How can I tell?How can I fix? 2nd GPU client not getting credit.

[dfonda]

I should be pushing 15k ppd right now. This is according to fahmon

5500 ppd Gpu and SMP

9000k ppd 2 GPU's(Same rig)

1000k SMP

These are very conservative numbers it actually shows more.(In fahmon) Here is what shows up.On EOC http://folding.extremeoverclocking.com/user_summary.php?s=&u=195217

At some point I thought I would see at least 1 spike up to 15k Maybe I am just bad with math??

I am guessing that the 2nd GPU client on the dual GPU rig,may be appearing to stanford as the same one.

I have 5 clients showing up on stanfords site if that holds true for the weekend ..then I guess all 5 are being credited.

Thanks

EDIT I was thinking the fastest way to tell would be to shut down my other clients for a day...then I could pinpoint what numbers this dual GPU rig is putting out.

[Adak]

I believe you do have a problem, Dave. Have you checked your machine ID on the GPU clients?

If the double GPU rig looks good, I'd check out the other GPU client. That would also cause a loss of about 4,000 ppd.

[harlam357]

Yes... make sure it's not Folding the same work... and double check the machineid on the dual GPU rig... mine is cranking along nicely. :)

[dfonda]

Thanks guys I went over every thing last night...but was really tired. I'll check the id now.

It should be the double rig.

[dfonda]

Thanks again guys, It was right in front of me...both id's were set at 2.

Here is what I did:

Closed both clients.

Used notepad to open client.cfg I changed GPU1 to machineid1. I also changed the core from 1 to 0 thinking I could get both cores to 100%(Doesn't seem to have worked.) We should be able to do this right?

I checked the log file and I now have a machineid of 1 and the other log shows the machineid remaining at 2.

This seems right am I missing anything...I'll know in the afternoon I reckon...when I see a boost hopefully.

[harlam357]

I also changed the core from 1 to 0 thinking I could get both cores to 100%(Doesn't seem to have worked.) We should be able to do this right?

Do you mean you're trying to get one GPU client running on core 0 (not-default) and the other on core 1 (default) by setting the NV_FAH_CPU_AFFINITY environment variable?

Don't worry about trying to run each client on its own cpu core... both clients can co-exist on the same cpu core without any hit in ppd.

[dfonda]

Ok H I'll leave the cpu alone.

Looks promising I have 6 active processors in the last 7 days:thup:...thats up 1 from this morning. I finished using a client about a week ago so it should go back down to 5 tomorrow or the next day.

Now all I need is a good spike today and I'll be positive.

Thanks for helping me tie my shoes this morning.;)

[harlam357]

Hehe... hey dude, in my mind your idea was perfectly sound. I did the same thing... but found that there's no hit in production keeping both GPU2 processes on the same CPU core. So why max out two cores for no gain? :D

As hungry as the GPU2 core is on XP I figured to get full production out of two GPUs it would need two CPUs (one for each GPU)... that would make sense wouldn't it? But, for whatever reason that's not the case. :)

[Adak]

The GPU client only uses a small part of the cpu client that is servicing it, so two of them on one core is no problem.