How about adding something about which ones have optimized apps.
Milkyway and Collatze show huge gains on the ati cards with optimized apps.
my 4890 can get anywhere from 70k to 90k ppd gpu alone. and only uses 0.02 of the cpu
and seti shows decent gains with with nvidia cards.... my i920 alone will do about 8kppd, but when matched with a gtx260 and the modified apps, gets @16kppd.. that's an awesome gain.
yeah, i'm at 20k ppd right now on an x3350 @ 3.36 with a gtx 260 (192) and 9600gso. admittedly, it is used for basically nothing else, but that's still quite impressive for seti.
do you have to do anything special to gpu crunch in aqua? I only use that when seti is out of work (which seems to be quite often lately) and I don't think that I've ever seen a gpu-specific wu.
Now that I have an nVIDIA GTX460 GPU, on Ubuntu 64-bit, allow me to express my frustration with the state of GPU work today:
PrimeGrid - Doesn't work with Fermi-based GPUs without an app_info.xml file. Plus they're pausing this week to switch apps.
Collatz Conjecture - Doesn't work with Linux 64-bit without an app_info.xml file.
Milkyway@Home - Doesn't work with Fermi-based GPUs! This is a bug they've known about for months!
Seti@Home - No work! Period!
Einstein@Home - No current GPU work.
I haven't tried the following, but...:
Aqua@Home - long WUs, right? And sounds like no work either.
GPUGRID - long WUs, right?
The Lattice Project - No Linux GPU apps, plus huge memory requirements?
Not BOINC, so I might rather do PrimeGrid Project Staging Area work:
Folding@Home - Biology/Medicine - Geforce 8 and up - Windows, but I hear it works with Wine?
RC5-72 - Cryptography - Geforce 8 and up - Windows/Linux/Mac OS
That's the reason, and that's no problem. FreeHAL, for some weird reason, uses about 1% of your CPU time, and no GPU time. You can configure it to run something like 25 tasks at once, if you have the memory. What it does with that tiny sliver of computing I really don't know.FreeHAL@home ____running
Is your gpu doing any work at all? You may want to check the boinc preferences at the project website (your account page) and make sure the answer for "Suspend GPU work while computer is in use?" is "no". If your gpu is not crunching at all, then something else is wrong. Boinc version 6.10.58 is the latest stable version that supports GPU crunching.
Well it has some GPU tasks, and has completed some too (2k today!) but half the time it switches away from the GPU tasks to different products that are non-GPU. I upped Collatz's resource share to 1000 hoping it'll get preference, and it's only downloading GPU so we'll see if that fixes it. As I type this it isn't doing any GPU tasks though, just 4 AlmereGrid units...
Check the boinc program preferences. On the processor usage tab check "use GPU while computer is in use" and then enter "0" in "while processor usage is less than "0" percent (no restrictions).
You can configure any project (that haven't got too old server-version that is) to be a "backup"-project, by setting it's resource-share to zero.On a much more related topic to the OP, I find it very difficult to split GPU resource share between Collatz and Milkyway, as Milkyway has a 24 task limit, whereas Collatz seems to download over 100 tasks if I allow it. Collatz tasks also last longer, and Milkyway tasks don't seem to begin until all Collatz tasks are finished. The only workaround seems to be updating Collatz manually, setting it to not fetch any more work, and aborting some of the tasks. I do wonder if this is related to the problem in my post above, but as it is project-specific I'm not too sure about that. I'm not even sure this one is a bug, or if the code is just written that way deliberately!
Thanks. I have actually suspended Collatz now, as my poor 4870 seems to have intermittent problems sometimes causing blue screens, and Collatz seems to be the common feature in these BSODs.
No BSODs for the last 2 days since I stopped Collatz.
I will bear in mind how to set a backup project for after the f@h race