Gigabyte GA-EP45T-EXTREME Living Review

[Toben]

Everything was ok last night, did some benchmark tests in SiSoft, all ok and temps good too. CPU was around 31c on air with stock cooler, and System temp was 38c. When I powered on this morning the PC reset a couple of times before post, when I went into BIOS I got red message, saying some of the voltage settings were wrong. Strange, cos as far as I can see I haven't set anything to high. Anyway, I reduced the MCH Core from 1.22 to 1.180, and also the CPU VCore to 1.243. PC Seems ok now.

[Toben]

Also forgot to mention, changed the DRAM Performance Enhance to Turbo. Memory running at 1066. BIOs version is F9.

[supershanks]

Hi Toben.
Sorry for not replying earlier, been away for a few days & spent most of today in traffic jams , coming home.

Performance Enhance to Turbo.

in most cases should be fine, If stability is an issue try going back to 1st option - normal.

voltage settings were wrong

Usually just means over clocking has failed. Might be best to try bios settings you listed earlier, and running Prime95 25.6 and make a note of what it fails on. If large FFTs then it's probably ram, if small ffts then a cpu problem

[Toben]

Pain in the backside being stuck in traffic, I know what thats like. I do quite a bit of driving in my job. Drove to north Wales and back from Swansea the other day! Ok, I see what I get with Prime95. Not used Prime95 before, is it easy enough to use?

[supershanks]

yep you got it re traffic, gets worse. Yes if you follow the link i gave in last post there's afurther link with a bit of help on run options & resuolt interpretation.

[Toben]

Did some tests with prime95, this was the result.

WORKER 1
[Dec 11 00:04] Worker starting
[Dec 11 00:04] Setting affinity to run worker on logical CPU #0
[Dec 11 00:04] Beginning a continuous self-test to check your computer.
[Dec 11 00:04] Please read stress.txt. Choose Test/Stop to end this test.
[Dec 11 00:04] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Dec 11 00:04] FATAL ERROR: Rounding was 0.5, expected less than 0.4
[Dec 11 00:04] Hardware failure detected, consult stress.txt file.
[Dec 11 00:04] Torture Test completed 0 tests in 0 minutes - 1 errors, 0 warnings.
[Dec 11 00:04] Worker stopped.


WORKER 2
[Dec 11 00:04] Worker starting
[Dec 11 00:04] Setting affinity to run worker on logical CPU #1
[Dec 11 00:04] Beginning a continuous self-test to check your computer.
[Dec 11 00:04] Please read stress.txt. Choose Test/Stop to end this test.
[Dec 11 00:04] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Dec 11 00:05] Test 2, 4000 Lucas-Lehmer iterations of M19922943 using FFT length 1024K.
[Dec 11 00:06] Torture Test completed 1 tests in 2 minutes - 0 errors, 0 warnings.
[Dec 11 00:06] Worker stopped.

If I run the test again I get the same Fatal Error, but on the other Worker.

[supershanks]

hi toben,

your test failed ruinning

FFT length 1024K

compare that length to that link I mentioed

Large FFts
>512k
When processing these the CPU must use memory because of the size. So the calculations are done by manipulating the value in ram. So though both memory & CPU are being used it's memory that is getting stressed more.

As a consequence if Blend fails on an fft (>= 512k) then the best starting point to correct it is memory.

Small FFts
<512k
The data in this case can be mainly processed by the CPU with less memory usage. Consequently a failure on small fft (<512k) is likely to be linked with CPU issues.

you can see that it's a large fft (>512k) and so is likely ram related.

are all your ram timings on auto ?

[Toben]

Hi Supershanks. The memory timings are set to auto, showing 5-5-5-16. All I have changed is the voltage for the DIMM which is 2.0v. My memory is OCZ 4Gb 1066 Gold kit. For a successfull OC test with Prime, would you not expect to see any errors?

[supershanks]

For a successfull OC test with Prime, would you not expect to see any errors?
Edit/Delete Message

yes you would normally run prime in blend mode for an hour, if it does that without errors then it pretty stable.

You might try enabling XMP, if you do you'll have to re-enter your overcklock, CPU Host Frequency (Mhz) , etc., then rerun prime again.

[Toben]

Thought so, with Prime showing faults after ten mins, didn't think that was good. I dont have XMP in my Bios. My board is the GA-EP45-UD3P.