Sandybridge OC mysteries

Associate
Joined
16 Feb 2011
Posts
320
Hi all,
Brand new build. Motherboard is a P8P67 PRO (B2 - doesn't bother me) and CPU is a 2500k. CPU cooler is a noctua d14.

Was trying to get a steady 4.5ghz overclock on the default bios that came with the mobo, but I had to set it 1.32v manual vcore for it to run stable. Prime 95 with no errors. However, idle voltage, as reported in cpuz and hwmonitor, was 1.32 at idle and 1.26 at 4.5ghz. Idle temps were about 27/28deg c and load was about 48 deg c. I was confused as to why vcore was higher at idle. Before the oc idle vcore was lower than load. I therefore decided to get the new bios to see if it would remedy the problem, but it hasn't, and at the same vcore prime 95 fails within 15 seconds. I am lost, and I dont want to start pushing over the top voltages for 4.5ghz when I don't believe it is necessary. Can anyone tell me what I am doing wrong, or provide me with settings that they know work well? Is what I am getting normal? Doesn't sound like it is. What should I be setting as LLC?

Thanks for your help.
 
Just found out my ram is now not running at the proper speed. 800mhz instead of 1600. Ive already entered the timings and voltages for it. OCZ platinum 1600 ddr3 2x2gb. Could this explain prime95 failing? As I havnt had a blue screen at any vcore above 1.27 yet. Cant say that I am enjoying sandybridge atm :(
 
800MHz = double pumped so it's actually 1600MHz, take it you're reading from CPU-Z?

Fair enough, I know theres some issues with the voltage readings on it, but it doesnt explain why voltage @1.6ghz was much lower than with my OC, and the fact that the vcore increased with the turbo enabled before my OC, rather than decrease. Is it really running at 1.32 v idle when the temps are below 30 deg c?
 
Fails intel burn test after about 120 seconds. Its using 1.26v at load even though Ive specified a manual vcore of 1.32. Makes no sense to me, even after hours of reading on the web. Some stuff about vdroop and voltage offset, but I am still just as lost.
 
it wll be runnng at that vcore at idle yes. however, if you download the realtemp beta for SB it wll give you a readout of the vid as well as the TDP of the cpu at the time.

when my 2600k speedsteps down to 1600mhz, the tdp drops to around 20w.
obvously setting an offset voltage wll allow you to make the voltage drop in parallel to the core speed, but this is tricky to get right.
 
it wll be runnng at that vcore at idle yes. however, if you download the realtemp beta for SB it wll give you a readout of the vid as well as the TDP of the cpu at the time.

when my 2600k speedsteps down to 1600mhz, the tdp drops to around 20w.
obvously setting an offset voltage wll allow you to make the voltage drop in parallel to the core speed, but this is tricky to get right.

I'll download it and have a look at the tdp. At vcore set to 1.32 Im failing prime95 at 4.5ghz. I dont think my 2500k is particularly good, but at least Ive got a decent air cooler. It does pass 3dmark 11 though, do you think that is a good test of an overclocked cpu if the most intensive thing you will be doing is playing games? What vcore and multi are you running day to day on your 2600k? I'll give the offset a go when ive got a couple of days free, sounds quite arduous.
 
I dunno man, that's a high vcore for that frequency, you'd have to be seriously unlucky not to get 4.5 stable at that voltage and that's a low temp. I had the same thing with my 2600k and I reinstalled Window7 and I'm solid @ 1.27v through SuperPi, 3Dmark looped for hours and Prime. Good luck :-(
 
not at all.
i gave up on the offset voltages and used manual for a bit but it seemed that my vcore/power consumption wasnt changing anyway

from my research, prime95 seems to max SB chips out more than other stress tests so that is what i have been using.

my current clocks were these...
4.9 [49X102] @ 1.42v
4.6 [46X100] @ 1.3v

i have swapped my board to a ud4 today as my asus pro died on me :(
straight in at 4.6 at 1.275v ;)
 
nah
anything up to 103 is ok

anything more gives voltage stability issues for me on all boards i have tried.
by that i mean that prime will run for ages, then you restart and a random bsod happens
 
Well when I say prime fails at 1.32v @4.5, thats the voltage I set in the bios, at that clock cpuz says 1.26 volts , which sounds more normal (but then again 1.32 at idle). When you say your vcores, is that in the bios or the actual observed figure?

On my first bios, 4.5ghz was stable at 1.32 and never went above 50deg c. So confused, 1.32 manual vcore does sound like a lot for 4.5ghz, compared to everyone elses sandybridge. Disappointing.
 
Hmmm, just run prime using the small FTT (mainly cpu) test rather than the blend (RAM also tested) and its working fine. Will take the volts lower and run the small FTT again. Beginning to suspect RAM could be the problem. Already set the timings and vram in the bios. Might try leaving the vram at auto. Dont fancy clocking it down to 1333mhz, as I didnt pay for that.
 
Update (not that anyone is following this thread anymore lol) :
Using small fft on prime95, 4.5 ghz overclock is stable at 1.27v with load temps of 53 deg c :) (havn't tried lower than 1.27 yet). Ram was the issue, failed memtest at 1600mhz. Will try fiddling with ram voltage tommorrow, or slacking the timings slightly.
 
seriously is all this necessery to get get the best oc from a sandybridge system? I thought they did it automatically. I set my pc up next week, my spec is in the signature, will i have to mess about with settings or will it oc automatically, im a noob at this stuff although i do know to download bios updates ect. I decided to get a sandybridge system for the new bios and easy auto oc
 
seriously is all this necessery to get get the best oc from a sandybridge system? I thought they did it automatically. I set my pc up next week, my spec is in the signature, will i have to mess about with settings or will it oc automatically, im a noob at this stuff although i do know to download bios updates ect. I decided to get a sandybridge system for the new bios and easy auto oc

To be honest mate it isn't really that hard. I was stress testing using tests that used the cpu and lots of ram, and because there was an issue with my ram it kept failing even at higher volts. When i tested only the cpu it worked perfectly. I should have realised earlier. If it wasnt for my faulty ram, I would have had a solid overclock in minutes. It really is so easy if you follow the right guides, that I wouldnt bother with autotune. An hour configuring and testing will mean a lower voltage, cooler and therefore a safer overclock in the long run. msg me if you have any problems. I had never oc'd before and I can look back and say it was easy. Have a go
 
Back
Top Bottom