- Joined
- Jan 27, 2011
- Location
- Beautiful Sunny Winfield
I'm having a go at OCing my I7-4770K (CM 212 Evo air cooler and ASRock Z87 Extreme4 board in Antec 300 case with fans coming out the wazoo.)
Before OC - System runs all day long at about 56°C running FAH on all cores and GPU. All cores are reported at 3.7GHz.
My goal is a to find settings that I can run 24x7 to eek a little more performance (and PPD ) out of my system.
Preliminary tests:
Multiplier to 46 and Vcore to 1.20 => POST completes and boot menu (Grub) comes up. Crashes during boot.
Bump Vcore to 1.25 => System boots but crashes when I try to log in.
<fiddle with settings until I arrive at 4.2 and 1.23V.>
System will run stressers like FAH and Rosetta with no difficulty and temps in the 60-70° range. At this point I saw a post about using at least v28 of Prime95 for Haswell chips so I check and see that I'm using 27.9. I find 28.5 and switch to that. About 15 minutes into the test, the temperature shoots to 98°C and I stop the test.
I'm now at 4.0 and 1.23V. Prime95 (v28.5) will take the temp up to 94°C at times but typically runs below 90. I've had Rosetta running all night long and the highest temperature seen was 68°C and it is cruising at 55-57°C as I type.
Here's my question. Need I concern myself with the temperature that Prime95 achieves (when run in torture mode and with code expressly designed to stress the CPU) or should I base my OC on the high CPU loads that I'm actually running? If it was only a matter of a few degrees I wouldn't question this and would choose settings that were good for all possible loads, but in this case there is a 25°C difference which seems huge. I'm pretty sure that I could get another 0.2GHz and stay under 1.25V Vcore with the typical loads I'm running. (I'd have to test with Handbrake as well since that is a heavy CPU load.) If I had to choose an OC that was stable and cool with every possible load I could throw at it, I might wind up with little or no OC at all. I guess I should run Prime95 v28.5 on standard settings just to see what it does.
I'm curious what other OCers would choose to do when faced with these conditions. Yeah.. I know. Water cooling. I suppose this is a good reason to explore that. But let's consider what I could do w/out going that route for now.
Thanks!
====================================
Addendum for Linux overclockers:
I needed to run sensors-detect manually (Debian Sid) to get it to read Vcore and I don't believe the reading anyway.
Search the download server at Gimps for newer versions of Prime95.
The following command prints the CPU temperature to the console about every second.
You can watch sensors to monitor temp but it is more convenient to sort | tail the output of the previous command to find the highest samples.
Before OC - System runs all day long at about 56°C running FAH on all cores and GPU. All cores are reported at 3.7GHz.
My goal is a to find settings that I can run 24x7 to eek a little more performance (and PPD ) out of my system.
Preliminary tests:
Multiplier to 46 and Vcore to 1.20 => POST completes and boot menu (Grub) comes up. Crashes during boot.
Bump Vcore to 1.25 => System boots but crashes when I try to log in.
<fiddle with settings until I arrive at 4.2 and 1.23V.>
System will run stressers like FAH and Rosetta with no difficulty and temps in the 60-70° range. At this point I saw a post about using at least v28 of Prime95 for Haswell chips so I check and see that I'm using 27.9. I find 28.5 and switch to that. About 15 minutes into the test, the temperature shoots to 98°C and I stop the test.
I'm now at 4.0 and 1.23V. Prime95 (v28.5) will take the temp up to 94°C at times but typically runs below 90. I've had Rosetta running all night long and the highest temperature seen was 68°C and it is cruising at 55-57°C as I type.
Here's my question. Need I concern myself with the temperature that Prime95 achieves (when run in torture mode and with code expressly designed to stress the CPU) or should I base my OC on the high CPU loads that I'm actually running? If it was only a matter of a few degrees I wouldn't question this and would choose settings that were good for all possible loads, but in this case there is a 25°C difference which seems huge. I'm pretty sure that I could get another 0.2GHz and stay under 1.25V Vcore with the typical loads I'm running. (I'd have to test with Handbrake as well since that is a heavy CPU load.) If I had to choose an OC that was stable and cool with every possible load I could throw at it, I might wind up with little or no OC at all. I guess I should run Prime95 v28.5 on standard settings just to see what it does.
I'm curious what other OCers would choose to do when faced with these conditions. Yeah.. I know. Water cooling. I suppose this is a good reason to explore that. But let's consider what I could do w/out going that route for now.
Thanks!
====================================
Addendum for Linux overclockers:
I needed to run sensors-detect manually (Debian Sid) to get it to read Vcore and I don't believe the reading anyway.
Search the download server at Gimps for newer versions of Prime95.
The following command prints the CPU temperature to the console about every second.
Code:
while(:); do sensors | awk '/Physical/ { gsub("[\+C°]",""); print $4 }'; sleep 1; done