Guide to running multiple GPU work units

[t_ski]

Sorry, no rest for the weary...

 

[manofthem]

"Project is temporarily shut down for maintenance" popped up again this morning, looks like it started around 3:20am est and still down. Hopefully it won't last too long.

 

[t_ski]

The only thing I keep seeing is that all my systems are sitting at "waiting to report" for several hours. I keep clicking on "do network communication" to make them go through. Maybe the server thinks I'm spamming it? lol

 

[manofthem]

The only thing I keep seeing is that all my systems are sitting at "waiting to report" for several hours. I keep clicking on "do network communication" to make them go through. Maybe the server thinks I'm spamming it? lol

Yes, mine sat there saying "waiting to report" under Tasks in the manager, but I saw the other thing in the Even Log, which doesn't log as an error, just an entry. I didn't check how long it was down, just that it was working again later this morning. This is 2 consecutive days now; maybe it's going to be an everyday thing.

 

[[Ion]]

My log file showed that the project was done for about three hours again. I really wish they'd stop doing this--I don't like to run out of GPU WUs

 

[manofthem]

My log file showed that the project was done for about three hours again. I really wish they'd stop doing this--I don't like to run out of GPU WUs

That's what I'm talking about! Yesterday and today, 3 hrs of downtime each day: that's already 6hrs of WUs that could be running, and we are missing them!!!

 

[[Ion]]

That's what I'm talking about! Yesterday and today, 3 hrs of downtime each day: that's already 6hrs of WUs that could be running, and we are missing them!!!

Well, do realize that the cache should have mitigated part of the loss each time. But still---either I'd like them to stop doing maintenance, or, better yet, just let us store 1000+ WUs in our cache.

 

[manofthem]

Well, do realize that the cache should have mitigated part of the loss each time. But still---either I'd like them to stop doing maintenance, or, better yet, just let us store 1000+ WUs in our cache.

True, I didn't really take that into consider. ....but.... I go through 20 WU's in under 10 minutes. Yesterday I had 153 ready, which means that it must have taken ~70-80 minutes for those to finish. Then the rest was downtime, and me no likey downtime

 

[MetalRacer]

Is it possible to run multiple GPU's that are not crossfire compatible?

 

[mstenholm]

Is it possible to run multiple GPU's that are not crossfire compatible?

I'm almost sure that it is. One problem could be that you have to use the same app file (the one that defines how many GPU WUs you will run at the same time) and if one card can do 1 and the other 10....

A workaround is to make two BOINC instances.

Are we talking NVidia?

 

[t_ski]

Since he's asking about Crossfire capable, I'm going to assume ATI and not Nvidia. If both cards will run in the system, you should be able to crunch on them. You may get mixed results if you're using a custom app_info.xml file to load multiple WU's on each card. A higher-end card will handle the WU's better than a lower end card.

 

[mstenholm]

Since he's asking about Crossfire capable, I'm going to assume ATI and not Nvidia. If both cards will run in the system, you should be able to crunch on them. You may get mixed results if you're using a custom app_info.xml file to load multiple WU's on each card. A higher-end card will handle the WU's better than a lower end card.

Yes my first thought but in his spec 7970s are mentioned and I assume that identical GPUs can run Crossfire. It was only to get most information for the next question...

 

[Daimus]

Is it possible to run multiple GPU's that are not crossfire compatible?

I'm running HD7850 and HD6950 in the same system. They are not crossfire compatible.

 

[KieX]

Thought I'd post this here, as it's GPU related.

Message from uplinger (WCG Tech):

Low Work for Help Conquer Cancer ( Nov 20, 2012 )
Greetings all,

We are currently running low on work for Help Conquer Cancer. At the moment of posting this we have about 1 hour of work left to send on the grid. We are running through many work units per day as you may already know and we are working with the researchers to get more work units onto the grid as soon as possible.

Normally we would slow a project down to help keep the project running on the grid. But this solution will not work since it is our only GPU project. So, to keep things fair between CPU and GPU we have decided to keep the weight normal until we run out of work.

This is not the end of the project, we are low on work from the researchers.

Thank you for your participation and patience as we get more work for Help Conquer Cancer. Please let me know if you have any questions.

-Uplinger

Source

 

[[Ion]]

Thought I'd post this here, as it's GPU related.

Message from uplinger (WCG Tech):


Source

Ggahhhh :shadedshu


I guess on the upside it means we're doing an amazing job at this!

 

[Daimus]

hought I'd post this here, as it's GPU related.

Message from uplinger (WCG Tech):

Thanks for posting this. Now I understand why the server does not send the GPU WU's.

Without GPU WU's we may lose challenge.

 

[KieX]

Thanks for posting this. Now I understand why the server does not send the GPU WU's.

Without GPU WU's we may lose challenge.

Well, it was bound to happen sooner or later. And when we had the chance we showed a lot of other teams how it should be done

 

[Daimus]

Well, it was bound to happen sooner or later. And when we had the chance we showed a lot of other teams how it should be done

"thanks" button disappeared.
One of my crunchers is empty of GPU WU's, but 2 other have they enough. I'm going to reconfigure videocards.

 

[KieX]

"thanks" button disappeared.
One of my crunchers is empty of GPU WU's, but 2 other have they enough. I'm going to reconfigure videocards.

That's happened to me very often over past few weeks

I got a big cache of CPU WU with my app_info.. hopefully new WU's come in before they're completed.

 

[manofthem]

Thanks KieX for posting, a big answer to a lot of things lately. This sucks, as my PPD is going to plummet lol. I really hope they get it going again soon; I hate to see my gpus sitting there with nothing to do.

I'll have to get my CPU going tonight when I get home

 

[KieX]

Thanks KieX for posting, a big answer to a lot of things lately. This sucks, as my PPD is going to plummet lol. I really hope they get it going again soon; I hate to see my gpus sitting there with nothing to do.

I'll have to get my CPU going tonight when I get home

Yeah.. going to be quite a difference. Your CPU should be good for 6.7K PPD. Wasn't that long ago it was considered a really good score for the Wattage. But then the 10-80K PPD GPU came along to distort all notions of points

 

[manofthem]

Yeah.. going to be quite a difference. Your CPU should be good for 6.7K PPD. Wasn't that long ago it was considered a really good score for the Wattage. But then the 10-80K PPD GPU came along to distort all notions of points

Yeah I understand. The gpu WUs have been a tremendous start for me, which has been nice, but now I hope this drop doesn't kill our team's output for the competition. I suppose we just have to wait and see.

 

[Norton]

Yeah I understand. The gpu WUs have been a tremendous start for me, which has been nice, but now I hope this drop doesn't kill our team's output for the competition. I suppose we just have to wait and see.

Remember it will affect the other Teams as well... if they have been moving into using the app_info file for gpu's then some Teams will be caught with their pants CPUs down.

 

[mstenholm]

Remember it will affect the other Teams as well... if they have been moving into using the app_info file for gpu's then some Teams will be caught with their pants CPUs down.

Remind me what happens when you remove the app_info file. Will it delete all? No, I think it was the other way around - when you add the file it would delete all. I got around 4-5 hours left in my 2x7770 (8 GPU WUs at the time) and I don't want to stay up until 3 am to optimize the right time to delete the file.

 

[KieX]

Remind me what happens when you remove the app_info file. Will it delete all? No, I think it was the other way around - when you add the file it would delete all. I got around 4-5 hours left in my 2x7770 (8 GPU WUs at the time) and I don't want to stay up until 3 am to optimize the right time to delete the file.

If you have any tasks with the app_info installed and remove it you'll lose those tasks too. Any significant change need to have the WU aborted or copleted.

Remember it will affect the other Teams as well... if they have been moving into using the app_info file for gpu's then some Teams will be caught with their pants CPUs down.

SG have about 200+ more users than us. If it's a CPU shootout it would be tough.