i5-2500k OC

For the hot summer I ran some short P95 tests to find a low voltage, and settled it at 3.8GHz with 1.24V back in May and even ran a 4h P95 run and shut it down myself. Not a long test at all, but I just figured it was all right since 3.8 is pretty low. Never noticed any issues at all over the summer, hardest thing I ran was BF3, never crashed once.

So yesterday I ran a few quick checks and P95 crashed each time in ~10-15 minutes. Could be my ram OC aswell, and some chip degradation too I suppose, but 1.24V is low.

So now I'm at 4.2GHz w/ 1.335V, been running P95 for 5h50m, volts are mostly 1.34 in HWmonitor, 1.325 crashed in 10min so I went to 1.335 and disabled C1E.

Does [email protected] sound about right? Last spring I did 8h run of P95 of [email protected], and then just settled with [email protected] and never checked again.

And silver arrows roc, 22C outside and now 6h into P95 blend temps are
50 57 56 52 -/+ 2C
package temp??? is ~59, whats that the "socket temp" ?

You're voltages seem a bit high. I can get 4.3 at 1.2v on my 3820, but that's a whole different cpu/platform. Your temps look fine though.

Make sure your ram timings and volts are set manually and then try for lower voltages. But your temps are looking good. You want to keep it below 70 or 80 generally.

What are your system specs. Maybe take some screen shots of you bios.

Your voltage is pretty high for 4.2, im under that for 4.5.

M33Cat said:

Could be my ram OC aswell,

Click to expand...

Overclocking DDR3 is generally considered useless unless you're benchmarking. If you really want to overclock your ram then you should do it AFTER you've found your stable CPU oc. One thing at a time.

and some chip degradation too I suppose

Click to expand...

no

So now I'm at 4.2GHz w/ 1.335V

Click to expand...

Is LLC turned on ? load line calibration? Or Vdroop as it's called on some motherboards.

Does [email protected] sound about right?

Click to expand...

My 2600k does 4.5 at 1.35v, yours should be even better. Less heat and less voltage without hyperthreading.

Last spring I did 8h run of P95 of [email protected], and then just settled with [email protected] and never checked again.

And silver arrows roc, 22C outside and now 6h into P95 blend temps are
50 57 56 52 -/+ 2C

Click to expand...

Why is your overclock changing as the seasons change? Anything under ~75c is phenomenal for these chips. I'd consider anything under 80c fine, myself.

I think your LLC needs to be turned on or adjusted (or ram set back to stock settings), and you could shoot for a 4500 "set it and forget it" overclock.

package temp??? is ~59, whats that the "socket temp" ?

Click to expand...

Safe to ignore, its a diode inside the cpu socket I think.

Gigabyte z68xp-ud4 with updated bios F9, my LLC is at 5 of 10, it its needed for about +1.25 and it keeps it very very close
PSU: XFX 750W pro black edition ~1yo

Yesterday I tried some P95 runs with 4.4GHZ and 4.2 @1.28, ... 1.335, 1.345, they all failed in the 1st 30min, had a few BSOD RAM related 0x0A and 0x1A. But I'm not going over 1.6V on them.

So I set my 2x2GB 1600 g skill eco's from 6-8-6-20-24-1T @1.590-1.6V BACK to stock 7-8-7-24-40-2T @1.35V. Before I had the QPI/VTT at 1.100, I set that back to stock 1.050.

Next
4.2 @1.335 w/ stock ram, and that gave another ram BSOD

Next
4.2 @1.335 upped ram to 1.4V, ran P95 2800MB blend, ran for 12H50m until I stopped it, temps were normal.

Today
C1E, C3/6, EIST, thermal, are disabled

4.2, LLC 5, Vcore is set to auto, not an off-set just AUTO

HWmonitor said 1.31 at idle 4.2GHz, and now in P95 large FFT voltage dropped to 1.27-1.28, and its now 2h3m into it and the temps reflect the lower voltage

So I'm a bit confused, I'll try a manual 1.280V again, but why did it fail a few times before

because when you apply the heavy P95 load to it, the voltage drops. HW monitor shows 1.31v at idle, and 1.28v under load.

That's what LLC is for, it tries to keep it from dropping so much by applying an offset of X amount, but only when it detects a drop in voltage.

So you will set manual 1.28v and then apply p95 load and it will probably bsod again. Also you keep changing your testing conditions. You are playing with way too many variables at once. You need to play with one thing at a time. And test with the same options each time.

\dev\goat said:

because when you apply the heavy P95 load to it, the voltage drops. HW monitor shows 1.31v at idle, and 1.28v under load.

That's what LLC is for, it tries to keep it from dropping so much by applying an offset of X amount, but only when it detects a drop in voltage.

So you will set manual 1.28v and then apply p95 load and it will probably bsod again. Also you keep changing your testing conditions. You are playing with way too many variables at once. You need to play with one thing at a time. And test with the same options each time.

Click to expand...

thats not the problem, LLC is at level 5, when I manually set 1.28 and run P95, it stays at about 1.28, just like when I set 1.335V and run P95, it goes to about 1.36V

Also you keep changing your testing conditions. You are playing with way too many variables at once. You need to play with one thing at a time. And test with the same options each time.

Click to expand...

I want to quote this post.. Make ONE change at a time man so you know EXACTLY what the difference maker is.

You should just set your ram at stock speed, timings, and voltage. This shouldn't be the problem w/ you overclock unless you have defective ram. Which if this is the case i would set everything back to stock and run memtest.

Why are you running your vcore on auto? I would try leaving C1e and speedstep enabled an overclock in offset mode. Try running Small FFT

As earthdog was saying your making too many changes so its going to make it harder to identify the problem

Why is your overclock changing as the seasons change?

Click to expand...

Just to take it a bit easier on the MB

Lucky23 said:

Why are you running your vcore on auto?

Click to expand...

I was just seeing what it would do

I seem it have it stable at 4.3GHz @ 1.300, ram at 7-8-7, C1E, C3/6 enabled.

I've ran IBT using 3000MB ram for 20pass, 100pass, 100pass, worked fine avg about 111.5 GFlops

I've ran a 12h21m run of P95 LFFT, its still running now on this rig, I just changed back from my troubled backup rig

[email protected] made IBT stop itself at pass 57/100, and P95 had troubles too

[email protected] passed 100pass of IBT and a 14h05m P95 blend until I stopped it.

So I'll try 1 last custom blend at 4.3GHz, but I'm happy it plenty stable for day-to-day and gaming. Next I'll have to try the offset method.