After hours and hours of reading forum posts and guides here and yonder, I am forced to cry out for help... so "HELP!!" lol! I have hit a wall at 3.9 Ghz and can't seem to break through. I ran OCCT for an hour with the following settings:
CPU clock ratio: 14
CPU Northbridge Freq: x7 (1995 Mhz)
Core Performance Boost: Disabled
CPU Freq: 285 Mhz (x14=3990 Mhz)
HT Link Freq: x7 (1995 Mhz)
NB Voltage Control: 1.15
Mem Clock: x5.33 (1519 Mhz)
DDR3 Timings: 7/7/7/20
DRAM volts: 1.510v
CPU volts: 1.5000
Temps in OCCT- 40-42 C
When I bump the CPU freq up to 286, temps go up to around 44-46 C and OCCT takes an error on a random core at around the 30 to 40 minute mark. I tried bumping the NB voltage control up to 1.17 but it still failed at 32 minutes. I have read so many different posts and guides that my brain hurts! What other tweaks should I try to get it stable at 4G? I am soooo close, I just need a little nudge!
Just a couple other things:
Is the "NB Voltage Control" the one I want to tweak or is it "CPU NB VID Control?"
Should I use some other program to stress test besides OCCT? There seems to be a bazillion opinions out there- Everest, LinX, 3DMark, Prime95 etc.
And last but not least, Temperatures:
Is the "core temp" in OCCT the temperature I want to monitor? Is it accurate?? I can't make heads or tails of SpeedFan and I understand there's a difference between the actual core temp and the socket temp. I read somewhere that the 1055t's can be 10 degrees off but I could only find one post about it and it wasn't backed up with anything concrete. I have been watching mine to ensure they don't go over 45 in OCCT which should keep me safe even if they are off a bit. I do suspect them to be off, as my idle temp has been pretty steady at 22 degrees C which is about the temp in the room here. I don't think my rig cools that well but who knows. I have a pretty good tunnel of air moving through the front, over the cooler and out the back. I moved all my drives to the bottom and tidied up my wiring just for this reason.
Picture quality might be a little lame but its the best I can do on my company issued Blackberry (aka Dingleberry!)
I'm still stuck in the stone age hardware wise, (2001 pentium 4 with locked bios) but, I do have a few suggestions. Not knowing the specs of the ram, the errors could be caused by ram errors. I would set it at 3.9ghz and run Memtest for a few hours and see if any errors come up. If so, increase your ram voltage or drop the cpu:ram link speed and try again. From my experience, AMD am3 cpus are real picky with 7-7-7 timings on the ram, drop them down a notch and run another error test.
I'm still stuck in the stone age hardware wise, (2001 pentium 4 with locked bios) but, I do have a few suggestions. Not knowing the specks of the ram, the errors that may be happening could be caused by ram errors. I would set it at 3.9ghz and run Memtest for a few hours and see if any errors come up. If so, increase your ram voltage or drop the cpu:ram link speed and try again. From my experience, AMD AM3 cpus are real picky with 7-7-7 timings on the ram, drop them down a notch and run another error test.
Sounds like a good place to start. I'll let ya know what happens!
This will allow you to monitor core temps while adjusting voltages and frequencies. It also has a quick stability test which stresses all cores at 100%, so you can monitor stability, temps, and adjust all in one program.
I use this to dial in everything, then exit windows and in BIOS I set everything as it was in windows.
Success! I loosened the memory timings to 9-9-9-24 and bumped the HT link volts, DRAM volts and NB Volts up just one notch each. If I can pass LinX or Prime95 I am going to start backing them down 1 at a time and find out where the bottom is, no sense over=volting stuff that I don't need too!
I am still curious about the temps... Even it they are 10 degrees off or so I should be OK with a max of 45 degrees under full load dontcha think?
Well, 45 degrees on a phenom ii x6 is not bad under full load, especially overclocked past 4ghz. As long as the temps don't exceed the mid 50's under full load I wouldn't worry about it, even if the sensors are a bit flaky.
If you can figure out which voltage needed to be raised you may be able to increase it slightly on its own and tighten the timings back into the 8's for the ram. Keep me posted on which voltages needed to be changed and the tightest timings you can achieve.
Oh, for stability if the temps flatten out after an hour and no errors come up during 100% load I wouldn't worry about general desktop usage. If you are doing distributed computing or running a server that has to very reliable I would run it for a full day.
In my opinion, one hour is good. Others may disagree. I think it is likely you'll see errors w/in the first hour. The true test though would be is it stable when playing your most demanding games/apps?