2500k @ 4.5 passes stress tests, fails to boot to windows days later

danielsix

Honorable
Dec 28, 2012
2
0
10,510
This is my first attempt at overclocking a CPU, and it's left me very confused. I followed the Sandy Bridge guide on Clunk, and overclocked my 2500k to 4.5 at 1.325v(with VDroop is was more like 1.3 under load). It passed a 12 hours on Prime95 overnight with no errors and seemed stable, temperatures were pretty low. Everything worked perfectly for 24 hours, it ran fine during gaming and normal use, but the today the computer would not even boot to windows. 3 times it either froze at the start screen or almost immediately after loading the desktop. I double checked all my bios settings and everything was the same as when it was stable. I restored the settings back to factory default and the computer ran perfectly again. The overclock was done via manual voltage, not offset.

CPU: i5 2500k
Mobo: Asus p8z77-v LE

These were the only settings I changed:

AI Overclock Tuner - Manual
By All Cores - 45
Internal PLL Overvoltage - Disabled
Load Line Calibration - High
CPU Voltage - Manual
CPU Manual Voltage - 1.325v
DRAM Voltage - 1.35v (correct for my RAM)

Could it just be some instability that Prime95 didn't catch? It just seemed odd that it could go from stable and running perfectly for an entire day, to not even being able to boot.
 

steddora

Honorable
Nov 13, 2012
686
0
11,160
Did you slowly work your way up to 4.5Ghz? Or did you simply set it to 45x multi and that voltage? I'd honestly recommend you back down to something like 1.2v at 4.2Ghz and see what happens there. 1.35v seems a tad high for the voltage at 4.5Ghz but I may be wrong. However I do know my 2600k actually runs 4.5Ghz at 1.315v very stoutly with level 3 LLC.
 

danielsix

Honorable
Dec 28, 2012
2
0
10,510

I started off at 4.0ghz, and was stable at 1.23v. Then went up to 4.5ghz. I found a googledoc from overclockers that suggested the average voltage at 4.5 on a 2500k was about 1.31v or 1.32v, with some as high as 1.38v. My cpu/mobo seem to have quite a bit of VDroop though, the voltage in CPU-Z was more like 1.296v-1.308v.

Do you have any recommendations for testing stability besides using Prime95? Like I said in the OP, it seemed extremely stable over a 12 hour test.
 

steddora

Honorable
Nov 13, 2012
686
0
11,160
Well make sure your cooling is good and run IntelBurnTest. It'll heat your processor up a lot more than any prime95 session will. But it won't pick up on minor instability. It will however pick off any major instability within about five minutes. But it will skyrocket temperatures. Example.. Prime can only get me to 55C at 4.4Ghz while IntelBurnTest pushes me clear to 65C. So be ready on the cooling level.

I hate CPU-Z on some setups. I actually had a friends 2500k reading that it was full load at 0.95v while clocked at 4Ghz.


As for the voltage ranges you're showing, that's about normal for the Sandy Bridge processors. 4.5Ghz usually seems to be the wall for different chips, especially on fixed voltages.

From the looks of it you need to watch that voltage in bios at idle and compare with a few different applications to find one reporting the best. LLC does it's best to control spike voltages from going from idle to full load and vice versa. Sometimes dropping to idle from full load while running at say 1.2v can actually cause a spike from 1.6 down to below 0.8v. So LLC probably isn't the determining factor since it's more than likely windows isn't running a full load on that processor during boot.

I'd go back to a lower clock and voltage and slowly work up. I'd try 4.0Ghz at about 1.1v or so and find a stable voltage from there and slowly work up. A good example though..

My 2600k runs about these on fixed and offset voltages

4.0Ghz @ 1.10v fixed -0.850v offset
4.2Ghz @ 1.15v fixed -0.050v offset
4.4Ghz @ 1.25v fixed -0.020v offset
4.5Ghz @ 1.31v fixed +0.005v offset
4.6Ghz @ 1.37v fixed +0.025v offset
4.7Ghz @ 1.41v fixed +0.075v offset
4.8Ghz @ Not stable +0.125v offset

At 4.8Ghz my processor needed over 1.45v and I wouldn't push it. How it was able to withstand IBT at those voltage offsets, the maximum voltage pushed was really jumping a lot higher than the offset shows. At 4.8Ghz it was pressing against the 1.45v barrier I refused to break on fixed and even went over a few times during IBT and absolute maxed at 1.46 I believe.

So each voltage is going to render something different. I personally have had better luck running offset voltages and controlling myself from idle stability issues with LLC. But you can see how profound the differences are the harder I pushed this processor. You'll see the same thing with yours as you go. But if you're having boot stability issues, I'd reset and start again like I said. Each chip is different and I've heard of 2500k's out there that needed over 1.3v just to hit 4.4ghz to get windows to boot. So see where you're at and don't think you're stable just because a few tests told you so. Don't always rely on Prime95 and IBT to tell the tale. Play your games, run your apps, that's where stability is going to matter.
 
@Danielsix .. try use auto voltage and + offset .. and in 4.5 everything still can enabled or auto

BCLK: 100.0, Turbo Ratio: by all cores, Multi: 45x, PLL Overvoltage: Auto, EPU Power Save: Disable, OC Tuner: Cancel/Off
CPU Ratio: Auto, Speedstep: Enabled, Turbo: Enabled / disabled , VRM: 350, Phase Control: Optimize , Duty Control: Extreme
CPU Current Capability: 110% ( i had still 100% and stable and don't use 140% .. it seem too high for me)
CPU Voltage: Offset Mode , Offset Mode Sign: + .. ( maybe 0.01-0.10) , Offset Voltage: Auto
also your RAM ... i don't know detail just try default first will be good for stability

my i52500K is 5.1ghz @1.44 .. seem too high and not good but just for trial and fery fast for render movie
my SS
good luck do trial and error