New member...OC'ing an older combo (P8Z68-V and i5-2500k)

[texasniteowl]

I've had this combo for a couple years and to be honest, it is still overkill for the majority of my uses. However a few weeks ago my Corsair H60 stopped cooling, so I replaced my cooling, and since I'm "playing around" with my system at the moment I decided to go ahead and try to OC it.

Basics: i5-2500k, ASUS P8Z68-V (bios 3402), 8gb Corsair ram, Thermalright True Spirit 140 BW Rev. A cpu cooler, 4 GT AP-53 case fans, Plextor M5 ssd, WD black 1tb, no gpu...using integrated (don't shun me!). Corsair HX-520 PSU (I think) that is about 6 yrs old I think.

Right now, I have set my turbo multipler to 44, LLC is regular, VRM is manual/350, optimized phase control, t.probe duty control, 110% cpu current capability, offset + 0.005, cpu spread spectrum disabled. I understand with such a low OC, I can set VRM back to Auto?

My idle vcore with these settings is 0.992-1.000 (cpu_z) and my highest load vcore is 1.320-1.328 (light load, during the image editing/GIMP) with sustained 100% load at 1.280-1.288. I did run 1 cycle of the RealBench benchmark tests and watched cpu_z while that was running to note those vcores. Max temps reached during the 1 run of RealBench were: 60/56/61/62. Note...at these temps, my fans only run at about 70%, not full speed.

I suppose I'm looking for input on how those voltages seem. I don't really have much interest right now in pushing the OC higher. I was getting input on another site, but their focus tends to be solely on the 100% load (ie. prime95) vcore which runs lower than the light load (ie. GIMP image editing) part of the RealBench tests. So is a light load vcore of 1.320-1.328 normal when the 100% load vcore is 1.280-1.288? On my system I am much more likely to be constantly jumping between the 1.320 and idle 1.000 than I am to be at 100% load. After all, it jumps to the 1.320 just opening a new tab in Chrome.

I have saved some screenshots of my bios, so if need be I will figure out how to post them here.

Thanks for any comments.

[Arne_Saknussemm]

The voltage difference is normal yes....that is vdroop...a protection feature which lowers voltage at very high loading so any overshoot is kept within safe limits.

How did you arrive at that offset figure...manually testing voltage? I ask because it might be possible to run 4.4GHZ with a touch less voltage..

Have you experimented with negative offset? like -0.005...-0.01?

But generally OC sttings...volts and temps fine...just saying..if you can drop voltage even more by doing a bit of testing why not 🙂

[texasniteowl]

Thanks for responding!

I've been using offset all along, never used a fixed volt. I started at 38, and stepped up to 40, 42, then 44. I have once or twice tried a negative offset. Most recently, with settings: 44, medium llc, auto vrm, 120% cpu current capability and offset of -0.010. With those, my idle voltage bounced between 0.976 and 0.984 but my 100% load (prime95) remained slightly higher at 1.280 +/- 0.008.

Also, this morning, with: 44, regular llc, manual/350 vrm, 110% cpu and offset -0.005 I got idle of 0.984 and 100% load of from 1.264-1.272 during h264 and heavy multitasking benchmark and 1.280-1.288 durng OpenCL.

About idle voltage...how low is too low? Is it possible to go much lower than 0.976 and still boot?

Also, since the system realistically is at idle more than it is at full load, should I opt for settings that give me lower idle volts over lower load volts? (As long as load volts are reasonable?)

Thanks!

[Arne_Saknussemm]

Hmm can't remember the typical idle voltage for those chips off the top of my head...but about 1v is about right I think...so yeah if you increase a negative offset it is possible to have idle voltage too low and cause instability but you'd have to do a bit of trial and error to find that for sure since each CPU is its own world

Maybe just leave it as it is and see if it will run 45 and 46 😉

[texasniteowl]

Well, as I'm already running below 1v, I will probably not play around too much chasing a lower idle voltage.

I have not tried 46 yet, but with the following settings, 44 and 45 were basically identical.
LLC: Regular
VRM Freq: Auto, VRM Spread Spectrum: Disabled
Phase Control: Optimized
Duty Control: T.Probe
CPU Current Capability: 120%
CPU Voltage: Offset -0.005

With both 44 and 45, idle was running 0.984. During 1 iteration of RealBench benchmark vcore was as follows:
image editing/GIMP: 1.312-1.320
x264 encoding: 1.272 +/- 0.008
OpenCL: 1.280-1.288
multi tasking: 1.264-1.288 (mostly 1.264 to 1.272 after the first 30 sec)
temps: 44 (60/58/64/62), 45 (61/57/64/62) = a draw

I might try 46...but then again...I may just stop here and run a couple hours of the stress test. If I try to lower offset any more, idle voltage will drop into the 0.97x range...and maybe my chip would be ok. But for now I might be tired of chasing voltages. 😄

[Arne_Saknussemm]

Heh...sure sometimes the tweaking has to stop...well, at least I try and tell myself that 😮

[texasniteowl]

Well, apparently i wasn't ready to stop. 😉

I have not tried 46 yet, but I went back to 44 and changed the offset from -0.005 to -0.010 just to see what would happen. I have not done a stress test, just 1 benchmark iteration, but it did survive that. And the vcores were not that far off from the -0.005 marks.

idle remained in the 0.976-0.984 range with one drop to 0.968.
gimp was primarily at 1.304-1.312
h264 was 1.264-1.272
opencl was 1.280-1.272
multitasking was 1.264-1.272 with occasional rise to 1.280 early and occasional drop to 1.256 late
max temps reached during the benchmark were 58/55/61/61

Since on previous settings, marks for 44 and 45 were the same, I may be able to go back to 45 with this -0.010 offset.
And this is all still on *Regular* LLC...aka 0%...

I am sort of tempted to see how low I can drop offset before I have idle problems...sort of.

[Arne_Saknussemm]

texasniteowl wrote:
Well, apparently i wasn't ready to stop. ;)....

I am sort of tempted to see how low I can drop offset before I have idle problems...sort of.

Heh heh...welcome to the eternal tweakers club...that's sort of what I'd be tempted to do 😄

[texasniteowl]

😛
Is there any good way other than just using the system to test for idle instability? Since all stress/benchmarking programs are more about testing load voltages?

[Arne_Saknussemm]

Hmm not really...I think using the OS is pretty much the only way. Drops from load to idle can help show instability though so running something for 30 seconds then stopping and watching system drop down...

I have to say Win 8 is really easy to corrupt so I usually do stability testing on a separate install on a separate drive....keep my 24/7 OS out of harms way as it were...win 7 was a lot less prone to this...