Message boards :
Number crunching :
MultiBeam application for ATi GPUs released
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11
Author | Message |
---|---|
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I once had the same issue on my card. It was a bunch of bad downloads. Had more than 30 units errored exatly the same second. So i aborted the rest of them. It never happened again. So i expect it was the units. With each crime and every kindness we birth our future. |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
I noticed mutliple errored WU's all ending at exactly 2988 seconds. I thought this was a bit odd and did a little search. I found that the Flops in the App_info may be a cause of this error. I also found this also my be due to registry entries. Since I've been uninstalling and installing the ATI drivers its a possibility that the old drivers left behind traces on the registry which may have caused the OpenCL to fail to load. The next time I install the ATI app, I plan on using the 1.0e11 flops entry as that is what works fine on Collatz and Milkyway. |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
I read on the BOINC FAQ that the flop count is a reason for it to fail. IIRC it was mentioned on Lunatics that the flops wasn't necessary. In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Miep Send message Joined: 23 Jul 99 Posts: 2412 Credit: 351,996 RAC: 0 |
Flops are not necessary - the server will take care of flops once ten WUs have validated. The problem is the transition phase until that tenth validation comes through. Your client will start adjusting dcf accordning to the shorter runtimes seen. The server then uses the same dcf once the tenth validation goes through - expecting client dcf to be near one where it can be as low as oh .1. The result is that WUs are suddenly grossly underestimated - until the next gets finished and dcf shoots up again. In that window between the 1st fetch after 10th valid and next completed WU you can get massive overfetch. Having correct flops set in the app_info will keep dcf near 1 and avoid that particular problem. Without flops you better pay close attention to your valid count or you run with small cache until it's through. Carola ------- I'm multilingual - I can misunderstand people in several languages! |
aad Send message Joined: 3 Apr 99 Posts: 101 Credit: 204,131,099 RAC: 26 |
You could bypass that problem by using Freds Rescheduler. It's also fixes the -177 problem. |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
rescheduler isn't necessary for ATI cards. The App runs the VLAR WU's just fine. In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
rescheduler isn't necessary for ATI cards. The App runs the VLAR WU's just fine. The rescheduler doesn't only reschedule, it also (separately) fixes the -177 problem. Gruß, Gundolf |
©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.