Intel Z87 and Haswell 24/7 OC Guide

[Enterprise24]

nice guide !

 

[The Von Matrices]

Thank you for pointing out to use FPU only in AIDA64 as opposed to the default options. Using that showed that my system is unstable at it's settings that I previously thought worked. Even after I lowered my clock speed to fix the stability issues, then the FPU stress test caused my processor to reach 100°C and throttle. Now I need to reduce the voltage to below 1.2V in order to prevent throttling. These chips really do get hot!

One thing I was confused about was the importance of VCC-in. You explained that lower default values indicate a more overclock friendly chip. But is there any reason to change this voltage from the stock value? The usefulness of changing the value isn't described in the article.

 

[cadaveca]

Thank you for pointing out to use FPU only in AIDA64 as opposed to the default options. Using that showed that my system is unstable at it's settings that I previously thought worked. Even after I lowered my clock speed to fix the stability issues, then the FPU stress test caused my processor to reach 100°C and throttle. Now I need to reduce the voltage to below 1.2V in order to prevent throttling. These chips really do get hot!

One thing I was confused about was the importance of VCC-in. You explained that lower default values indicate a more overclock friendly chip. But is there any reason to change this voltage from the stock value? The usefulness of changing the value isn't described in the article.

Yeah, I was purposefully vague on voltage overall, since it really seems to change from chip to chip, how you should set things. You really want to set iVR voltage as low as possible to help control heat.

VCC-in provides voltage to all domains, so it will drops, or increase, automatically based on workload, unless your BIOS allows manual adjustment options that disables Intel's automatic scheme. A small offset can help pull in stability, at times, regardless, but not all boards deal with this in the same way so there was no general rule I could come up with other than trying to stay under 1.9V, and saying that I think 1.865V is more than enough for 24/7 unless your chip is really sensitive.

I might have been able to come up with something more specific if I had more chips to play with, pushing the limits and killing hardware gets expensive quick, and I've killed a LOT of hardware since this platform launched, CPUs, boards, memory, VGAs....

 

[Nabarun]

...pushing the limits and killing hardware gets expensive quick, and I've killed a LOT of hardware since this platform launched, CPUs, boards, memory, VGAs....

I'm getting so nervous. The ambient temps here are ~35C on an average and more than 40C in summer. There are no "custom" WC solutions available in India. There are some AIO with potential to expand (like the Swiftech, CM Eisberg etc) but I don't know how to do the expansion since the individual parts aren't available....

Do you think the Swiftech H320 AIO in push-pull will be a good-enough solution? I can't find many reviews of the H320...

 

[cadaveca]

I'm getting so nervous. The ambient temps here are ~35C on an average and more than 40C in summer. There are no "custom" WC solutions available in India. There are some AIO with potential to expand (like the Swiftech, CM Eisberg etc) but I don't know how to do the expansion since the individual parts aren't available....

Do you think the Swiftech H320 AIO in push-pull will be a good-enough solution? I can't find many reviews of the H320...

Yeah, the H320 should be more than adequate for 24/7 OC'ing.


I killed a fair bit of hardware, but at the same time, I have kind of been trying to kill stuff, and find the limits. Then, when something dies, I have to try the same thing again to confirm the cause of the death. So I don't think you need to be so concerned about it...just be sure to keep everything within respectable limits as I posted, since maybe a wee bit more is OK.

 

[Nabarun]

Yeah, the H320 should be more than adequate for 24/7 OC'ing.


I killed a fair bit of hardware, but at the same time, I have kind of been trying to kill stuff, and find the limits. Then, when something dies, I have to try the same thing again to confirm the cause of the death. So I don't think you need to be so concerned about it...just be sure to keep everything within respectable limits as I posted, since maybe a wee bit more is OK.

Thanks man. So H320 it is then

 

[tacosRcool]

great article!

 

[GC_PaNzerFIN]

I must have absolutely terrible 4770K. Barely stable at 4.35GHz 1.29V (3900MHz cache) and temps sky rocket in two seconds from 30c to +90c in Intel Burn Test. Contact with H100i is perfect, but can't say the same for IHS and die then.

Yes, I am looking at you 4770K. You and I will visit hammer & vice soon

 

[d1nky]

haven't read it yet but im saving it to favourites for when I eventually try some blue!

thanks in advance....


p.s I just typo'd really bad and said cadavance LOL

 

[bearcub]

Aida/fpu?

Yeah, I was purposefully vague on voltage overall, since it really seems to change from chip to chip, how you should set things. You really want to set iVR voltage as low as possible to help control heat.

VCC-in provides voltage to all domains, so it will drops, or increase, automatically based on workload, unless your BIOS allows manual adjustment options that disables Intel's automatic scheme. A small offset can help pull in stability, at times, regardless, but not all boards deal with this in the same way so there was no general rule I could come up with other than trying to stay under 1.9V, and saying that I think 1.865V is more than enough for 24/7 unless your chip is really sensitive.

I might have been able to come up with something more specific if I had more chips to play with, pushing the limits and killing hardware gets expensive quick, and I've killed a LOT of hardware since this platform launched, CPUs, boards, memory, VGAs....

Great article and thanks...question about the AIDA stability test. I also use AIDA (paired with a G19 for monitoring temps/volts/etc, I can't imagine NOT having that nice little screen to monitor things while overclocking...)

Am I reading it wrong or are there ADDITIONAL settings you use, or do you just de-select everything but FPU? Not 100% sure that makes sense, if FPU is checked along with the others (CPU, etc) doesn't the test include the FPU tests? I turn off the GPU and HDD tests for sure, but why would having ONLY FPU checked be an advantage?

 

[cadaveca]

Great article and thanks...question about the AIDA stability test. I also use AIDA (paired with a G19 for monitoring temps/volts/etc, I can't imagine NOT having that nice little screen to monitor things while overclocking...)

Am I reading it wrong or are there ADDITIONAL settings you use, or do you just de-select everything but FPU? Not 100% sure that makes sense, if FPU is checked along with the others (CPU, etc) doesn't the test include the FPU tests? I turn off the GPU and HDD tests for sure, but why would having ONLY FPU checked be an advantage?

Having the other tests selected uses resources that the AVX-based FPU test needs, in order to be able to run fully. The physical design of the chip makes it this way.

Try both, and you should quickly see the differences.

So, you need to do FPU-only test to ensure that AVX-based stability is true. Likewise with the other tests...to do them properly, they need to be run separately, really.

 

[EarthDog]

So, let me understand... if everything else is selected it doesnt run the AVX instructions? A bit confused and want to ask our contacts at AIDA about that, LOL!

 

[cadaveca]

So, let me understand... if everything else is selected it doesnt run the AVX instructions? A bit confused and want to ask our contacts at AIDA about that, LOL!

You can find Fiery's post here on TPU saying just what I did, when [H]ardOCP first posted their Haswell review.


Here's the thread:

http://www.techpowerup.com/forums/showthread.php?t=184875




You should KNOW this, Mr man.

 

[EarthDog]

Thanks for the link. Read that in the past, and again, and still missed the part where he says "FPU must be the only item enabled to run the AVX/AVX2 instruction set in the stablity test. If others are enabled it will not run those instruction sets even if it FPU is selected"...

Please excuse my denseness...today.

....reading it again, LOL!

THis is as close as I can see:

If you enable only the FPU and GPU subtests, the stress on the system is very high. Using a power draw meter device, you can see that it draws more power from the system than using other stress testing solutions.

But that states you need FPU and GPU? Not sure why he mentions GPU in there for a CPU stress test. Perhaps that is where I am getting confused. I should be at the benching/review rig tonight so I will check it out.

 

[cadaveca]

But that states you need FPU and GPU? Not sure why he mentions GPU in there for a CPU stress test. Perhaps that is where I am getting confused. I should be at the benching/review rig tonight so I will check it out.

Just try it and check the power consumption.

AVX still runs when the other tests are selected, but at the same time, all the tests are contending for CPU time, meaning the AVX instruction set isn't running as fast as it would be under normal conditions. It's not a voltage boost that gives the extra power draw...the voltage boost is there when all four tests are selected, so...ergo the utilization is higher.


GPU test doesn't actually use THAT much CPU time compared to other CPU-and memory focused testing, and adding a GPU itself is usually far higher power usage than even adding a second CPU. To get the max power draw at the wall AIDA can show, you need to do both of those two tests. to get the most power draw out from the EPS 8-pin (or the CPU), you need to not do the GPU test either, and only do the FPU test.

Pick up one of those Zalman power meters, or put a clamp meter on the 8-pin. I report those numbers for a reason! Without that sort of testing, knowing exactly where the power draw is coming from can be a bit difficult, for sure.


Overclocking basics...test everything INDIVIDUALLY. Then, test it all together, in various mixes, as the workload will be different. I don't need the guys at AIDA to tell me any more than exactly what you quoted above...I simply tried all the tests and watched what happened. I could be wrong (which is possible about ANYTHING!!!)...but I don't think so.

 

[EarthDog]

AVX still runs when the other tests are selected, but at the same time, all the tests are contending for CPU time, meaning the AVX instruction set isn't running as fast as it would be under normal conditions. It's not a voltage boost that gives the extra power draw...the voltage boost is there when all four tests are selected, so...ergo the utilization is higher.


GPU test doesn't actually use THAT much CPU time compared to other CPU-and memory focused testing, and adding a GPU itself is usually far higher power usage than even adding a second CPU. To get the max power draw at the wall AIDA can show, you need to do both of those two tests. to get the most power draw out from the EPS 8-pin (or the CPU), you need to not do the GPU test either, and only do the FPU test.

THAT makes sense. Thanks!

I have a KillaWatt and CAMeter so I can see both if needed. Will use KaW though, LOL!

 

[springs113]

Hey Dave for the life of me I can't seem to get pass base clock of 100 and max multi of 45. I wanna lower my multi like you said to about 39 but can't seem to get pass that(multi of 39 and higher base clock) for some reason.

 

[EarthDog]

What baseclock are you trying to reach? Are you hosing on 101 or did you shoot for 125 out of the gate? Details would be helpful...

 

[springs113]

What baseclock are you trying to reach? Are you hosing on 101 or did you shoot for 125 out of the gate? Details would be helpful...

Sorry, been busy(2month old newborn). I have tried in between all the way up to 117 I can't get my chip stable what so ever. I had it boot into windows @ 4.6 but eventually bsod.

msi mpower
4770k
g skill 2400 MHz sniper

I had the latest 1.5 bios and have since turned to the 1.5b4(beta) and have since stabilized. I have booted with little tweaking @4.6 but backed off to 4.5ghz for now till my next day off. I'm idling around 26-27 degrees celsius and maxed out @65 stress testing.

 

[Footman]

Wouldn't mind seeing Devils Canyon z97 guide.....

 

[cadaveca]

Wouldn't mind seeing Devils Canyon z97 guide.....

You will, soon. I only have had some chips so far, testing every voltage and setting across several boards with several CPUs takes time.