Z77A-GD65 + i5 2500K Overclocking - How to start?

[profJim]

As I recall:
CPU:OCCT is similar to Prime95 and uses small, medium, or large datasets. I don't know if this mode uses AVX.
CPU:LINPACK is similar to IBT and LinX
GPU:3D is similar to Furmark, but it doesn't appear to support DX11
POWER SUPPLY: Does all of the above and puts a maximum load on your power supply!!

It looks like the only test that supports AVX is the Linpack option.
Much as I like some of the OCCT features, I think that you will be better off with Prime95.
My personal preference is to go with Prime95 (AVX), HCI Memtest and if you want a GFlops score, IBT or LinX for a brief 5 loop test using all available memory.

It might be useful to run the OCCT Linpack test if you want to test with a another AVX program.

A web search using: prime95 vs occt might provide more detailed information.
http://www.ocbase.com/index.php/support might provide more information.
A web search using: occt forum should provide useful feedback.

[Zach_]

Grazie

[Chike]

I have never yet seen a SB system in my own eyes.
You know the voltages just keep playing with them.
Just for the kick try lower PLL, I mean under 1.8, 1.78 -1.74.
I didn't say VTT 1.2 is safe I said I have seen people that use it as high, or close, and also people that set Vccsa over 1.1v, how safe it is I don't know.
Try also to lower the DRAM voltage, some modules operate and OC better with lower voltage rather than high.

[Zach_]

Update:

So I raised Vcore to 1.42 (1.40 in CPU-Z) and tested.

10 loops of IBT on Max (6700MB), passed.
Did a large FFT test for 6.5 hours, passed.
Did a small FFT test for 4 hours, passed.
Then I did a blend test which amazingly, failed after 7.5 hours... I don't know the exact size of the FFT but it had to be something in between 100-200K according to the pattern.

I have no way to explain this, the only possible one I can think of is that while I was asleep the CPU temperature got close to 80C as I turned off the AC and the system may have BSOD'ed itself due to the high temp, the error code was again 124 but there's no way that there wasn't enough Vcore.

I wanna call this stable, what do you think? I don't think Prime failed because of a stability issue because it passed a large FFT test of 6.5 hours.

[profJim]

What are the cpu and memory speeds that you used with vcore = 1.42 volts?
What are the other bios voltages and custom settings that you used?

CoreTemp allows temperature logging and you can set the logging interval in seconds.
I think that the default setting is 10 second intervals.
RealTemp probably has a similar interval logging option.

Is your system stable? Possibly.
How long does each Prime95 test need to run for a system to be considered stable?
8 hours? 12 hours? 48 hours?
There isn't one definitive answer.
I'm satisfied if each test completes 12 hours. This is my personal optinion based on SWAG.
Be nice to your cpu, don't run IBT or LinX for more than 5 loops using all available memory.

Run some additional AVX stress tests using different programs for 8 hours or longer.
This should help remove any doubt as to how stable your system is.
Don't forget the run the multi GB zip/test integrity/unzip test that I described earlier.

Food for thought:
As the bios matures for your motherboard, higher overclocking and better stability might be possible.
This is very common with any recently released motherboard/chipset release.
When you upgrade to newer bios versions, it's very common that you will need to tweak some of your previously stable bios settings.
Remember that tweaking requires repeating stress testing.
When you flash to a different bios version (new or old) you will probably lose all your saved bios settings.

From what seen, it looks like many or most z77 boards can o/c to 4.0 - 4.2GHz with nearly stock voltage settings.
YMMV

Nice job!!

[Zach_]

I tried 4.6GHz with:

Vcore - 1.42
S/A - 0.965
I/O - 1.07
PLL - 1.84
DRAM - 1.504

Memory at 1600Mhz with hard set timings of 9-9-9-24 and the DRAM voltage is manually set to 1.504
At first I tried IBT with PLL at 1.8 but each loop took 180 seconds and yielded 92 GFlops as opposed to 130 seconds and 120 GFlops before.

So I raised PLL to 1.84 again and presto, loop completed at 130 seconds with 120 GFlops.

Is there a CoreTemp log that can tell me what was the temp was at the time of the BSOD?

Yeah I should upgrade my BIOS to 10.5, currently I'm using 10.3 because previously when I used 10.5, I was unable to change the voltages but after that I realized that they can only be changed withe the +/- buttons or simply by typing them in, so I will update my BIOS and try the same voltage settings in Prime 95.

After that, I'll run the Zip test, are there any more AVX stability tests you can recommend aside from IBT and P95?

Thanks

[profJim]

With Prime95, if one of the worker threads has an error, only that worker will stop and it will display an error message.
Each of the remaining worker threads continue to run.
The enclosed screen-shot shows that two worker threads had errors at the same time.
The Main Thread display at the top of the window shows the starting time.
If you get a BSOD, you can find that time of the failure from the last entry in the temperature log or get an approximate time from Prime95's Results.txt log.

I used WhoCrashed over the weekend when I got an unexpected BSOD = 0x21.
WhoCrashed displayed a summary of 15 BSOD's that I've had going back to early December 2010.
FWIW, almost all of the BSOD's occurred while stress testing my overclocked settings.
You can optionally set the number of displayed BSOD's to a different number instead of the default setting = 15.
Resplendence Software - WhoCrashed, automatic crash dump analyzer
Resplendence has freeware and paid versions for their programs, check them out.

OCCT has limited AVX testing.
I don't think that Orthos uses AVX.
A web search using: stress test AVX might provide links to other suitable programs.

Sample Prime95 worker thread errors:

[Zach_]

Thanks, I'll try again using the updated BIOS, but Jim, you have to tell me how is it possible that AVX IBT passes 4.6 with 1.37 Vcore and AVX Prime fails at 1.42, it's beyond me... I'm thinking that my Prime is bugged...

[profJim]

Thanks, I'll try again using the updated BIOS, but Jim, you have to tell me how is it possible that AVX IBT passes 4.6 with 1.37 Vcore and AVX Prime fails at 1.42, it's beyond me... I'm thinking that my Prime is bugged...

WRONG oh misguided one.
IBT or LinX can quickly fail after a short test using 5 - 10 loops, where Prime95, OCCT, Orthos and similar programs might not fail until many hours have elapsed. Many experienced overclockers have known about this for several years. That's why many of us use a brief IBT or LinX test only as a preliminary stress test. If either of these fails, there is no sense running Prime95 or similar tests unless you want to see which FFT size will fail.

I've seen some posts where the OP ran IBT or LinX for 50 or 100 loops or sometimes overnight with no problems, but Prime95 would eventually fail.
My bottom line, IBT and LinX are NOT good enough to reliably stress test your system.
Their GFlops scores are a good indicator that shows if your system is performing at an expected level.
My overclocked dual core P35/E6300 (R0) system maxes out at 23 GFlops.

Even if HCI Memtest doesn't use AVX, I think that a 4 - 6 hour long test is still useful.
Once your system appears to be stable, run your most demanding PC games with maximum stable quality settings for several hours at a time.
It's a tough job, but someone's gotta do it.

It's possible that Prime95 AVX has a bug.
You should post your findings in their forum and see what the developer and others have to say.

[Zach_]

:)