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.
Peter Hucker wrote:Zalster
)
My gamma tasks are 46 in progress and 131 pending on a single gpu here. I finish a wu in about 1328 seconds
TBar wrote:So... watch my
)
What do you mean by "lockdown buster"?
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:Zalster
)
I appreciate the offer but you don't have to if you want to keep doing Gamma. If I was really, really frustrated, then I would just fire up another machine and set it for GW only as well. Then I could be my own wingman on the results, lol. But I'm not at that point yet. My fleet is spread out across 3 projects so I have enough to keep me occupied.
Quote:So you don't need
)
No, back to 4 pages of pendings with no wingmen assigned yet for the Velajr1 tasks. Basically everything sent, crunched and returned today. All of yesterday's pendings I was speaking of earlier this morning have been assigned a wingman.
Tom M wrote:Thank you for
)
Here is my newest version of my cc_config.xml I have restarted the boinmanager/clients.
<cc_config>
<log_flags>
<sched_op_debug>1</sched_op_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
<save_stats_days>365</save_stats_days>
<exclude_gpu>
<url>http://einstein.phys.uwm.edu/</url>
<device_num>0</device_num>
<app>einstein_O2MDF</app>
</exclude_gpu>
<exclude_gpu>
<url>http://einstein.phys.uwm.edu/</url>
<device_num>1</device_num>
<app>einstein_O2MD1</app>
</exclude_gpu>
</options>
</cc_config>
---edit---
I have also switched my location to "home" which is a GW gpu only profile (the Windows box is running on it). Its going to take a while to see if I download any GW gpus at all.
I am NNT Cosmology@Home to free up that thread for cpu GW processing. I was going to do the same with WCG (4 threads) until I discovered that I am now processing a Rosetti Beta task on that project.
----edit---
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!
Interesting. On
)
Interesting.
On https://einsteinathome.org/host/12806167 which is my Ubuntu Intel mining rack system I just reverted to Nvidia driver 4.10 for Linux and now my Gtx 1060 3GB is successfully running GW gpu tasks. (I only have 2 Nvidia cards on it right now. The AMD cards are still in the mail).
gtx 1060 3Gb GW gpu tasks were failing under 4.40 as previously noted by Tbar who has a lot of 1060 3Gb'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:Interesting. On
)
i don't think it has to do with the driver. some GW workunits require more memory than others, even under the same "type", such as VelaJrs: Ive seen some use more than 3GB, ive seen some use less than 1GB. you will probably still fail if you get sent another one requiring more than 3GB again.
_________________________________________________________________________
Ian&Steve C. wrote:i don't
)
I watched it erring out one right after the other. I changed the driver version to 4.10 and it stopped. I agree correlation is not causation.
--edit---
I haven't been able to locate an error from the short tasks or any short tasks at all which is suspicious. But as Ian/you said, 4.10 doesn't keep the 10 second tasks from happening completely because I did spot some more of them.
However I finally got my exclude working. I am not sure if I can get it to download any new tasks but at least it has stopped running GW gpu on the gtx 1060 3GB card.
--edit---
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!
mikey wrote:My gamma tasks
)
I have 4 decent GPUs (the 5th is a mini one in my normal desktop which is quieter!), each can do a gamma in 943 seconds. But they're shared between 2/3rds Gamma, and 1/3 Milkyway Seperation.
I've got 1010 Gamma awaiting validation! My wingmen are taking too long.... I've reduced my buffer to 3+3 hours, so WUs are returned faster, which I assume is better for the project, aswell as keeping things tidier on my end, especially if I want to change projects or weights.
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.
Zalster wrote:I appreciate
)
I want to do whatever the Einstein server wants to hand out. But 4 of my 5 cards are limited to Gamma only due to 3GB GPU RAM and being attached to wimpy CPUs. The 4th GPU is 4GB GPU RAM with a good CPU, so it does whatever the server gives it, which is surprisingly a fair bit of Gamma, even though the GW queue is much larger.
I wonder if the server deliberately chooses a wingman to be someone with a different model of GPU? That would make more sense in terms of checking for mistakes. In which case if yours are all the same you couldn't er.... play with yourself.
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.