Question about IntelBurnTest

[BonzaiDuck]

HOW I SPENT THE LAST THREE DAYS OF MY SUMMER VACATION

By BonzaiDuck, Public School #69, 3rd Grade (third time around)

1) Found my "quick fail" point with this E8600 for 4.1 Ghz, CPU_FSB 410, DDR-820, FSB 1640.

2) Adjusted the voltage to 1.30000V. Tweaked the memory tRC to 16 (for 4,4,4,12)

3) Ran Small-FFT for 16 hours (no errors)

4) Ran Blend-test for 16 hours (no errors)

[ . . . Saying . . . . "Ahhh . . . . Gee . . . . I dunno . . . . I jus' dunno . . ." as I consider running 64-bit IntelBurnTest.]

So I fired it up, and chose the "half-of-memory-usage" test. Ran only 3 iterations to see what happens with the core-temperatures and TCASE.

Not bad . . . not bad . . . . TCASE of 57C, core values go up to 64C/64C.

Of course, I have issues with these sensors and a possible need for BIOS revision, but these temperatures are biased upward from what they should be. Just wanted to see what happened with my big toe here.

NOW . . . QUESTION . . . . Is there any difference in the stress on the CPU for running the "half-of-memory" test as opposed to the "maximum stress" options?

Just curious. I guess I'll find out, either way. [but . . . . I dunno . . . . . gee . . . . I dunno . . . . maybe . . . . maybe . . . . I feel like Morgan Freeman with Gwyneth Paltrow's cardboard box in "Seven." ]

 

[error8]

Yes there is. The more memory the test eats up, the tougher it becomes. The problem is that setting maximum memory usage gives an error and the program crashes. I guess it tries to take more ram that it's available and that's the cause of the crash.
I would advise you to try using the 4'th option and set the memory quantity by hand, trying to find the maximum ram the program can take before crashing. For example, I'm on Vista 32 bit, that gives me a total of 3500 mb of ram and the maximum IntelBurnTest can take is 2000 mb. Setting 2050 mb or more and the program crashes.

EDIT: For example , at 4 ghz, if I set half memory test, I can pass something like 7-8 tests out of 10. But setting the memory at 2000mb I only pass one test. So the more memory it eats the more reliable it becomes in "smelling" instability.

 

[BonzaiDuck]

Error --

Thanks for the insight. I'm running 20 iterations now. We'll see whether it reports an error.

As far as the thermal dissipation, I want some bragging rights: "IntelBurnTest is sissy compared to what people had reported."

So far, on "Maximum Stress":

E8600, 4.0 Ghz using 4.1 Ghz voltage settings, 1:1 with moderately tight timings (also withen spec.):

Peak temperatures: 64C/65C (or viceee-versah.)

Now -- I'd reported before, and it has been discussed -- about "sticking" sensors. The sensors move when TCASE shows the "TCASE-TJunction-Delta" pushing above the "stuck" level. All sensors as read by software or by BIOS are biased upwards -- for the TCASE diode, this is a BIOS problem. PRIME95 load temperatures (57/57 at this same room-ambient) reflect a Load-minus-Idle spread in TCASE characteristic of this heatpipe cooler with other C2D processors. We've seen statements that "INtel does not mean for the core temperatures to report idle temperatures." I suspect that's a warranty-disclaimer, if you choose to RMA the processor because idle core values that should be 38C are always shown as 46C in RealTemp or 56C in other monitoring software.

Of course, my assumptions about "upward bias in idle TCASE" and "core readings unstuck above the level shown at idle" extrapolate to my other remarks, but it makes sense.

We'll see if this VISTA-64 and IntelBurnTest x64-version crash -- it hasn't happened yet.

EDIT: Task Manager continues to report under "Applications:" "Running."

It's grabbing about 3,030 MB of RAM. Monitoring software in background; graphic display; TAsk Manager -- no problem so far. This is going to be interesting . . .

[Pardon my enthusiasm. Y'all probably maintain an honest level of privacy when doing this. I'm just stoked. And by this evening, I will have total thermal control of fans with "ESA" and no noise for the most part. {YEeeee-Haaaah!}].

EDIT AGAIN: Haven't yet installed the screenshot software -- but "YEEEE-HAAAAH!!" How to stop worrying and Luv Yur Bomb!

Someone said they didn't think 20 iterations was yet enough. Thing ran 68 minutes without missing a lick -- passed 100%.

Here's my advice to the Minions of Professor Chaos and General Disarray:

ATTEND TO COOLING.

READ THE ARTICLE FROM FALL 07 (Anandtech) ON "OVER-CLOCKING THE QX9650."

FROM THERE -- DON'T PUSH YOUR OVER-CLOCK TO THE EXPONENTIAL RISE IN VCORE.

FIND A SETTING ABOVE WHAT YOU WANT, CERTIFY IT UNDER PRIME95 WHILE YOU GO FISHING, AND THEN BACK IT OFF 5 TO 10 MHZ IN THE CPU-REFERENCE-FREQUENCY.

TEST WITH INTELBURNTEST. In my case, VISTA-64, I-B-T x64-version, and 4GB of RAM seemed to work fine on the #1 "Maximum Stress" setting. It may be different with different OS version and different memory size.