How common is it for a quad core to have the cores not finishing the tests at exactly the same time? I started an instance of Prime95 25.4 to test the stability of an overclock on my new Q9550/Gigabyte EP45-UD3P system (3.4ghz, stock vcore) and it ran for 8 hours with no errors. However, I noticed that after that time Cores 0 and 2 were about 5 tests behind cores 1 and 3. Incidentally, cores 0 and 2 were also 3 degrees cooler than cores 1 and 3.
This is normally a sighn of lack of voltage, go to the Bios and add an extra .0125V to the Vcore, this should be enough for the 4 cores to keep up.
If this doesnt help much then continue to add extra Vcore untill it sorts itself out.
there are different stages to the Vcore shortage, 1: No Boot, 2: Post but no Boot to windows, 3:Boot to Windows but crashes, 4: Boot to Windows, but crashes in Prime95, 5: Prime runs for a period but 1 core may fail, 6: prime runs but 1 or more cores will lag behind.
Windows has to do some things in the background of the computer and these will take time away from number crunching leading to some cores getting out of sync. One big example is that processing time has to be found to display the results and move the mouse cursor.
If the voltage was to low you would end up with errors reported in the calculations where binary bits flip or are mis-read, at the worst case you will get a crash.
when using prime95, the usage of the cores is maxed out, meaning that the windows ops are already being taken in to consideration.
when O\C my Q6600 go to 3.5 i had the same minor problems (although no where neer as bad), a slight change in Vcore voltage corrected the problem, cores will fall behind before they fail, if it dont crash at the 8 hour point then it is fine, although i added more Vcore to make sure that it was as stable as poss.
it may not be the voltages, but this was my experience on my system and also on several other systems i have built and O\C for friends and family, trying the Vcore up just 1 notch may or may not help, although it will not hurt to try.
There will ALWAYS be a difference between when the cores finish. However, a difference that large means you do need to boost the volts. It is normal for a core to fall behind on 1 round in 1 test, but not 5.
cool, thanks guys. I'll try bumping the volts. This is my first Core 2 platform, so I was a little confused and wasn't sure that was normal or not. One thing for sure though, this thing overclocks so much easier than my AMD 3800x2!
I tried bumping up Vcore a few times, ending up at 1.2v actual (1.2325 I think in Bios) and ran Prime for an hour each time to gauge if they were in sync or not. It seemed different cores would be falling behind every time, and the problem was not really consistent, worse some times than others. The system seems rock solid in all other aspects of computing, games, audio production (particularly CPU intensive), etc.
At this voltage, I decided to play around a bit, and took the machine up to 450fsb, resulting in 3.8Ghz and see if I could get a failure out of Prime95. I ran it overnight, about 11 hours and 20 minutes, no failure, but the cores were off by a spread of about 4 tests still. So it seems no matter what I do to it, it's Prime stable but not in sync.
Should I keep going with the Vcore bumps? The temps hit 60C on the hottest cores on RealTemp last night so I think if I have to keep bumping voltage, I'll have to start bringing down the clock to keep things somewhat cool. Or since it seems stable in everything else I should just let it fly?
Its not a problem. Remmeber you ran the test for 8 hours, thats a long time. 5 test behind in 8 hours is nothing to write home about, your system is good to go. There is no reason to go bumping up voltages for no reason, not to mention it wont help at all. Like Randomizer said, if it was a voltage issue cores would have started failing or just general shutdown/reboots.