Ok here it goes I got a amd 945 95w oc'ed to 3.75 w/ core voltage of 1.394 my issue is I ran prime95's max burn test and the log said it completed but sometime after that my pc BSOD I wanna say it was an event 41. I ran 3dmark06 w/ no errors, then windows 7 upgrade advisor no errors, then did 2 memtests with no errors. Im trying to figure out why Im getting a bsod AFTER my prime95 run not during. If it was during it would make more sense to me but the fact that its after makes me wonder. This is the only time that it happens too day to day stability its fine, nothing happens after gaming for hours or surfing the web so right now Im just confused as hell.
If you passed memtests, try running large and small FFT for several hours each as they test CPU more than a Blend which is partial RAM. Might want to try LinX/Intel Burn Test as it's a different method, seems to come up with errors faster. I have a totally diffent CPU so I can't really comment on the voltages, but I found VTT can help with stability a lot although for my CPU the VTT max voltage setting is only 1.21 while Vcore is 1.55.
I'm actually having an issue at 3.8ghz where I pass sever hours of Prime95 but fail within 5 iterations of LinX... pretty annoying!
Thanks for the response. I run all FTTs in place so there shouldnt be any of my ram being tested, as far as p95 goes see the thing is I passed all the tests and completed it but it BSOD after the test otherwise the log wouldnt have showed 128k completed, Im not at my pc right now but Im pretty sure thats as low as prime goes so that would have been the last test. So sometime after prime was done and when I got home (5hr window) it BSOD on me. I will be giving LinX a try when I get home and hopefully I would have to worry about this anymore.
The BSOD might have been an anomaly. I've heard that sometimes there's been issues with crashes due to Sleep mode. Since you passed the intel burn test and prime95, I'd say you're ok. Probably won't see another BSOD. Hopefully not, anyway. I've found most of my BSOD's caused by RAM, btw. If I change some settings on it just for fun or whatever, occasionally it'll BSOD but then, at the recommended settings it's fine.
ok just ran about 20min of occt and it crashed my pc lowered my fsb to 247 so that gives me an effective 3.705 oc. I re ran intel burn twice at 20 runs and no errors. Since I lowered my fsb though I also lowered my vcore to 1.356 and thats the run of occt that crashed. So right now im running at fsb247 and vcore 1.376 and running occt again. This seems like the best option to run since it found the instability and fast, not waiting overnight for prime95 to take its time.
Yea I like Intel Burn Test, although for my CPU I spent a good 5 hours tweaking voltages. At 3.6ghz I found that Vcore of 1.181-1.188 gave me the highest Gflops, anything higher or lower reduced it by anywhere from 1 to 3 GFlops. So, keeping the Vcore at that amount, I set the VTT low, 1.13V and slowly did testing. I passed 10 IBTs but then tried to do 10 more just after and failed after about 4. Couldn't get it to pass again so raised VTT a little... retry... again and again. Finally I thought I had it stable at 1.144V but running Prime95 over night, one core stopped. I raised it one notch this morning and have Prime95 going right now. I'll check it at lunch. Basically what I'm getting at tho is to ensure stability, use both tests.
IBT for quick checks but Prime95 for long term stability. Since you're getting 20 IBTs you're probably fine but might want to do an overnight P95 just to be sure.
Ok last night ran prime 95 overnight at 60min per test all ftts in place and when I got up I was happy to see that my pc was still running just fine all tests completed. So far I did 2x20 pass intel burn tests, ran occt for an hour, and did a prime95 overnight, all were successful. So Im running at 3.705 with a 1.376 vcore, idle temps are down 1-2c now around 29-30c and max load temps are 48c with all my fans on low. Im very pleased with the results I dont like my temps for my cpu >50c and I only lost about 1.5% out of my original oc that was at 3.75 which would max out at like 54-55c. So for a 23.5% oc I think I got some good results thanks for the input guys.
Just out of curiosity, what is your GFlops at? If that's a quad core, the max theoretical GFlops is 16x3.705 = 59.28. From what I've experienced and read, within 10% is pretty good. Best I've had was 8.5% off at 3.8ghz, but at 3.6ghz it's more like 9-9.5% oddly enough. Anyway, just curious
Just wondering where did you get the 16x from? If thats the multiplier then I would be at 15 so it would be 55.57. With the intel burn test I was getting about 49gflops. I havent messed with any other voltages besides the vcore I just left everything else on auto. I might go back later tonight and mess with the VTT see if that will change anything because I would like to get every last drop of performace out of it. And heres a question I would imagine that a cpu is a very efficient component using only the power that it needs so if I was to up my vcore to say 1.4 do you think that it would change temps by drawing more power or will it consume only what it needs?
I got the math from a discussion on LinX, basically you get 4 GFlops per core so 16 in a quad core, times your speed gives the max theoretical GFlops.
Your Vcore will raise the power consumption and temperatures. If you have stability, you might try lowering Vcore slightly and running the test. Since we're on totally different cores it might not be the same, but I found that there is a sweet spot for Vcore voltage that gave highest efficiency, anything higher or lower didn't effect the stability too much but it did reduce efficiency.
Ok I guess I'll give it a try raising it cause at 1.376 that was the lowest vcore that I attained stability at so I'll try raising it some and see if it improves my Gflop throughput. It could be though that we are using different platforms though. I was reading somewhere that equivelent clocked intel and amd cpus the amd would perform anywhere from 5-15% less than the intel cpu. Going with just 10% less if thats true it would be around 53.36 Gflops and you said being within 10% of that is pretty good and 10% of 53.36 is 48.02 and I know mine was around 49ish. God I think Im looking into this way to much...........Anyways since I know my lowest stable vcore Ill go home and mess with it some after work and see if I can get better efficiency out of it.