Message boards :
Number crunching :
APv6.01 via Lunatics 0.40 installer has errored out all APv6 tasks
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
hbomber Send message Joined: 2 May 01 Posts: 437 Credit: 50,852,854 RAC: 0 |
[quote] 8 tasks simultaneously? |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
thats the idea In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
[quote] Sure, yes but i only had 5 max yet. 3 cores were running MBs. With each crime and every kindness we birth our future. |
hbomber Send message Joined: 2 May 01 Posts: 437 Credit: 50,852,854 RAC: 0 |
thats the idea It might be not an idea, as he uses GPU also, he may have left some core to serve GPU. That's why, I ask him. i7-3820 4.65 GHz, running 4 cores only - 16-19K seconds per AP 6.01 WU, r557, but OS is not AVX aware. Gonna run 8 threads now, finally got able to run whole 8 tasks at once. |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
As nice as the AP's run on the Bulldozer, I think I'll keep them running on my GPU only. 1-3 hours vs 14-15 hours. GPU wins In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
As nice as the AP's run on the Bulldozer, I think I'll keep them running on my GPU only. 1-3 hours vs 14-15 hours. GPU wins Yes, but now its credit efficient. Taking 5 times longer than MB but getting 6 times credits. With each crime and every kindness we birth our future. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
As nice as the AP's run on the Bulldozer, I think I'll keep them running on my GPU only. 1-3 hours vs 14-15 hours. GPU wins LoL, GPU wins vs other CPUs too :P For example, my Q9450 doing 4 AP in ~12h and HD6950 GPU (especially when healed from ATI low usage bug) spends ~1.5h doing 3 tasks at once. With bug in place it spends ~3h doing 3 tasks at once so let say 3 tasks in each 3 hours vs 4 tasks in each 12 hours - GPU wins clearly :) |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
I noticed a few months ago with r409 on v505 that if I ran all six cores of my 6100, the times increased on all tasks by ~33%. So I just ended up using <ncpus>3</ncpus> in my cc_config and they seem to run about as fast as I can get them to go on this hardware. That shared FPU deal really hurts when two cores in a pair try to do floating-point stuff at the same time. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I have different experience. I found out i get much better performnce running on all cores on my 8150. Even you can´t compare r409 with r555 or r557. With each crime and every kindness we birth our future. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
I have different experience. That is a good point. I suppose I can see what happens when all six run now that there is AVX. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
I have different experience. Gave it a shot for a little while. 3 at a time yields a median run time of 40,450 seconds according to my spreadsheet, which translates to roughly 8.8889%/hour. Running all six cores yields about 6.5%/hour. That does in fact correlate to the roughly 33% slowdown that I observed in the battery of benchmarks I threw at this machine on day one. Going back to 3 cores. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Karsten Vinding Send message Joined: 18 May 99 Posts: 239 Credit: 25,201,931 RAC: 11 |
3*8,9=26,7 6*6,5=39 Much more throughput with six cores. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
From what I read this AMD shared FP resembles Intel's HT. It was shown that in most cases enabled HT gives better throughput thouh each single task runs slower. I checked HT on Atom netbook and observed some speedup too. Same here. Each individual task runs slower but whole system throughput is better. |
cliff Send message Joined: 16 Dec 07 Posts: 625 Credit: 3,590,440 RAC: 0 |
My 2nd rig is an AMD Phenom2 4core, its got 1 validated AP6 WU, about 3secs:-) of course there was 100% blanking as well.. On a 6core bulldozer its between 13.5 hrs and 18 hrs.. I expect its blanking dependant given the 2nd rigs result.. All others are still awaiting validation. [edit] 1 from rig 1 validated now at 51,208.42secs running time.. Cheers, Cliff, Been there, Done that, Still no damm T shirt! |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Are you aware you are running without AVX on your FX host ? What clock speed are you at ? With each crime and every kindness we birth our future. |
cliff Send message Joined: 16 Dec 07 Posts: 625 Credit: 3,590,440 RAC: 0 |
Nope I wasnt aware of not running AVX on my host, particularly since siv reports a curent AP WU is being processed under ap6.01R557 SSE2_331_AVX.. Cpu is currently running all cores at 3.81Ghz.. How does one determine if AVX is on/off? Cliff, Been there, Done that, Still no damm T shirt! |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Nope I wasnt aware of not running AVX on my host, particularly since siv reports a curent AP WU is being processed under ap6.01R557 SSE2_331_AVX.. That only means the app is AVX capable. You need service pack 1 to enable AVX. Would be interesting because i tested this app on win 8 CP. I´m running without AVX as well on my Vista host. With each crime and every kindness we birth our future. |
cliff Send message Joined: 16 Dec 07 Posts: 625 Credit: 3,590,440 RAC: 0 |
Ahh, that a pita, because no matter how I try I simply cannot get SP1 to install on this W7, and it IS a legit copy.. Its daft 'cos I have W7 [from same dvd [oem] on 2 other drives on this rig as backup in case the main copy falls over, and SP1 installs fine on those.. Trouble is both the other are on small SSD's, I've got about 56Gig free on one though.. But I also have a fairly large number of WU in hand.. Take a good while to work through then even with NNT. And I doubt I can just shut boinc down and copy all the files over to the SSD and reinstall boinc and not have a problem with the outstanding WU's:-( Cheers, Cliff, Been there, Done that, Still no damm T shirt! |
hbomber Send message Joined: 2 May 01 Posts: 437 Credit: 50,852,854 RAC: 0 |
Just finished several, calculated with 8 active logical cores. 29-31K seconds per task. Impressive, but still MB pays better. Gotta run with AVX to see how it will be. Btw, AP with this application loads my CPU heavier, temperature increased 2-3 degrees compared to SSE x64 MB application. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
3*8,9=26,7 From what I read this AMD shared FP resembles Intel's HT. It was shown that in most cases enabled HT gives better throughput thouh each single task runs slower. That is true, but I prefer lower times, since they are also a bit more consistent. When I have all six cores loaded, the durations vary quite a bit, independent of % blanked. When I only run three cores, the times are consistent--only adding about 2,000 seconds for 90% blanking. Bottom line: I'll run it my way, you can run it yours. :p Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
©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.