Intel Burn Test, Q6600, and Bizarro Land

With the modulation boxes ticked and throttlestop turned on it will try to overide any throttling. To monitor only, uncheck the modulation boxes, the multiplier box and set "Turn off". Anyway, you can see something is going on with core 0 and core 1 by the fact that those cores are only running about halve the time. (C0%)

Seems your Tjmax should be 90C but HWMonitor is using 100C. This would mean your worse temperature of ~78C is infact ~68C. The DTS reading still has quite a bit of headroom so it doesn't seem like your overworking your CPU.

It's a bit weird it's happening on only 2 cores. Are you able to measure NB temperature? Have a look in the event log for any power messages.
 
TJMax for a Q6600 - G0 is 100C

The misleading information that Intel released (90C) is TJ Target which has nothing to do with actual TJMax.

A Core 2 Quad consists internally of 2 separate Core 2 Duo CPUs so it's not too unusual for 2 of the 4 cores to be doing something different than the other 2 cores. It could be either a bios or a defective CPU problem.
 
Captain Picard said:
About an hour ago, I switched the fan to pulling air through the radiator into the case. The temperature of all the components in the case, CPU included, rose significantly. I should have figured that this would happen, as my PC sits in a desk cubby, and the only direction to draw fresh air is through the front of the case, not the rear. I've put the fan back on the radiator, blowing air through it and out the rear again. Yes, the pump is plugged in.

I use core temp to monitor CPU temperature.

The friend I bought the CPU from said to me a little while back, "It's not necessarily a worthless OC if there's an error in IBT." However, I don't like the idea of my CPU delivering any sort of error or abnormal behavior regardless of the program in question.

I'm not sure what else to try at the moment, I might slap another 120mm fan onto the radiator and see what happens to my temperatures.

If anyone comes up with something, I'd be glad to hear it. Thanks for checking it out so far.
Click to expand...

Core Temp is not accurate for the Q6600, at least in my case it was not. I use Real Temp in conjunction with a thermistor I build into the heatsink the hardOCP way.
 
So, I went to bed all those hours ago, but now I'm up again.

It turns out the bios I was running, F9c, was a beta bios. Version F9 is available at Gigabytes website, so I flashed to that one. I'm going to fire up IBT with RealTemp, and Throttlestop to see if anything has changed. 9 x 376 @ 1.350V.

Update: I ran two runs (Very High stress level, 4096 MB) of IBT at the above speed and voltage. The second run produced an error, but no cores throttled during the test, and the time for each run was nearly the same. I haven't calibrated Real Temp, but the default TJMax was 100*C, and it reported the same temperatures as core temp. I suspect the previous bios was the culprit, but I need to give the CPU more voltage so it can produce errorless runs. Previously, the throttling would happen about 40 or 50 seconds into the first run, and 20 into the subsequent runs as the CPU began to "warm up".

Update 2: I bumped the CPU voltage up to 1.3875v @ 9 x 378, and did three runs of IBT. The third run produced an error, but scores were consistent between the three runs, and no throttling was occuring. With my current cooling/airflow situation however, I don't have much headroom left for voltage and MHz. I hit 83*C during these runs.

I'm almost 100% certain it was a bios bug that was causing the throttling. Thank you everyone for reading and posting your ideas. I really don't want to futz around with switching around the fan/radiator setup or re-mounting the waterblock, but I may do that at some point to see if I can get better temperatures. The stock Corsair fan is quiet at 1200RPM or under, but pretty loud at full speed 1600RPM.
 
Last edited:
On a 65nm Core 2 based CPU, you are going to find it very difficult (impossible) to run reliably at 3400 MHz with the core temperature over 80C. I've done a significant amount of testing when living on the edge like this. You can run a Core 2 Quad hot or fast reliably but not both at the same time.

On the 45nm Core 2 CPUs, running reliably at close to TJMax is not a problem when overclocked to 3600 MHz but when you push further to 4000 MHz, you'll need to leave about 35C of headroom before TJMax. Depending on voltage requirements, you'll need to keep the core temperature down around 65C to 70C to remain Prime stable at 4GHz. The 65nm CPUs are the same but their range is lower. More like 3200 MHz to 3600 MHz. As the speed goes up in this range, the core temperature must go down or you'll lose stability.

Here's a fun E8400 test I did. A mild overclock and full core temperature for 3 hours of Prime95. Turning the CPU heatsink fan off really helped get the temps up there. :)

It bounced off the thermal throttle for 3 hours but never skipped a beat. Zero errors. At 4000 MHz, it doesn't run stable for long as the temps go up. At 70C this CPU is not stable but if I move my computer to the basement and get the core temp down about 5C, then it is fine. Just a matter of finding what temperature vs speed your CPU is happy running at.



When using ThrottleStop you can change the DTS column to core temperature by adding this option to the ThrottleStop.ini configuration file:

TJMax=100

You can also monitor GPU temperatures by adding:

GPU=1

for Nvidia cards or adding:

GPU=2

for ATI video cards.
 
Last edited:
Those temps are very bad for an H50 and only 1.3875v. There might be something really wrong with your setup.

With my Q6600, I do in fact require the exact same voltage for 9x378 = 3.4GHz and I just tried 3 runs on your settings and on air, I didn't hit 65° C and my CPU has 20 watts more TDP than yours.

Have you tried reseating the water block? What are you using for thermal paste and how did you apply it? The first time I installed my heatsink I used too much paste and temps were 10° higher than they are now after I cleaned it up and did it again.
 
Spare-Flair said:
Those temps are very bad for an H50 and only 1.3875v. There might be something really wrong with your setup.

With my Q6600, I do in fact require the exact same voltage for 9x378 = 3.4GHz and I just tried 3 runs on your settings and on air, I didn't hit 65° C and my CPU has 20 watts more TDP than yours.

Have you tried reseating the water block? What are you using for thermal paste and how did you apply it? The first time I installed my heatsink I used too much paste and temps were 10° higher than they are now after I cleaned it up and did it again.
Click to expand...

i agree. i have a b3 with the same cooler as OP and mine isnt nearly as hot as his. i really think something is:

wrong with the cooler
not properly installed
really poor ventilation in the case
 
I've only got one exhaust fan in the case, and it's doing double duty as the radiator fan. As for thermal paste, I just used the pre-applied paste on the water block. The block is mounted so that the corsair logo is upright, waves parallel with the bottom edge of the CPU socket. The fan is pushing air through the radiator and out the back of the case. My case is also sitting in a desk cubby. The front of the chassis and bezel have plenty of perforations for drawing air, but there are no fans at the front pulling air in.
 
Captain Picard said:
I've only got one exhaust fan in the case, and it's doing double duty as the radiator fan. As for thermal paste, I just used the pre-applied paste on the water block. The block is mounted so that the corsair logo is upright, waves parallel with the bottom edge of the CPU socket. The fan is pushing air through the radiator and out the back of the case. My case is also sitting in a desk cubby. The front of the chassis and bezel have plenty of perforations for drawing air, but there are no fans at the front pulling air in.
Click to expand...

One fan for the entire system is a complete no-no. It's working extremely hard and half as efficient because there is no fan feeding air into the system for it to use. Why don't you install more fans? Combined with the desk cubby, that whole environment is very unhealthy for everything in your case.
 
unclewebb said:
The misleading information that Intel released (90C) is TJ Target which has nothing to do with actual TJMax.
Click to expand...
Was it really misleading or was it mis-interpreted?

Isn't the "Tj target" supposed to be the minimum temperature for Tjmax. As I see it as Intels processing and calibration is not perfect, ie not all CPUs are equal, then it can be expected most of those CPUs will have their Tjmax happen a little higher than the minimum expected Tjmax. So if a Tj target is quoted as 90C then it would not be expected for the CPU to hit Tjmax below that temperature and it would be likely that Tjmax, when the DTS reaches 0, could occur a few degrees above the minimum. 91C, 92C, 93C etc. I can not see how Tj target has nothing to do with Tjmax.

As for the OP, the important thing is your distance from Tjmax whether Tjmax be 90C or 100C so the CPU itself is not being overworked temperature wise and naturally throttling. Happy to see you've made some progress with the BIOS.
 
I don't think there are any screw holes to mount fans on the front of the chassis. I'll have to see if I can rig something up. My case was a freebie, and it's just a cheapo inWin.
 
When I'm testing an E8400 and one week Intel tells me that has a TJ Target of 100C and then I test a Q6600 G0 which tests identically to the E8400, I have to assume that they have the exact same TJMax.

Some of the original TJ Target numbers that Intel released were so ridiculous that when the complaints came in they decided that they had made a mistake and decided to release some new numbers to keep users happy. This was after they contacted some of the programmers that provided evidence to show that their numbers were a farce. My "opinion" is that some of the numbers released by Intel were deliberately misleading. If you encourage users to use a TJMax that is 10C less than the actual TJMax then they suddenly have core temperatures that are being reported 10C less than the actual core temperature.

Kind of funny how RealTemp, CoreTemp and HWMonitor all decided to ignore Intel's 90C TJ Target BS for the Q6600 G0. Actual TJMax on some of the core 2 / core 3 cores is closer to 105C, not 90C.
 
Captain Picard said:
I don't think there are any screw holes to mount fans on the front of the chassis. I'll have to see if I can rig something up. My case was a freebie, and it's just a cheapo inWin.
Click to expand...

In your case, you might as well run it with the case sides off inside your cubby. Having the single fan trying to pull air through whatever intakes and gaps there are through your case while inside a cabinet is making it work extra hard contributing to heat and noise.

Put another fan on the other side of the radiator at least and make push/pull configuration.
 
Unfortunately I can't fit two fans on the radiator due to there being too little space between the block/pump assembly on the CPU socket and the back of the case. I tried fitting the fan sandwich into the case, but it's a no go.
 
Until you get that computer out of the cubby(computer oven), you are going to be stuck where you are...

I saw the one post where you mentioned 75c on an H50 and then later over 80c...

That right there is your problem.

My old system had a Q6600 G0 in it.

Even with my old modified Thermaltake Big Typhoon, I was only seeing 62c max temp with Orthos. IBT was a couple degrees higher, but nowhere near what you are reporting.

And that was with decently high voltages.. right around 1.4v.

The temps inside that cubby are probably at least 30-40f higher then room temperature.

And the more heat you pump into that cubby without proper ventilation is just going to made the problem even worse.

The easiest thing you could do to lower temps is to get it out of the cubby.

If that isn't a possibility for whatever reason, you can cut the back off of the cubby provided it isn't right up against the wall.
 
Yeah, I just need to get a new case and get the PC out of the desk cubby. I sort of need a new desk for that too. :(
 
Captain Picard said:
Unfortunately I can't fit two fans on the radiator due to there being too little space between the block/pump assembly on the CPU socket and the back of the case. I tried fitting the fan sandwich into the case, but it's a no go.
Click to expand...

You could try mounting it outside the case, so that the whole thing hangs "out" instead of "in" and frees up that space near the cpu.
 
Spare-Flair said:
In your case, you might as well run it with the case sides off inside your cubby. Having the single fan trying to pull air through whatever intakes and gaps there are through your case while inside a cabinet is making it work extra hard contributing to heat and noise.

Put another fan on the other side of the radiator at least and make push/pull configuration.
Click to expand...

Yes, this is definitely worth doing.
 
krakerman said:
You could try mounting it outside the case, so that the whole thing hangs "out" instead of "in" and frees up that space near the cpu.
Click to expand...

cause the tubes dont come off you cant do that, but you could mont 1 fan and rad in the case 1 fan out
 
tricky0502 said:
cause the tubes dont come off you cant do that, but you could mont 1 fan and rad in the case 1 fan out
Click to expand...

Gotcha, thanks! That's good to know, as I've been considering picking an H50 up for a second building and doing just that. Will have to look into it a bit more now.
 
Captain Picard said:
I've been hesitant to use P95 to verify stability, since it was my understanding that if you have issues with IBT, then the system isn't stable. There's definitely something strange going on, which most likely indicates that the system isn't actually stable.
Click to expand...

I just thought that I should mention that I have passed 40+ runs of IBT with a C2Q in the past only to fail the custom prime run that I mentioned and with my wolfdale I have failed prime after passing 40+ runs of IBT as well. Linpack is not the end all be all test. I only run it last since it doesn't show errors reliably. If you are borderline stable you can pass 20 or so runs only to fail the next time that you run 10 which has happened to me in the passed. If you show errors 25 minutes into small fft the next time that you run it you will show errors 25 minutes in which can really help with tweaking.

To be honest the only time that I have ever failed linpack after passing the custom run that I mentioned ,which is more stressful on the MCH than IBT, and small fft is because of the added vdroop if you aren't using loadline since the insanely high load that you will never see with any other program causes more vdroop than you will ever see again causing instability.

You can use IBT but I would never recommend using just IBT.
 
What is your MCH voltage ?
What is your FSB voltage ?

Run the IBT and put your finger on the MCH/Northbridge heatsink, how high can you count, one Mississippi , two Mississippi ... until you have to take your finger off ?


Everything points to heat as the root cause aggravated by the four sticks or ram which is a known OC killer with that board.
 
You must log in or register to reply here.
Back
Top