On my desktop at home I'm running bigadv and a GTX460 and its used all the time for general stuff (web surfing, email, word, excel etc) on a 6900 it drops to around 33:10 per frame or around 32K PPD. Thats running -smp 7 with one HT thread for the GPU.
Folding@Home News 12th November - 19th November 2010
- Thread starter dekez
- Start date
Soldato
What CPU/overclock do you have?
I am getting 33K PPD, give or take, with moderate browsing/office stuff on all 8 cores on an i7 920 at 4GHz. Going to OC my server this evening and see if I can get that running bigadv. Server gets somewhat busy overnight, so will be interesting to see if it can crunch through the WUs fast enough!
I am getting 33K PPD, give or take, with moderate browsing/office stuff on all 8 cores on an i7 920 at 4GHz. Going to OC my server this evening and see if I can get that running bigadv. Server gets somewhat busy overnight, so will be interesting to see if it can crunch through the WUs fast enough!
Same CPU, its almost 4.2Ghz, seems to be the most stable its ever been, just crunching through with no errors, for a while I couldn't finish a bigadv but then I tweaked it a bit changed the multi and memory settings and now its sweet as a nut.
My server (currently a 920@4, but soon to be meatier) runs VMware with three VM's a Domain Controller, Exchange server and SQL/Web server. Its running -smp 8 but gets busy off and on. Has 12GB memory too so harder to OC. But stable as well at 4Ghz solid.
My server (currently a 920@4, but soon to be meatier) runs VMware with three VM's a Domain Controller, Exchange server and SQL/Web server. Its running -smp 8 but gets busy off and on. Has 12GB memory too so harder to OC. But stable as well at 4Ghz solid.
Soldato
Ahh excellent. I know my CPU can do 4.2GHz, but I get the Grey Screen of Death when it runs at that speed. Seems something causes my PCI-E ports or the 5870 to throw a wobbler at 4.2! 
Still, have a server to OC soon!
Still, have a server to OC soon!
Soldato
I assume you've tried locking the PCI-E clock to 100MHz? Also read somewhere that underclocking the PCI-E clock helps on certain motherboards with high CPU clocks.
Soldato
Tried both, nothing worked!
I spent a long time trying to work it out. At first I assumed it was a dodgy 5870 as it was a graphics issue. But the minute I lowered my OC the GSoD never came back. I might push it back to 4.2 and see what happens - maybe 8 months running at 3.8 has burned my CPU in a little or some other voodoo nonsense!
I spent a long time trying to work it out. At first I assumed it was a dodgy 5870 as it was a graphics issue. But the minute I lowered my OC the GSoD never came back. I might push it back to 4.2 and see what happens - maybe 8 months running at 3.8 has burned my CPU in a little or some other voodoo nonsense!
Soldato
Just cranked up my server's i7 to 4GHz [well, 3999MHz but shh!] and got it running bigadv. Couldn't be arsed doing extensive testing. Did a few runs of IntelBurnTest on Very High and passed, that'll do me! *waits for explosion*
I have noticed in Core Temp that I am not quite hitting 100% load and the multi is bouncing around all over the place. I might have to turn off speed step, though I have almost identical settings on my main PC and it sits at a solid 100% load and the multi stays at x20. Hmm.
*waits for explosion*I have noticed in Core Temp that I am not quite hitting 100% load and the multi is bouncing around all over the place. I might have to turn off speed step, though I have almost identical settings on my main PC and it sits at a solid 100% load and the multi stays at x20. Hmm.
Soldato
I find LinX with max memory hard to beat as a stability tester, couple of hours of testing are well worth it when you could easily lose a BigAdv to a slightly unstable clock.
Soldato
Finally got through the first frame on a 2684. 45 minutes! That works out at 3 days total to finish the WU at 19.8K PPD. That seems a bit low for an i7 at 4GHz ish. I think I will have to tell speed step to sod off and see if the CPU usage settles down at 100%, instead of bouncing around.
Any ideas why it keeps dropping as low as 80% on all cores?
On the plus side with all 3 clients I am on 64K PPD according to HFM.
Any ideas why it keeps dropping as low as 80% on all cores?
On the plus side with all 3 clients I am on 64K PPD according to HFM.
Soldato
i7 920 @ 3999 [190x21]
Gigabyte X58A-UD3R
GeIL Black Dragon PC3-10660 C7 1666MHz
I think the load is what is causing it to step down. When it hits 100%, the multi goes back up to 21. So, for whatever reason the SMP client isn't pushing the CPU hard enough. It is running -smp 7 to give the GPU client a bit of breathing room.
I doubt it is overheat protection, it is barely budged past 73C, the hottest core is 76C. My desktop runs at 85+ if the window is closed and no twitchy multi.
Gigabyte X58A-UD3R
GeIL Black Dragon PC3-10660 C7 1666MHz
I think the load is what is causing it to step down. When it hits 100%, the multi goes back up to 21. So, for whatever reason the SMP client isn't pushing the CPU hard enough. It is running -smp 7 to give the GPU client a bit of breathing room.
I doubt it is overheat protection, it is barely budged past 73C, the hottest core is 76C. My desktop runs at 85+ if the window is closed and no twitchy multi.
Finally got through the first frame on a 2684. 45 minutes! That works out at 3 days total to finish the WU at 19.8K PPD. That seems a bit low for an i7 at 4GHz ish. I think I will have to tell speed step to sod off and see if the CPU usage settles down at 100%, instead of bouncing around.
Any ideas why it keeps dropping as low as 80% on all cores?
On the plus side with all 3 clients I am on 64K PPD according to HFM.
2684s are crap bigadv WUs. ~20k sounds about right for that setup. I get roughly the same for a 920 @ 4GHz. Same rig gets about 30k for a 2686 or 31k for a 6900.
Soldato
I think SpeedStep was confused rather than there being an issue with my OC.
I restarted the client with all 8 cores running plus my GPU. GPU TPF hasn't changed and neither has the bigadv frame time! :/
However, all cores are running at 100% and the multi isn't jumping around. So my theory is the "spare" HT core I left for the GPU client was bouncing from core to core, pulling down the average CPU usage and messing with SpeedStep.
Doesn't help if 2684s are crap in the first place!
I restarted the client with all 8 cores running plus my GPU. GPU TPF hasn't changed and neither has the bigadv frame time! :/
However, all cores are running at 100% and the multi isn't jumping around. So my theory is the "spare" HT core I left for the GPU client was bouncing from core to core, pulling down the average CPU usage and messing with SpeedStep.
Doesn't help if 2684s are crap in the first place!
Soldato
My PPD hasn't changed for either the SMP or GPU clients, so I will just leave all 8 cores crunching. At 20% on my server now and still at 20K PPD. So looks like what my server gets up to at night doesn't adversely affect the WUage.
My PPD on EOC has taken a slight hit though. Still, not to worry, 72K points coming my way at around midnight!
My PPD on EOC has taken a slight hit though. Still, not to worry, 72K points coming my way at around midnight!
Soldato
Well so far so good, running bigadv and have not been plagued with the bsod like i was last time i tried. Picking up some higher value wu's but not as high as some 
Soldato
Right ho sah! will implement that this afternoon