Haswell i5-4690k BSOD need advice

greyfox92

Distinguished
Feb 17, 2011
39
0
18,530
Hello I just made my rig I'm using ...

i5-4690k
DH-14 Noctua
z97-a asus mobo
2x4 ddr3 gskill sniper ram 1866 freq

I oc'd to 4.5 ghz at 1.25v while setting up xmp profile for my ram. My temps on HWmonitor for all cores was about a 72c max (for both IBT and p95) and it passed 10 IBT tests on very high and another 10 passes on maximum.

However, when I used p95 blend, I would get a BSOD at around 4 hours into the test. I upped my voltage to 1.257v to see if it would change anything but still BSOD BCCODE 124 4 hours into the test.

I googled a bit and found that p95 doesn't work too well with haswell chips....is this valid? If so, should I run aida64 instead and rely on that information? I'm concerned somewhat with the longterm stability of my rig because I'm a pretty hardcore gamer and sometimes do 10+ hours gaming.

I understand that these are synthetic stress tests aka my computer will most likely never be this stressed using real life applications, but I feel that a bsod only 4 hours of p95 blend is really really something to worry about in terms of stability.

Please pass some wisdom my way!!

I'm not trying to lower the freq of my ram btw. I will stick to using the xmp profile for my ram.

I've considered lowering my cache ratio by 3-5 compared to my oc multiplier. I really want to get at least 4.5ghz on 1.25v which I heard from many haswell users that this is easily possible. So what do you guys suggest?

EDIT: I'm trying to use aida64 but never used it before. I did a 10min test with just cache stress and ran it stable for 10 min and now I'm about to run the fpu test.
 

jasonite

Distinguished
Apr 2, 2012
345
0
18,960
I hear you bro, I just made my first gaming rig also. In my case I also got a 4690k, a Z97 Asus mobo and 1866 ram, and even a Noctua cooler!

Honestly I would not use Prime95 to stress test your CPU. Here are two rationales why. The first is a youtube video: http://youtu.be/0CHs5_TdpXE?t=13m25s

I set that one to right where he talks about Prime 95. The other is an article from Anandtech where he briefly discusses why he does not use it: http://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/2

I go along with that guy and instead run the POV-Ray benchmark and OCCT for about 5 minutes.

Honestly I used ASUS's AI Suite 3 and did their 5-way Optimization. Set it to extreme tuning, ratio only, and per core, as well as a 30-sec memory stress test. That is straight from an ASUS rep who was doing a video presentation. Then I'd do it again and this time instead of per core do all cores, and see which version gives you the higher result. That is a great base for overclocking, because it will take you quite high and also adjust your voltage and CPU cache as well, among other things. The Youtube video I linked to, if you watch the whole thing the guy recommends not increasing the base clock so I haven't. AI Suite 3 is great according to every source, and you can manually clock it up from there if you want to as well. That's my advice.

Jason
 
bugcheck 124 is a common overclocking /overheating problem. You can use the windows debugger and dump the reason that the CPU called the bugcheck. Use the !errrec command on the address of the WHEA_ERROR_RECORD

it will indicate why it failed and where. if you get several of the same failures you might find you hit a limit to your overclocking. if you get several failures but in different locations (memory banks) it might fixable. (increase cooling, or change voltages)
For best results: make sure you have the most current BIOS, and the most current CPU chipset updates installed.
These will give you microcode fixes for errors in the electronics of the CPU and motherboard.