2500k 4.5 OC requesting help!!

[tomsanity]

As I am typing this thread I'm running prime 95 on a 2500k clocked at 4.5ghz. I've been playing around with the offset settings and am currently creeping closer the + end (currently still on -)

I came here to ask why the only core out of all the cores that are failing is my 3rd (4th worker)? It has the coolest temps about 5 degrees Celsius cooler but I always get an error saying rounding was .5 less than .4.

Currently at offset voltage of -.35 which is around 1.248-1.272 volts with max temp being 73. VID : 1.4161v

Does anyone have a fix or insight into my situation?

Thank you in advance!

 

[Smartazz]

Your overclock isn't stable. Try raising the voltage a bit more.

 

[Kenmitch]

Might help if you up the vcore a little more as suggested.

What motherboard? What other voltages have you changed? Playing with offset vore the additional turbo vcore?

 

[tomsanity]

Hey guys thanks for the reply. I'm using an asrock z77 extreme 4. I just woke up and at -.35 I was able to run 11 hours with 3 cores, but the core I was concerned about stopped at 5 hours. I got FATAL ERROR: Final result was 01B03033, expected: 974DA6CC

I haven't really changed any other voltage settings.

I was planning to change my LLC to level 3, do you think that would help?

 

[khurios2000]

Tom,

I got Z77 Extreme 4 and 2500k too, I found that my CPU need 1.296-1.31V to get 4.5Ghz stable.

I leave the memory at 1600Mhz 11-11-11-30-1T 1.5v, all other voltage options at auto.

My CPU voltage (offset) +.005v LLC 5.

Under CPU advance configuration:
Enhance Halt State (C1E) = enabled
CPU C3 = Disabled
CPU C6 = Disabled
Package C State Support = Disabled

 

[tomsanity]

Thanks khurious! Would changing the settings in my ram have much if an impact I currently run at 9 9 9 24 2

On a side note, if it's just one core failing hours before all my other core, could it be because of a faulty ram? I'm running small ftts instead of blend currently.


I'm currently running LLC at level 3 with a -.50 offset which is giving me the voltage that you are insisting. Well see how it goes.

 

[Yuriman]

You should run your RAM in spec. If you have 1600mhz RAM, use a divider so it stays at or below 1600mhz. Use the timings that are printed on the side of the stick. Overclocking your RAM will have relatively little impact on overall system performance with a large potential for instability (unless you know specifically what you're doing).

On a side note, if it's just one core failing hours before all my other core, could it be because of a faulty ram? I'm running small ftts instead of blend currently.

Unlikely. I'm surprised you're able to get such a high overclock with negative offset. Ivy Bridge's voltage tables are different from Sandy's of course, but I need about +80mv (with load line calibration at "level 5" which is most stable on my system) to hit that same clock. What is your actual, loaded voltage with that offset? Voltage is the number that counts, offset is just a way of getting there.

Also, I would probably avoid using a negative offset in general because you run the risk of making your system unstable at idle.

 

[tomsanity]

At -.50 LLC 3 I get around 1.296-1.31V . I haven't touched my rams at all, everything is still at default settings.

LLC 5 -.35 gives me about anywhere from 1.240-1.280. I can't go any higher due to my temps. Its breaching 80. I'm planning on removing my 212 evo and re-seating it.

 

[Yuriman]

If you're needing to use negative offset I'd leave LLC @ 5 in that case. 1.3v is a fairly safe voltage for that chip.

 

[tweakboy]

As the gurus said, 1.28v for 4.5 is too low. IM surprised you can boot into windows.

I would personally not do offset and actually put a voltage. 1.35 to 1.39v for 4.6Ghz gl

 

[FalseChristian]

I'm running my i5 2500K at 4.5GHz stable. It's currently using 1.275v at idle up to 1.34v at full load. Works great!

 

[stahlhart]

A couple of things that worked for me for getting Prime95 stable:

-- set your memory timings manually (do not use an XMP profile, even if the settings are identical; set them individually and manually)
-- manually set CPU PLL and VCCIO voltages, as opposed to leaving them at "Auto"... for VCCIO you'll more than likely need to bring it up from a default setting that's less than 1V, but don't exceed 1.15V. CPU PLL you'll have to play around with, I think it's going to individually depend on the buiild, but you'll more than likely end up somewhere in the vicinity of 1.79V to 1.85V

Are you paying attention in the log files for which iterations are failing? There are a few specific values that Sandy Bridge has trouble with (1344K, 2688K, 1792K), especially if you're attempting to certify a stable overclock.