Sorry, maybe it was mentioned here somewhere already but I can't find the answer to my question. Are the new MDGW CPU tasks level 3 cache intensive? If so how much does one task require?
Makes you wonder if it is possible to buffer those I/O's in a way that the 900 transactions are not the equivalent of 900 physical I/O's.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Makes you wonder if it is possible to buffer those I/O's in a way that the 900 transactions are not the equivalent of 900 physical I/O's.
Tom M
I think you are talking about a batch write and that depends on how the database is setup, if it's task by task probably not since each task needs individual data
I removed this project (Multi-Directional Gravitational Wave search on O3 (GPU) v1.02 () ) from my preferences as both my NVIDIA Quadro K2200 (4095MB) driver: 527.27 and my NVIDIA GeForce GTX 1060 with Max-Q Design (4095MB) driver: 527.56 are having 100's of WU in error.
I couldn't find a definite answer on why/how to correct. (but I may have missed it)
We are aware of an issue that can affect the Windows GPU app right now. We'll look into it ASAP but it'll take until the first week of January, unfortunately (see above). We'll update this thread as soon as we think we've resolved the issue. Until then it's of course perfectly fine to opt out of the app for the time being.
Sorry for the hassle, sometimes these bugs only manifest themselves when launching the apps full-scale, despite all beta testing we do.
Is anybody running multiple GPU WUs on this? My GPU is a Geforce RTX 3060 w/12Gb of memory. My CPU is a Ryzen 7 3800X (8 cores/16 threads). I am running my rig at 50% of CPUs @ 100% of time. Originally I ran a single unit and found the WU took about 5 minutes but then the GPU went to 0% or 1% for about 10 minutes (I assume the CPU was doing its thing at that point). So I went into settings and Preferences > Project > GPU Factor GW and set it for 0.50. Now I get 2 wus running at the same time (sometimes) without the 10 minute of GPU doing nothing.
Would there be any advantage to running the GPU Factor GW at 0.33 or 0.25? I am also running Meerkat GPU BRP7-cuda55. Interesting that at least half the time the GPU shows one GW wu (0.50) and two Meerkats wus (0.25) running at the same time. Along with my FGRPSSE CPU wus, it shows that Im working 3 different projects at once. Pretty cool and well rounded!
Any suggestions would be much appreciated! Im mainly wondering about how many GW wus would be optimal. I know... try it out and see for myself. Just thought Id get some input before I do... Thanx All
you could run 3 at a time on your 3060, but 4 may cause it to run out of memory and cause errors. unless something changed with memroy use in the last week or two, these were using about ~3200MB per task.
Sorry, maybe it was mentioned
)
Sorry, maybe it was mentioned here somewhere already but I can't find the answer to my question. Are the new MDGW CPU tasks level 3 cache intensive? If so how much does one task require?
Please note:
)
Please note: https://einsteinathome.org/content/holiday-season
Einstein@Home Project
Oliver Behnke wrote: Please
)
Enjoy your Holidays and hug your family!!
Makes you wonder if it is
)
Makes you wonder if it is possible to buffer those I/O's in a way that the 900 transactions are not the equivalent of 900 physical I/O's.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote: Makes you
)
I think you are talking about a batch write and that depends on how the database is setup, if it's task by task probably not since each task needs individual data
I removed this project
)
I removed this project (Multi-Directional Gravitational Wave search on O3 (GPU) v1.02 () ) from my preferences as both my NVIDIA Quadro K2200 (4095MB) driver: 527.27 and my NVIDIA GeForce GTX 1060 with Max-Q Design (4095MB) driver: 527.56 are having 100's of WU in error.
I couldn't find a definite answer on why/how to correct. (but I may have missed it)
Hi Zou,We are aware of an
)
Hi Zou,
We are aware of an issue that can affect the Windows GPU app right now. We'll look into it ASAP but it'll take until the first week of January, unfortunately (see above). We'll update this thread as soon as we think we've resolved the issue. Until then it's of course perfectly fine to opt out of the app for the time being.
Sorry for the hassle, sometimes these bugs only manifest themselves when launching the apps full-scale, despite all beta testing we do.
Best,
Oliver
Einstein@Home Project
No worries, I thought I had
)
No worries, I thought I had missed something, I'm glad it's under control :D
Enjoy !
Is anybody running
)
Is anybody running multiple GPU WUs on this? My GPU is a Geforce RTX 3060 w/12Gb of memory. My CPU is a Ryzen 7 3800X (8 cores/16 threads). I am running my rig at 50% of CPUs @ 100% of time. Originally I ran a single unit and found the WU took about 5 minutes but then the GPU went to 0% or 1% for about 10 minutes (I assume the CPU was doing its thing at that point). So I went into settings and Preferences > Project > GPU Factor GW and set it for 0.50. Now I get 2 wus running at the same time (sometimes) without the 10 minute of GPU doing nothing.
Would there be any advantage to running the GPU Factor GW at 0.33 or 0.25? I am also running Meerkat GPU BRP7-cuda55. Interesting that at least half the time the GPU shows one GW wu (0.50) and two Meerkats wus (0.25) running at the same time. Along with my FGRPSSE CPU wus, it shows that Im working 3 different projects at once. Pretty cool and well rounded!
Any suggestions would be much appreciated! Im mainly wondering about how many GW wus would be optimal. I know... try it out and see for myself. Just thought Id get some input before I do... Thanx All
you could run 3 at a time on
)
you could run 3 at a time on your 3060, but 4 may cause it to run out of memory and cause errors. unless something changed with memroy use in the last week or two, these were using about ~3200MB per task.
_________________________________________________________________________