GPU choice is currently low

[green birch]

Hello, can you tell me how to choose a GPU for stable operation?
Serveless is often throttled these days with the message "Supply of your primary GPU choice is currently low. Update worker configuration."
And the configuration screen is all red. What should I do? Change the location? I'm using a 4090, which is sufficient for my tasks, but I don't know which way to go to keep 3-5 workers running under heavy loads.

[arctic void]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[fluid sand]

If you set up Active workers, they will always be available, though it costs money.

Also, there is a trick where if you set the available DCs to "Select all," you can use workers whose locations are unknown and cannot be explicitly specified.

If you do not want to use Active workers, you will have to make it work on another GPU or simply wait.
One thing I can say is that RTX 4090s and RTX 5090s are always popular GPUs, so it is a well-known, common issue that they are often out of stock.

[copper pendant]

how long do you thik the wait will be? currently you all have almost no available gpus for serverless solutions and the few you do have are low availability

[fluid sand]

Based on my experience, these are issues that occur about once a month.
They usually resolve within a day, and the service becomes available as usual by the next day or the day after.

[copper pendant]

Perfect youare amazing. In that case I will wait

[copper pendant]

Any idea why its once a month?

[fluid sand]

What the staff said previously was that users who use an excessive amount are monopolizing the resources.
My guess is that it might be due to mining users, or perhaps someone leaked a Runpod API key and it is being used for mining as well (though I am not sure if this is true).

[sharp ridge]

probably its a user that uses tons of resources for batch jobs

runpod's will try to better prepare for that

[green birch]

Hello. Am I correct in understanding that the lower the CUDO version I set, the more GPUs will be listed, and could this help solve the problem? Are there any lists of all servers by version and their real-time counts to confirm my theory?