Folding Forum

Hello,

I'm relatively new to this. I've been able to fold with CPU but GPU does not work.

I'm using an old dual Xeon system that I threw four GT210s in. F@H client sees them but tries to get the WUs from 0.0.0.0, how do I fix this?

Update: I also tested a Quadro FX 1700 (different machine too) and nothing happened. The only other GPUs I can borrow are Radeon X1300s from school. I did, however, get an AIO with a Radeon HD 7600A running both CPU and GPU.

Welcome to the folding support forum.

Your GT 210 cards use Tesla based GPU's. These older pre-Fermi GPU's are only semi-supported for folding as the main projects that were assigned to them ran out about a year and a half ago. A few projects that use a later folding core can run on some pre-Fermi GPU's, but they are also approaching end of life. For assignment purposes these need to be opted into and require two additional options to be set.

If you want to opt in, add the following options to your GPU slots:

max-packet-size set to small

client-type set to beta

While the beta flag is being used in combination with the size setting, these WU's are no longer beta. The two settings together just are being used to indicate to the AS that you want to process these older projects.

Edit: The same issue applies to the Quadro card, it is pre-Fermi. As for the Radeon card, that is too old. The minimum required ATI series Radeon card is the HD 5000 cards. The lower power cards from that series may not work at all, or not complete within deadlines.

Joe_H wrote:Welcome to the folding support forum.

Your GT 210 cards use Tesla based GPU's. These older pre-Fermi GPU's are only semi-supported for folding as the main projects that were assigned to them ran out about a year and a half ago. A few projects that use a later folding core can run on some pre-Fermi GPU's, but they are also approaching end of life. For assignment purposes these need to be opted into and require two additional options to be set.

If you want to opt in, add the following options to your GPU slots:

max-packet-size set to small

client-type set to beta

While the beta flag is being used in combination with the size setting, these WU's are no longer beta. The two settings together just are being used to indicate to the AS that you want to process these older projects.

Edit: The same issue applies to the Quadro card, it is pre-Fermi. As for the Radeon card, that is too old. The minimum required ATI series Radeon card is the HD 5000 cards. The lower power cards from that series may not work at all, or not complete within deadlines.

Thanks for the reply! I'll see if I can get that to work.

Joe_H wrote:
max-packet-size set to small

client-type set to beta

.

Okay, it's showing Work and Collection Server addresses now after a reboot. But again, the CPU gets right to business and the GPUs do nothing. This time, it says they're waiting on FahCore Run.

By default, the as installed settings for GPU folding slots are to run when the system is idle. That is defined by either the screen being shut down or the screen saver starting up. You can also set the slots to fold all of the time by right clicking on the slot in FAHControl.

Joe_H wrote:By default, the as installed settings for GPU folding slots are to run when the system is idle. That is defined by either the screen being shut down or the screen saver starting up. You can also set the slots to fold all of the time by right clicking on the slot in FAHControl.

Can you provide a screenshot please?

Many screen shots included. https://folding.stanford.edu/home/guide ... all-guide/

This is what it shows



Sometimes it does this too.

What happens if you remove the CPU-slot?

In those two screen-shots, you'll notice that status of the GPU WUs changed from READY to UPDATE_CORE, which is a good sign.

First, I would pause 3 of the 4 GPUs and see if it will run by itself. If it does, then allow a second one to start folding., etc. Sometimes when multiple slots try to update the core simultaneously, the FAHClient has problems although I've never seen four of them all trying at the same time.

In either case, it would be helpful to see the top segment of your log. (See instructions in my signature.)

Yes, if all 4 slots are running the same core type, and one or more slots is trying to update that core, all slots except the actively folding slots will stop in the Core Updating status until that last slot is no longer in use.

Pause all slots. Start one GPU slot and let it get going until it shows at least 1% completed. Then pause it, and start the next slot. Rinse, Repeat.

http://prostoporno.me/http://prostoporn ... oporno.me/

Which GPU drivers did you install?

Which Operating System are you running?

We really need to see the log (see my previous post).