Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
2 days for a genefer workunit?
| Author |
Message |
|---|
|
|
|
hi all
just wanted to ensure that everything is fine as I see a lot of people completing these workunits significantly faster than me.
does the following look right to you?
linux (kubuntu 14.04, 64-bit)
16gb RAM
i7 3.4 GHz cpu
nvidia GeForce GTX 650 (1023MB)
cudaGFN 3.03 for genefer tasks
4 LLR and one genefer task running simultaneously, but hyperthreading no deactivated
is about 2 days for a workunit about right here? or is something wrong?
| |
|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
                           
|
hi all
just wanted to ensure that everything is fine as I see a lot of people completing these workunits significantly faster than me.
does the following look right to you?
linux (kubuntu 14.04, 64-bit)
16gb RAM
i7 3.4 GHz cpu
nvidia GeForce GTX 650 (1023MB)
cudaGFN 3.03 for genefer tasks
4 LLR and one genefer task running simultaneously, but hyperthreading no deactivated
is about 2 days for a workunit about right here? or is something wrong?
It doesn't seem unreasonable. I was running about 22 hours on a GTX 460, and the 650 is a slower GPU. The x6x models are faster than x5x models (that's the digit that tells you where part falls in the lineup, with 8 being the flagship GPU and 1 being the bargain "value" part.) Furthermore, the 6xx series GPUs were actually slower (i.e., have poorer double precision performance) than the older 4xx and 5xx series GPUs. All told, I could see a 650 being half as fast as a 460. So, yes, 2 days doesn't seem unreasonable.
That being said, I looked at some of your results, and you're getting some errors, including the "maxErr exceeded" error. That's indicative of hardware failures, which is typically (on Nvidia) caused by overclocking. Even factory overclocking is sometimes problematic.
If your GPU is overclocked -- whether by you, or as a factory overclocked GPU -- you should try lowering the memory clock speed to stock speed. That won't affect performance much, but it will make the GPU a lot more reliable.
If it's already running at stock clock speeds, you can try lowering the memory clock below stock. I haven't heard of that being needed on a GTX 650, but it was necessary on some older GPUs.
____________
My lucky number is 75898524288+1 | |
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
                                     
|
|
Just to confirm, 2 days for a GTX 650 is about right. For comparison, my OEM GTX 645 cards (which are faster than a GTX 650--GTX 650 has 384 shaders vs. GTX 645 with 576 shaders) runds the GFN short in about 26 to 27 hours. Thus, depending on your clocks, a GTX 650 should be in the 40 to 50 hour range.
Also, to follow-up on Mike's note regarding your errors, the overclocking might be the culprit, but a memory overclock is probably not the issue here from my experience on very similar cards (e.g., GT 640 cards). Rather, the overclock and/or case environment might be resulting in a lot of heat issues, especially if you are using a stock GPU cooler as many of the coolers on these entry level cards are not very efficient.
One other common issue that arises with these lower end cards is that they don't handle other loads very well while crunching. Unlike high end gaming cards, my experience with entry level cards is that they produce more frequent errors if the computer is being used for other graphics work even if it is not very intensive (e.g., a simple game like solitaire, etc.).
| |
|
Message boards :
Generalized Fermat Prime Search :
2 days for a genefer workunit? |
