Monday, July 27, 2015

Had to stress test the 4790K after a long time


The last time I stress test day CPU was when I had the Maximus VII Gene board and the Noctua NH-U12S cooler. And it was with a low ambient temperature as well.

Fast forward to July of 2015 and we have a lot of changes. New case, meaning new motherboard and a new cooler. And Jul means, it's the summer. Sure, we have the A/C on, but the temps are around 26-27C, which is a couple of degrees higher than the previous months.

Wouldn't I have stress tested after all those hardware changes? I should have, but I didn't think it was necessary. For one, the stability of the overclock of Intel CPUs these days don't seem to be affected by the temperature. So the cooler or the lack of airflow inside the case shouldn't have made a difference. Then, Asus says that all of their Z97 Motherboards overclock exactly the same; only the feature set is different. More importantly, I didn't have the time to go through it again.

But yesterday, while I was transcoding a HD video in Handbrake, I got a BSODs. The dreaded "WHEA UNCORRECTABLE ERROR" error.

So the overclock wasn't stable after all. It was time for me to stability test the CPU. At least this time I had a starting point.

4.5GHz @ 1.21V Vcore with everything else set to default

That was where the CPU was at up until now.

But first, I wanted to make sure a stress testing tool would find the current overclock settings as unstable. I am not a fan of synthetic stress testing tools such as Prime 95 or AIDA64. Instead, I turned to Asus Realbench, which can be downloaded from here.

Sure enough; Realbench also found my current setting to be unstable – within minutes. That’s good. We are not in business.

I rebooted the PC into UEFI and changed the Vcore from 1.21V to 1.22V and re-ran the tests. It was still unstable.

Increased it to 1.23V. Still unstable.

Then I wanted to fiddle with the settings such as the Input Voltage and Load Line Calibration. Big mistake. You should never change more than one variable. A BSOD caused by unstable OC settings for Haswell CPUs almost always means lack of Vcore. Still, I increased the Input Voltage to 1.85V from AUTO (which was at 1.75V if I recall correctly). That did not make a difference, but I decided to increase the Vcore to 1.24V while keeping the Input Voltage at 1.85V. The Maximus VII Gene board had the Input Voltage at 1.85V.

1.24V passed. 8hrs of Realbench with 8GB RAM. I ran it overnight.


Luckily for me, the CPU temps did not rise into the 90s. The max CPU temp reported by RealTemp was 86C which is reasonable for this setup. RealTemp can be downloaded from here.

It might not be perfectly stable, but for my work, this is all I need.

No comments:

Post a Comment

Related Posts Plugin for WordPress, Blogger...