[QUOTE=petrw1;534444]Ran for 15 minutes then on restart this...[/QUOTE]

Still ….

Do I assume I am officially punted?

[QUOTE=Chuck;534562]I wish we could get a definitive answer about what is happening here[/QUOTE]
I am not google, but I can guess what's happening.

When pressed for resources, users who have been consuming resources a lot recently will get throttled.

Wait for a few hours and try again and you should get resources again.

I don't think it is anything to do with the type of work, but rather the duration of work. Only workload they have explicitly disallowed is cryptocurrency mining (I think).

[QUOTE=axn;534566]When pressed for resources, users who have been consuming resources a lot recently will get throttled. ... I don't think it is anything to do with the type of work, but rather the duration of work.[/QUOTE]

I agree. I also think it is fairly clear that we have not been blocked because of the particular work-type (again, they could simply block GPU72 if they wanted to), and that we will /never/ get official word from Google.

For the last few days, my main account has been getting one or two 10 hour sessions per day -- the other accounts nothing. This morning I managed to get backends for four out of my five accounts; all have been running for three hours.

I also see an additional four GPU72_TF users running at the moment.

Any chance of getting a combined package of Mprime and mfaktX set up? That way if we don't get a back end we can still do P-1 or a few ECM curves.

[QUOTE=Uncwilly;534618]Any chance of getting a combined package of Mprime and mfaktX set up? That way if we don't get a back end we can still do P-1 or a few ECM curves.[/QUOTE]There's a how-to for combined cpu and gpu use at [URL]https://www.mersenneforum.org/showpost.php?p=528073&postcount=8[/URL]
(part of the Colab oriented reference thread [URL]https://www.mersenneforum.org/showthread.php?t=24839[/URL])

This example includes mprime and mfaktc and has been available since mid October.

Doesn't even require gpu72 account. Perhaps it is adaptable to mfaktc via gpu72, yet mprime as primenet connected. If one background task or the other crashes or runs out of work and so terminates, top keeps the foreground process and the surviving background task going for however long the session lasts. So it should ride out gpu-backend-unavailable pretty well.

@Uncwilly:
See [URL="https://www.mersenneforum.org/showpost.php?p=533076&postcount=710"]post #710[/URL] on this thread.
Hope you won´t go through the same.

Note: ignore the "TPU" references in the post. They were actually CPUs.

Not seen a GPU now for a few days, guess that they just decided to starve us ...

[QUOTE=bayanne;534765]Not seen a GPU now for a few days, guess that they just decided to starve us ...[/QUOTE]

Same situation for me too.

[QUOTE=Chuck;534791]Same situation for me too.[/QUOTE]

Yup. Me too. Not a single instance across five different accounts in the last 48 hours.

Oh well, it was fun while it lasted...

For the last two days my one account hasn't been able to get a GPU either, but my other account gets one every time... no idea why.

[QUOTE=James Heinrich;534794]For the last two days my one account hasn't been able to get a GPU either, but my other account gets one every time... no idea why.[/QUOTE]

I suspect Google keeps track of how much total GPU time your account has used. I also suspect they recently instituted a lifetime cap on GPU time.