Experimenting with Ubuntu 11.10 on a Q6600, 4 GB RAM with 2-5830s in Xfire. All is running, it's using both GPUs but it's taking twice as long to run as it did in Win7. I checked and both GPUs are only around 50%. I even figured out how to O/C then as well as increase fan speeds, thank Google!
Is this normal for Ubuntu? Is there anything I can do to increase the GPU loads? Somewhere in the app files? I'm a noob using Ubuntu for this so be kind to Mr. noob.

Hmm, I've got ubuntu 11.10 too, but only with one hd5750. I've noticed that under linux the gpu process takes up a lot of cpu cycles as well. So it might be possible that your cpu tasks get in the way of the gpu task(s). Try limiting the cpu work in order to free a core or multiple cores if you have multiple gpu tasks (i'm unsure about this because I have no multi gpu setup).

Good luck :)

Experimenting with Ubuntu 11.10 on a Q6600, 4 GB RAM with 2-5830s in Xfire. All is running, it's using both GPUs but it's taking twice as long to run as it did in Win7. I checked and both GPUs are only around 50%. I even figured out how to O/C then as well as increase fan speeds, thank Google!
Is this normal for Ubuntu? Is there anything I can do to increase the GPU loads? Somewhere in the app files? I'm a noob using Ubuntu for this so be kind to Mr. noob.

There is a setting in Linux that ramps up the gpu's to 100% but I do not remember what it is, sorry!

*Update; When using the new proprietary drivers from AMD (ATI), version 12.2, cpu usage drops drastically.

Before:
http://moowrap.net/result.php?resultid=9297556

After:
http://moowrap.net/result.php?resultid=9390083

Even though still the cpu-priority of gpu tasks seems to be equal to my other cpu projects and therefore I need to leave one cpu core emtpy.
Also note that the gpu is fastest at Core #0 setting (one might expect #3)

Edit: I also see the application is using dnetc518-linux-x86-stream instead of dnetc518-linux-amd64-stream in this case. I've checked with benchmarks, there is no performance difference and both benches run at low cpu usage. I guess there are different tasks which autoselect their application? (x86 vs amd64)

I also see the application is using dnetc518-linux-x86-stream instead of dnetc518-linux-amd64-stream in this case. ... I guess there are different tasks which autoselect their application? (x86 vs amd64)

Scheduler usually sends you both 32-bit and 64-bit apps and after a while it will start picking one that is measured to run optimally (faster) on your host. Since both of those stream clients contain the same GPU-code, there won't be that much real difference (just like your benchmarking showed) so this shouldn't matter.

Sometimes a host can't run either version (like 64-bit only host that doesn't have 32-bit runtime libs) so scheduler will see failures and thus should learn to not send that version to the affected host. It can still sometime retest the host by sending the "bad" version, though.

-w