Discussion Thread for the Continuous GW Search known as O2MD1 (now O2MDF - GPUs only)
Language
Copyright © 2024 Einstein@Home. All rights reserved.
Copyright © 2024 Einstein@Home. All rights reserved.
For an RX 560 4GB gpu what is
)
For an RX 560 4GB gpu what is the "perfect" combination of CPU and gpu settings? eg. 1:1, 0.5:0.33 etc.
Thank you.
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!
My slowest PC, an HP Laptop
)
My slowest PC, an HP Laptop with an E-450 CPU and no graphic board, is running both GW 02 tasks and gamma ray pulsar tasks with slow but good results, all validated. It was not accepted in Science United because its 7.8.3 BOINC manager was declared "too pld". I have been running more than 40000 Milkyway@home tasks, both CPU and GPU, and thousands of Asteroids@home tasks, idem,very little else. When I see that the GPU is idle, I load GPUGRID GPU tasks, all validated, via my standard BOINC manager. I see they are listed in the Science United Scientific Projects list, although they were not called via Science United.
Tullio
Tom M wrote: Peter Hucker
)
Yeah I guess you're right. Competition invites invention. Just as long as we can do away with the copyright trademark patent crap.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
Tom M wrote:For an RX 560
)
That's exactly the GPU I have! In a system with an i5 8600K. I recently gave up doing gravity on it. If I run one at a time, it only uses 70% of the GPU. If I run two, it runs the risk of running out of GPU RAM and slows down tremendously. I now run gamma on it, 2 at a time.
As for CPU cores - I gave up on that too. I changed all GPU tasks to use 0.01 CPU cores (as in zero, but entering 0 means something special). Then I just set the global limit to the number of cores left over when the hardest GPU stuff was running. The main reason for this was LHC. If I allowed all my 6 cores to be used for Boinc, then told the GPU tasks to take 1 CPU core each, Boinc insisted on downloading 6 core multithreaded CPU tasks from LHC, which whenever they ran slowed the GPU down. Boinc sometimes just doesn't work right.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
tullio wrote: My slowest PC,
)
I think I'll stay clear of this "Science United" thing.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
On a SETI@home message board
)
On a SETI@home message board the say that OpenCl on nVidia boards can use no more than 27% of the Video RAM, not the system RAM. Since my Video RAM is 3 GB, 27% of it is about 800 MB and that seems to too small for GW 02 tasks, while it was sufficient for GW 01 tasks last year. As a comparison, GPU-Z shows that Asteroids@home tasks take about 700 MB and Milkyway@home take 400 MB, so they are successful , like GPUGRID tasks which I have not tested yet with GPU-Z, but I shall do when I get them running.
Tullio
tullio wrote:On a SETI@home
)
My AMD cards use all their own RAM and the system RAM just fine. I find it hard to believe Nvidia are that restrictive. I've seen many people mention that they run multiple work units (as many as 7 per card) on them. I think Keith Myers in this thread knows more about multiple WUs.
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
I always run only one task on
)
I always run only one task on a GPU card. If one task is not successful, two or more could not run. Only Einstein@home GW 2 fail on my GTX 1060, all other projects run just fine.
Tullio
tullio wrote:I always run
)
I'm surprised they fail on a card that good. It has 6GB? My 4GB AMD RX 560 can run one GW ok, but I stopped it as it was only 70% utilised (it's always waiting on the CPU, which is odd as the CPU is better than the card, maybe the card isn't good at most of the calculations so hands loads over to the CPU?). I've decided that it's best to let the people with better cards do gravity and to do gamma on all mine (which are 3GB and 4GB).
If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.
Peter Hucker wrote:I'm
)
If you clicked on Tullio's computers link, you could see that the card in question is reported as "NVIDIA GeForce GTX 1060 3GB (3072MB)".
The 3GB model is cut back a little from the standard 1060 in some other aspects than memory size, also.