Going about testing stability on the Haswell Refresh / Devil's Canyon ...

[gupsterg]

I only recently bought a i5 4690K, prior to that my rig was a Q6600 G0; had this over 7yrs. This was OC'd to 3.4GHz (378x9 @ 1.3375v) with DDR2 @ 1007MHz on an Asus P5K Deluxe WiFi/AP Black Pearl edition for 24/7 use. It could do all I wanted and run F@H for days on end without BSOD, all the stress testing I did was Prime95 on it.

Now the i5 4690K I'm finding is a total different ball game to have the same kind of faith!

After a few stages of testing I was happy with 4.4GHz with cpu cache ratio 39 and without XMP / Speedstep / Adaptive volts, VCORE of 1.16v and VCCRING 1.12v

Then I enabled cpu cache ratio 41 / XMP / SpeedStep / Adaptive volts with same voltages. Over a lengthy period of use / stress testing and particularly F@H it was unstable. So I had to lower cpu cache ratio to 40 and up VCCRING to 1.15v.

Now today it BSOD after 5.5hrs of F@H with downloading going on in the background for 3hrs of that 5.5hrs but prior to this it had been tested with XTU / Prime95 v26.6 / RealBench / x264 extensively and longer time period (24hrs +).

As I didn't wish to use those programs I fired up Prime95 v28.5 and ran the in built benchmark and hey presto it BSOD in under 3min. So I thought try default settings just to check hardware AOK, no BSOD.

Then I went through my usual staged testing from bios defaults without XMP / SpeedStep / C-States and found CPU 4.4GHz CACHE 3.5GHz stable @ 1.18v in the Prime95 v28.5 benchmark.

Next enabled SpeedStep / C-States and adaptive CPU voltage and finally XMP with CPU Cache ratio 40.

What are fellow members using and how long are they testing?

[NemesisChild]

Prime is not always a good stress test, in my opinion, it's results are inconsistent.
I think it's safe to say that most people here in the ROG forum use Real Bench.
Running the stress for an hour will usually be sufficient to test overall system stability.

But the true test for me is how the OC handles in games.

[gupsterg]

I have no issues in Realbench can loop it for hours, same with XTU , Prime95 v26.6 , x264.

I have no issues in games / 3dMark for hours.

What I'm finding is folding @ home system is BSOD'ing and IMO is stressing system more than Prime95 v26.6 / XTU / RealBench / x264 and games.

It;s not the GPU in folding @ home as that GPU has been tested in my Q6600 for 24hrs plus.

I have stayed clear of Prime 95 v28.5 due to the AVX code creating high temps / power use but this is the only thing via the in built benchmark showing instability very quickly.

*** edit ***

The hex value for the BSOD I had today whilst folding @ home was Bug Check Code : 0x00000124

On a info page by der8auer it states:-

Core clock is unstable. You need a higher core voltage

Which matches what I had to do to core voltage to get Prime95 v28.5 in built benchmark test to remain stable.

Now the real test will be once folding @ home has been run again.

[NemesisChild]

That error code is almost always related to inadequate CPU voltage.

[gupsterg]

Cheers for post earlier and now 🙂 , felt I should have put a smiley in my last post as I felt my text a bit abrupt 😉 .

Looking at crash data I think PC crash 2 times in real use between 02/05/15 and today (15/05/15) , as other crashes are 10mins of each other they were periods of OC / settings testings IMO, will now keep a better log than I do about things I do 🙂 .