Xeon Owners Club

[Regeneration]

How do you test for stability and for how long?

My method for 1000% stability is: 8 hours of IntelBurnTest with 85% RAM, 8 hours of Prime95 26.6 Large FFT, and finally 24 hours of blend.

Uncore overclocking instability usually takes 5 hours to occur (BSOD). Normally on 512K test on Prime95. Sometimes, when you almost stable, you get silent Kernel WHEA cache errors in the event log.

 

[Durvelle27]

How do you test for stability and for how long?

My method for 1000% stability is: 8 hours of IntelBurnTest with 85% RAM, 8 hours of Prime95 26.6 Large FFT, and finally 24 hours of blend.

Uncore overclocking instability usually takes 5 hours to occur (BSOD). Normally on 512K test on Prime95. Sometimes, when you almost stable, you get silent Kernel WHEA cache errors in the event log.

mine is if it doesn't crash in the games i play

 

[Knoxx29]

How do you test for stability and for how long?

Prime95 5 hours small FFT, OCCT 5 hours and Memtest overnight, honestly i shouldn't even test those Machines for that long because they aren't built for 24/7.

 

[Regeneration]

* IntelBurnTest is best to test core stability.
* Prime95 version 26.6 Large FFT is best to test uncore stability.
* HCI MemTest (with no pagefile) is best to test DDR stability.
* OCCT isn't bad, but some recent versions are buggy and crash for no reason.
* MemTest86 isn't good enough to stress test DDR overclocking stability.
* Prime95 version 26.6 custom blend with 80% of RAM is best to test everything.
* If you want to go crazy, you can run LuxMark in the background to apply more stress on the IOH and PSU.
* If you don't want to spend much time on stress testing, just go with a long run of IntelBurnTest.
* Don't forget to restart between stress tests since RTL values auto change all the time.

 

[phill]

There you go pal:
View attachment 92780

Is it me or is the CPU load at 100%??

 

[Zyll Goliath]

I like to do 10 runs in Intel burn test sometimes short runs in prime95...also I really love Asus Real Bench because it´s not that long but push various things to the limit.....after that games will do the rest of the Test

 

[Knoxx29]

Is it me or is the CPU load at 100%??

Yes CPU is at 100% load because he is running AIDA64 FPU stress test.

* IntelBurnTest is best to test core stability.
* Prime95 version 26.6 Large FFT is best to test uncore stability.
* HCI MemTest (with no pagefile) is best to test DDR stability.
* OCCT isn't bad, but some recent versions are buggy and crash for no reason.
* MemTest86 isn't good enough to stress test DDR overclocking stability.
* Prime95 version 26.6 custom blend with 80% of RAM is best to test everything.
* If you want to go crazy, you can run LuxMark in the background to apply more stress on the IOH and PSU.
* If you don't want to spend much time on stress testing, just go with a long run of IntelBurnTest.
* Don't forget to restart between stress tests since RTL values auto change all the time.

Thanks for the advice.

It's not necessary but maybe when i am bored i will give it a try, the Evga X58 Machine i will downclock it to 4.2GHz.

 

[Knoxx29]

it was time to give it a break to the CPU.

I downclocked it from 4.7GHz 1.448V to 4.2GHz 1.258V

 

[Zyll Goliath]

it was time to give it a break to the CPU.

I downclocked it from 4.7GHz 1.448V to 4.2GHz 1.258V

View attachment 92991

I am curious how much is the score now in CineBench R15 on that speed?

 

[Knoxx29]

I am curious how much score now in CineBench R15 on that speed?

With those settings for sure very low, unfortunately i have shut down the Machine a few minutes ago, tomorrow i will run CineBench.

 

[Knoxx29]

I couldn't resist till tomorrow

Here

 

[Zyll Goliath]

I couldn't resist till tomorrow

Here

View attachment 92993

That´s still Good&Solid score M8......I am "hovering" at 4005Ghz with my E5645:

And I am more than happy with this speed.......

 

[Knoxx29]

That´s still Good&Solid score M8......I am "hovering" at 4005Ghz with my E5645:

View attachment 92995

And I am more than happy with this speed.......

 

[Knoxx29]

because the 4.0GHz overclock was slightly higher than your I decreased the clock speed a little bit more

 

[Zyll Goliath]

because the 4.0GHz overclock was slightly higher than your I decreased the clock speed a little bit more

View attachment 93009

Thats just about right Score for that Speed ......also whats yours NB(UCLK) Frequency?I am getting a bit higher scores If I rise that but then machine it´s not that stable......

 

[Knoxx29]

Thats just about right Score for that Speed ......also whats yours NB(UCLK) Frequency?I am getting a bit higher scores If I rise that but then machine it´s not that stable......

 

[Regeneration]

I'd try to push the DDR frequency further if I were you.

 

[Zyll Goliath]

View attachment 93010

Yep.....thats it....

When I go on 3300Mhz overall results are 1-2% better but as I said stability issues appear.......

 

[CAPSLOCKSTUCK]

X 58 porn..............dirty girls

 

[Knoxx29]

Yep.....thats it....

View attachment 93011
When I go on 3300Mhz overall results are 1-2% better but as I said stability issues appear.......

I guess that's it.

The real NB frequency for my Overclock 3989.40MHz it's 23x 3800+ but i always use a little less 22x/21x, if i am not mistaken right now i have set at 21x.

 

[Zyll Goliath]

I guess that's it.

The real NB frequency for my Overclock 3989.40MHz it's 23x 3800+ but i always use a little less 22x/21x, if i am not mistaken right now i have set at 21x.

Yep....I guess that NB freq. should be at least 2x speed of RAM so you are just a bit higher than that and its perfectly fine,not sure how much higher you can go before the trouble appears.....nor that I am suggesting anything .......

 

[Knoxx29]

Yep....I guess that NB freq. should be at least 2x speed of RAM so you are just a bit higher than that and its perfectly fine,not sure how much higher you can go before the trouble appears.....nor that I am suggesting anything .......

Right now i am at work but tonight i will give it a try, however the right thing it's not to exceed the target memory frequency.

Here is an example:

Target memory frequency is 2020MHz, 2020MHz x 2 = 4040MHz, as you can see it's set at 3707MHz but i could use it up to 4040MHz.

 

[Zyll Goliath]

Right now i am at work but tonight i will give it a try, however the right thing it's not to exceed the target memory frequency.

Here is an example:
View attachment 93020
Target memory frequency is 2020MHz, 2020MHz x 2 = 4040MHz, as you can see it's set at 3707MHz but i could use it up to 4040MHz.

Well the speed of the uncore (the L3 cache and memory controller) should be at min 2x of the speed of ram in your case thats 3200Mhz but its easily overclockable.....if you don´t have any problems at 3707Mhz try to push it at 4040Mhz and see how much you can gain in tests,the good things is that this OC usually don´t rise the temps(at least in my case) but it´s tricky when it comes to the every day stability.......

 

[Knoxx29]

Well the speed of the uncore (the L3 cache and memory controller) should be at min 2x of the speed of ram in your case thats 3200Mhz but its easily overclockable.....if you don´t have any problems at 3707Mhz try to push it at 4040Mhz and see how much you can gain in tests,the good things is that this OC usually don´t rise the temps(at least in my case) but it´s tricky when it comes to the every day stability.......

I never had any problems running it at 3707, btw that's an old overclock that i am not using anymore, you're right the speed of the uncore should be 2x the speed of the ram and that's why i am not using anymore 1600MHz but instead 2133MHz.

 

[Zyll Goliath]

I never had any problems running it at 3707, btw that's an old overclock that i am not using anymore, you're right the speed of the uncore should be 2x the speed of the ram and that's why i am not using anymore 1600MHz but instead 2133MHz.

Well...I guess....in that case you can even push it beyond 4266Mhz.....