expensive and we would prefer not to keep them running unnecessarily, just because we are worried about not being able to find a suitable card when relaunching.

Thanks in advance.

marros

I've been having great difficulty getting support from Google for weeks. The machines in Azure have proven to be much better because the resources to use the A100 seem impossible to execute. Apart from the V100 too. All regions are always unavailable and support never seems to resolve it. I suggest better support as it is not worth depending on the resources of good machines to use with AI in GCP.

The regions reported on the website do not work for the A100 and V100:
https://cloud.google.com/compute/docs/gpus/gpu-regions-zones

legel

+1.

For the past week I've been trying to launch a A100 on GC but failing in all regions...

This should be a red alert for the Google Cloud team, as it's exactly the type of problem that will drive customers elsewhere...