Questions and Answers :
Preferences :
Could simultaneously running SETI@Home + GIMPS + a Bitcoin full node mess with eachothers results?
Message board moderation
Author | Message |
---|---|
Angelino Desmet Send message Joined: 18 Dec 19 Posts: 10 Credit: 708,004 RAC: 3 |
Even though SETI@home is GPU intensive, GIMPS is CPU intensive, and the Bitcoin full node isn't intensive, I definitely notice that running these programs simultaneously slows each separate process down. For example a GIMPS calculation goes from 11 days, to 11 days ~10 hours. That it takes each process slightly longer doesn't bother me. But what I wouldn't want is more inconclusive results or an increased chance of calculation errors. I overclocked my i7-3820 from 3.6 GHz to 4.5 GHz and the highest core temperature after running these programs for 24 hours is 68°C and the lowest ~63°C. However, the official Tcase temp is 66.8°C. Should I turn it down a notch? Or will the results will be consistent either way. [size=12]https://www.cynicusrex.com: a sentient stack of stardust pondering nothing and everything. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Running multiple projects won't cause issues with the results, but overclocking can. And overclocking that is OK for one project, may cause issues on another project due to the different types of work involved stressing different parts of the CPU to different levels. If you choose to overclock, all you can do is check the results for each project to make sure none are getting invalid or erroneous results. If so, then reduce the overclock. Grant Darwin NT |
Angelino Desmet Send message Joined: 18 Dec 19 Posts: 10 Credit: 708,004 RAC: 3 |
If you choose to overclock, all you can do is check the results for each project to make sure none are getting invalid or erroneous results. If so, then reduce the overclock.Can false positives happen? [Or should it be false negatives?] I.e. a valid result which actually has errors in it? For example this RX 5700 XT problem. [size=12]https://www.cynicusrex.com: a sentient stack of stardust pondering nothing and everything. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
Can false positives happen? I.e. a valid result which actually has errors in it?If it has errors, it's not valid. That's what can happen with overclocking. Of course if you get 2 systems producing the same error and they are paired together for validation, then you end up with the RX 5700 XT problem (which is well known- it's an issue with their drivers relating to OpenCL support. They work OK on some projects, but cause issues on many others). Grant Darwin NT |
Angelino Desmet Send message Joined: 18 Dec 19 Posts: 10 Credit: 708,004 RAC: 3 |
Alright, thank you. For now I've only had one inconclusive result which became valid, so I'll keep things how they are while being on the lookout.Can false positives happen? I.e. a valid result which actually has errors in it?If it has errors, it's not valid. That's what can happen with overclocking. [size=12]https://www.cynicusrex.com: a sentient stack of stardust pondering nothing and everything. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Even though SETI@home is GPU intensiveYeah, but for that GPU work runs on the CPU as well. Which is what'll cause your time differences. All that the GPU does is calculate the kernels, but the CPU does all the hard work of translating the raw data into those kernels, transfer those to the GPU memory, wait around for it to be finished, transfer the crunched data back to main memory, translate it back into something humans can read, then save it to disk. At some projects, and I don't remember if that's here as well, the FFT algorithm runs on the CPU as well, as it's more efficient to do it on there than on the GPU. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14649 Credit: 200,643,578 RAC: 874 |
At some projects, and I don't remember if that's here as well, the FFT algorithm runs on the CPU as well, as it's more efficient to do it on there than on the GPU.I think that's one we can cross off the list for SETI. Both CUDA and OpenCL have effective FFT libraries, and the SETI GPU apps use those, rather than the cpu-based FFTW library. |
Angelino Desmet Send message Joined: 18 Dec 19 Posts: 10 Credit: 708,004 RAC: 3 |
Yeah, but for that GPU work runs on the CPU as well. Which is what'll cause your time differences. All that the GPU does is calculate the kernels, but the CPU does all the hard work of translating the raw data into those kernels, transfer those to the GPU memory, wait around for it to be finished, transfer the crunched data back to main memory, translate it back into something humans can read, then save it to disk. At some projects, and I don't remember if that's here as well, the FFT algorithm runs on the CPU as well, as it's more efficient to do it on there than on the GPU.So are you discouraging what I'm doing? Or can I keep going if it doesn't cause errors. [size=12]https://www.cynicusrex.com: a sentient stack of stardust pondering nothing and everything. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13720 Credit: 208,696,464 RAC: 304 |
So are you discouraging what I'm doing? Or can I keep going if doesn't cause errors.Ok to continue if no errors occur. Grant Darwin NT |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.