Folding on vast.ai

[HandOverFist]

Does anyone here use this and is this headless docker still valid? ohgodatrini/fahclient:v7_5_1 I have had no success getting it started.

[Joe_H]

I don't know if anyone still uses it, but FAHClient v7.5.1 is old and out of date. Current version is v8.5.5, and the last v7 version was released about 5 years ago - v7.6.21.

[HandOverFist]

It is the only docker I have found related to FAH.

[Joe_H]

Containers are available from - https://github.com/FoldingAtHome/containers/releases. Latest version there is from just over 4 years ago and supports 7.6.21.

There is also this one specifically for GPU folding - https://hub.docker.com/r/foldingathome/fah-gpu

[HandOverFist]

So, not much of a update lol. I do have it working now as headless, but the gpu appears to be continually dropping load and starting again. I have yet to see any points generated under my username/team #/passkey. I do know that there are many on the team who still use the outdated/older versions without a problem.

Edit: With one 5090 records show it only managed to complete (2) WU's in almost four hours of (supposed) run time. I killed that instance and started another on a different machine with four 5090's. Will give it another four hours to decide if this is even viable.

[calxalot]

I asked for help on discord. I know some folks are running v8 on vast.ai.

[calxalot]

May be helpful: https://github.com/JWhyFR/fah-v8

[HandOverFist]

On a lark I changed the gpu from a 5090 to a RTX PRO 6000 and it appears to be holding steady atm.

Edit: Nvm...something is broken.

[toTOW]

I know two prebuilt FAH v8 templates for Vast.ai :

• This one is from Firedfly : https://cloud.vast.ai/?ref_id=27485&tem ... 88d13222b1
• This one is from JWhy : https://cloud.vast.ai/?ref_id=260637&te ... 580760e7fd

They are quite simple to use, all you have to do is to save it to your own Vast.ai account and add your FAH account token to the corresponding field in the template environment variables.

There's a full comprehensive guide on the Alliance Francophone's website (in French) to help about JWhy's template : https://www.alliancefrancophone.org/dossiers/gpu-cloud

[HandOverFist]

I know two prebuilt FAH v8 templates for Vast.ai :

• This one is from Firedfly : https://cloud.vast.ai/?ref_id=27485&tem ... 88d13222b1
• This one is from JWhy : https://cloud.vast.ai/?ref_id=260637&te ... 580760e7fd

They are quite simple to use, all you have to do is to save it to your own Vast.ai account and add your FAH account token to the corresponding field in the template environment variables.

There's a full comprehensive guide on the Alliance Francophone's website (in French) to help about JWhy's template : https://www.alliancefrancophone.org/dossiers/gpu-cloud

This appears to be working, but the ppd on (2) RTX 5090's appears rather low. I'll keep a eye on it and report back. Thanks so much for the help!

Edit: It appears ppd performance varies vastly between machines given the same gpu. It pays to experiment.

[muziqaz]

I know two prebuilt FAH v8 templates for Vast.ai :

• This one is from Firedfly : https://cloud.vast.ai/?ref_id=27485&tem ... 88d13222b1
• This one is from JWhy : https://cloud.vast.ai/?ref_id=260637&te ... 580760e7fd

They are quite simple to use, all you have to do is to save it to your own Vast.ai account and add your FAH account token to the corresponding field in the template environment variables.

There's a full comprehensive guide on the Alliance Francophone's website (in French) to help about JWhy's template : https://www.alliancefrancophone.org/dossiers/gpu-cloud

This appears to be working, but the ppd on (2) RTX 5090's appears rather low. I'll keep a eye on it and report back. Thanks so much for the help!

What do you mean by low?

[HandOverFist]

I know two prebuilt FAH v8 templates for Vast.ai :

• This one is from Firedfly : https://cloud.vast.ai/?ref_id=27485&tem ... 88d13222b1
• This one is from JWhy : https://cloud.vast.ai/?ref_id=260637&te ... 580760e7fd

They are quite simple to use, all you have to do is to save it to your own Vast.ai account and add your FAH account token to the corresponding field in the template environment variables.

There's a full comprehensive guide on the Alliance Francophone's website (in French) to help about JWhy's template : https://www.alliancefrancophone.org/dossiers/gpu-cloud

This appears to be working, but the ppd on (2) RTX 5090's appears rather low. I'll keep a eye on it and report back. Thanks so much for the help!

What do you mean by low?

I mean ppd output on a 5090 was about half of what it should have been per WU. Currently running again on a different machine and the gpus are getting over double the previous ppd with the same WU.

[muziqaz]

This appears to be working, but the ppd on (2) RTX 5090's appears rather low. I'll keep a eye on it and report back. Thanks so much for the help!

What do you mean by low?

I mean ppd output on a 5090 was about half of what it should have been per WU. Currently running again on a different machine and the gpus are getting over double the previous ppd with the same WU.

No, no, when I asked what do you mean by low, I asked what is the magic number. Your expectations might be misguided, for all we know

[HandOverFist]

Previously was 20M ppd...now 54M ppd on the same WU.

Edit: Just made over 213M ppd in less than 25 minutes with (4) 5090's. Lowest scoring gpu was 48M ppd.

[muziqaz]

Previously was 20M ppd...now 54M ppd on the same WU.

Edit: Just made over 213M ppd in less than 25 minutes with (4) 5090's. Lowest scoring gpu was 48M ppd.

magic
Thanks