2500K, stock cooler, mild OC, questions...

tech-wreck

Honorable
Jan 8, 2014
885
0
11,160
i have an i5 2500K in a gigabyte z68xpud3 mobo. haven't had the courage to mess with it until very recently, it only has the stock cooler at the moment. CPU temps 29c idle, peaks around 55c. airflow through the case is good, and it's regularly de-fluffed.

i've read http://www.tomshardware.co.uk/forum/265056-29-2600k-2500k-overclocking-guide and some other threads about OC'ing on a stock cooler, and now i'm a little more confident.

i've already bumped it up to 3.7GHz, and haven't noticed any noticeable rise in temperatures while testing. (test method: minecraft, visuals maxed, temp logging with speedfan)

without touching the voltages, how far can/should i push this?
and what noticeable improvements can i expect?

also, is the gigabyte app for tweaking the CPU from within windows any good? i haven't tried it yet. or is there a recommended alternative out there?

any other suggestions?

thanks in advance :D
 
Solution


Smart move. I smoked (literally) a Phenom II X4 965BE and the motherboard both using Intel Burn Test. Both the MB and the CPU were used items I had bought online. So there's no telling how bad they had been mis-used over their lifetime. The point being, do stress testing at your own risk, and don't depend on the CPU's ability to throttle down to save itself.

Btw, I never run for more than the default 10 passes with IBT. That's enough for me.

clutchc

Titan
Ambassador
Our systems are similar. (see my sig) You can get quite a bit of increase in performance with just the multiplier alone with that processor. I do all my OC'ing in BIOS. But right now, I just have the multiplier increased to 42. Run a program like Intel Burn Test for at least the default 10 run pass. If it passes, you can try increasing it more. Eventually, you will have to stop or raise the vcore voltage a tad to make it stable.
http://www.techspot.com/downloads/4965-intelburntest.html
 

tech-wreck

Honorable
Jan 8, 2014
885
0
11,160


dl'd :) i'll do the burn test tomorrow, it's nearly bedtime.
so far i've only increased the multiplier in BIOS, unless enabling XMP counts as an overclock.

are you running with a stock cooler? (i'm guessing not, looking at the rest of your specs (which i am very envious of, btw)) and what are your core temps like?

mine looks like this:
CPU_graph.jpg

^almost idle.
i'm guessing the green line at the bottom is ambient temperature. the other two unidentified speedfan readings are 25 and 36c, and i'm not sweating.


 

clutchc

Titan
Ambassador
I don't use speedfan, I prefer HWMonitor and Real Temp. The only important temps are when your processor is loaded up. Idle really doesn't matter, because if the CPU is idle it is bound to be good safe temps. (Unless there is a problem, of course).
You are correct, don't try to OC w/o an aftermarket cooler. The stock cooler is only sufficient for stock settings. I have the Cooler Master 212 EVO on the system right now. http://www.newegg.com/Product/Product.aspx?Item=N82E16835103099

Intel Burn Test is for checking to see if the OC you have set is stable. For long term verification, you can use Prime95 and let it run for an hour or more. Some folks leave it run overnight.
 

tech-wreck

Honorable
Jan 8, 2014
885
0
11,160
i've read a lot of good things about the 212 EVO, and i will invest in one at some point. why does anyone spend 2-3 times as much on other coolers when that model is arguably the best? anyway, i digress.
there seems to be some disagreement between whether to OC or not with a stock cooler, so i'm going to find out for myself.
my aim is to see how much extra heat the stock cooler can handle. it rarely goes over 65c at full load at the moment with the fan at 2/3 full speed, so i'm confident there is room for some minor tweakage without baking anything. and i'm expecting 'loud', i can always turn the music up.
i know, here be dragons. i'll report back with my findings.
 

tech-wreck

Honorable
Jan 8, 2014
885
0
11,160
done some stress testing, and the results are ... interesting; the intel burn test is aptly named. i set the multiplier back to stock (33x) and started, 20 passes. it seems i was way wrong when i thought i was taxing the CPU before. a continuous 92Gflops later and it was recording peak temps of 80c. wow.
then i remembered that the side fan has a manual speed switch, so i flicked that over to full speed and re-ran the test.
weird. peak temps of 84c... too much turbulence? i might ditch that fan and stick some polycarbonate there instead.

conclusion: i cannot overclock this computer with it's stock cooler, so i'm putting my OCing ambitions on hold until i get my hands on a 212evo. :D

 

Heroesneverquit

Honorable
Feb 13, 2013
251
0
10,810
Nah you can. Its different running burn tests and prime95 and such. I know prime 95 will overheat my motherboard. Doing whatever I want though whether it be gaming and streaming I have not gone over 44c. The current issue I have though is a motherboard issue. My mobo does not have any VRM heatsinks.
 


That's a good, sensible approach.

At 3.7Ghz the stock cooler is working at its absolute peak, and even then, it's only designed to sustain adequate cooling for brief periods when Turbo Boost kicks-in. Running the processor continuously at that speed definitely warrants an aftermarket cooler.
 

Heroesneverquit

Honorable
Feb 13, 2013
251
0
10,810
I had my FX 6300 OCed to 4.1ghz with the stock cooler without any problems. With an aftermarket cooler I would agree that I worry less about it but I was not all to worried about it when it was on the stock cooler anyways.
 

clutchc

Titan
Ambassador


Smart move. I smoked (literally) a Phenom II X4 965BE and the motherboard both using Intel Burn Test. Both the MB and the CPU were used items I had bought online. So there's no telling how bad they had been mis-used over their lifetime. The point being, do stress testing at your own risk, and don't depend on the CPU's ability to throttle down to save itself.

Btw, I never run for more than the default 10 passes with IBT. That's enough for me.
 
Solution