New Q6600 G0 Build Failing at Stock Speeds in Prime95

[adamsleath]

1 mem stick may not be like the others...
have you tried testing with 2 sticks?
perhaps u can alternate the sticks to establish possibly which one is the 'culprit'
...unless im barking up the wrong tree.

[226]

I have 2x2gig sticks so I still have to test them individually. So far I can't tell if this means it's a bad ram stick or if turning the ram speed down only made it stable because there is a voltage problem with the cpu or mobo. I also tried adjusting the ram voltage between 1.8-2.1 but it wouldn't boot when I got up to 2.1. Memtest didn't find any problems but maybe Prime95 brought it out because it's a more stressful test.

[SLi_dog]

Does this suggest that the memory could be faulty because memtest did pass?

memtest stable definately does not mean Windows stable.

I've had RAM that passed hours of memtest just to BSOD in Windows and error in Prime95 within a second.


What vdimm are your sticks running at and is that what they're rated for?
I'd see what the maximum warrantied vdimm is on your RAM and if there's some head room there, slowly keep bumping up the voltage until it runs stable

[226]

memtest stable definately does not mean Windows stable.

I've had RAM that passed hours of memtest just to BSOD in Windows and error in Prime95 within a second.


What vdimm are your sticks running at and is that what they're rated for?
I'd see what the maximum warrantied vdimm is on your RAM and if there's some head room there, slowly keep bumping up the voltage until it runs stable

My ram is rated between 1.8-1.9 I tried voltages up to 2.1 but it wouldn't boot with it that high.

I'm pretty sure I have a ram problem now. I ran both sticks by themselves with prime95 and everything at stock settings. The first stick failed after 50 minutes and the second stick went for 5 hours without a problem. I'm going to stress test both sticks again to be sure.

Running the stick that appears to be good I bumped some voltages up a little and was able to get into windows oc'ed at 9x312 but I didn't test it.

[Leeghoofd]

If I read up on other Q6600 with Abit IP35 mobo's there seem sto be a very high drop in voltages when bundled with quadcores ( bad PWM design ? ), probably you will need to set it up way higher than 0.1 extra on the stock Vcore volts....

Secondly have you tried both sticks in the same dimm slot ? or just removed one ? tried other dimm slots ?

[226]

I tried both sticks in the same slot. I just tested each stick individually again and they both passed but that was with the large FFT test and I found that when I had both sticks in they would fail much quicker with the blend test which is harder on the ram. Now I'm running the blend test on each stick individually to see if there is a bad stick but they certainly fail when using both sticks of ram. The stick that I'm running the blend test on now has gone for 6 hours without a problem so I'll see if the other stick can do the same.

Since it fails on the test that is harder on ram and underclocking the ram seemed to take away the problem I still think I have a memory issue. I tried running at 2.0v for the ddr2 and it seemed to fail quicker than it would at 1.8. Bumping up the vcore and mch seemed to help prime95 last longer before failing.

[226]

I'm getting mixed results testing the ram but it does fail quickest when using both sticks. When testing with one stick that I suspected was unstable by itself the system restarted the other day but then I ran it again and it went for 6 hours without a problem. I tested the other stick and it hasn't failed yet when run individually yet.

Does this sound like bad ram?

The things that point to it being a memory problem are:
-The test that is mostly intensive on memory is more likely to cause failure than the test that is harder on CPU and voltages.
-Underclocking the ram seemed to get rid of the problem.
-One failed when run by itself but I got mixed results on this one and the other stick hasn't had any problems.
-Running 2 sticks causes failure the quickest which shows that the problem is related to memory or it could suggest a different problem where the system can't handle the voltage from using both sticks at once. Also increasing the vcore may have helped the system last longer during the tests but this may not be true because I might have been running the large FFT test after increasing the vcore which is much less likely to cause problems.

[Leeghoofd]

I suspect more it's a NB/mobo/BIos problem, if both dimms pass the test when installed alone but fails when used in pair it means the NB memory controller is stressed out. Also these big dimms don't like too much voltage and trying to run thme at more than 2 volts without added cooling could ruin them...

Have you tried running them in single channel modus ? does that work at stock speeds ?

[226]

I haven't tried single channel mode because I didn't see that option in bios but I will try that next. The stock voltage for the memory is 1.8 and running it at 2.0 seemed to cause it to fail faster so I'm not sure what failing faster at a higher voltage implies. So would a voltage increase stress the ram or NB memory controller more?

One stick had failed 2 times running by itself but I'm not sure now because it didn't have any trouble the last time I tested it.

[Leeghoofd]

It isn't an option the bios , it's the way you place the dimms eg 2 dimms next to each other in Abit mobo is dual channel ( at least it was in the 939 days) if you place the dimms one slot apart it becomes single channel or the other way around in case it' s no longer valid...
with Asus you have to palce the dimms in the same colour to get dual channel...

[226]

I see what you're saying I'll do that next. When I get home I'll also try turning up the memory controller voltage more than I did before because if dual channel mode is causing problems that is probably where the issue would be. I also think that I couldn't boot at 2.1v for the ddr2 so I'll see if that's true for both sticks individually.

[226]

I think I figured out the problem Prime95 has been running for over 10 hours with a very small underclock of the ram and the processor oc'ed to about 3.0. The IP35 PRO had it set to a small OC of 408 which may have been causing problems but it seems to be running fine at 396.

Here is a screen shot of it still running:

[Leeghoofd]

Well the ram should do 400mhz not 396... so 333FSB and 400 ram still doesn't work ?

[226]

It wouldn't let me get that exactly with the ratios with the fsb that I had. I did try setting it to 400 with the cpu 8x400 but I got a blue screen in Vista after running Prime95 with those settings but that may have been related to the fsb and vcore. I'm at about 3ghz with the cpu so and it's a speed that I'm happy enough with as long as these settings are stable I think I'll leave it this way.

Having it at 396 doesn't seem like a significant enough of an underclock to be a problem. I was getting the problem at just 408 so a underclocking it a little might be good for stability. I'm guessing that this won't OC at all because 2x2gig sticks is stressful on the memory controller.

[Leeghoofd]

Try 333 FSB on the CPU then you will get 800mhz on the ram exactly... also might try 1.3 as CPU VTT just for stability reasons

[226]

Ok, I'll try that next. I wasn't planning on oc'ing the ram anyways so I'm not going to try and get any higher than 800 since the failure point is very close and I doubt that a few mhz would make much of a difference.

[Leeghoofd]

Well this is only 30mhz more on the CPU and 4mhz on the ram but at least the ram runs at it's rated speed at which they should deliver !!!

Let me know if they run man, I'm kindda interested in going 4gigs too...

[226]

I should know tommorrow because I'm currently running a test to see if it's still stable with the vcore turned down a little from 1.37. 4 gigs does seem worth it though especially for Vista. When running the blend test in Prime95 the system still goes fast with 4 gigs installed but when I only had a 2gig stick in it was almost completely unusable.

[226]

It ran fine at 9x333 and 800 on the blend test for 8 1/2 hours so that seems to be stable but I might set it back to 330 for long term stability. Now all I have to find out is the lowest voltage I can use these settings at because I had the vcore at 1.37 and I'm getting max temps of about 58 but it might be a little higher with the large FFT test.

[smashhell]

Well this is only 30mhz more on the CPU and 4mhz on the ram but at least the ram runs at it's rated speed at which they should deliver !!!

Let me know if they run man, I'm kindda interested in going 4gigs too...

I have a really noob question here.

What if the Ram are too set higher than rated speed ?

Like DDR2-800 but running it 832 ?

Will there be a problem ?

Thanks in advance !

[Leeghoofd]

Well all depends on your ram, I never run ram over it's rated speed but most users get easily at the stock voltage 10% more out of it... 832mhz instead of 800 won't be that big a deal no, will work fine mate , but if you got 4 dimms I can highly advise a ram cooler... what FSB are you running then ?

[smashhell]

Well all depends on your ram, I never run ram over it's rated speed but most users get easily at the stock voltage 10% more out of it... 832mhz instead of 800 won't be that big a deal no, will work fine mate , but if you got 4 dimms I can highly advise a ram cooler... what FSB are you running then ?

Ahh .. Thanks or the quick answer.

I am running 3.2Ghz (356x9), DDR-2 800 4Gb.

With ram running at 1:1 right now, which is 712 Mhz.

I wanted to make it 1:1.2 but scared going above spec.

Shouldn't be a problem right?

[Zucker2k]

@226. Here's what I'll do; bump your nb voltage by about .25, and vcore by .25 for your initial overclock. It should work. Goodluck.