5930k Overclock

[Johnny5]

Hello All,

I wanted to share my overclock and hopefully get some feedback on some possible tweaks.

Core Speed - 4.6
CPU Core Voltage - 1.36250
DRAM V - 1.2
XMP - 2666mhz
Multiplier is 100x46 sync all cores
CPU Cache Voltage is 1.2v
CPU Cache Ratio - 24
BCLK - 100
Temps - ~mid 30s idle with pump set to trickle and rad fans running at low rpm. Max temps have reach mid to upper 50s.

System specs are in sig. I like to run this 24/7, gaming, and the occasional benchmark.

Any thoughts or feedback would be greatly appreciated.

Thanks!

[ERIU-OC]

Lelouchvi wrote:
1.770 input
Just a typo 1.16 =p
at 4.2Ghz 100x 42

2666 g skill ram xmp settings

idle around mid 30s on the cpu and 40 on the gpu

loads temps just touching 60s with realbench after 4 hours of it. and 47 on the gpu

looking really good right now, keep on tuning!
you might wanne find out if your OC is stable when using AVX instruction set at this point?!
if so run prime95 (Ver. 27.9) on custom FFT length 1344, 1344 FFT length is usually good for checking Vcore requirements and FFT length 448 for checking input voltage requirements

EDIT: pay attention on your temps when using prime95, it produces more heat than realbench...

if you run into BSOD:
0x124 Weha uncorrectable blabla -> usually more Vcore is needed
0x101 Clock Watchdog blabla -> usually to high or low input voltage
these are recommendations to watch on and not to be taken as hammered in stone

Cheers

[Lelouchvi]

ERIU-OC wrote:
looking really good right now, keep on tuning!
you might wanne find out if your OC is stable when using AVX instruction set at this point?!
if so run prime95 (Ver. 27.9) on custom FFT length 1344, 1344 FFT length is usually good for checking Vcore requirements and FFT length 448 for checking input voltage requirements

EDIT: pay attention on your temps when using prime95, it produces more heat than realbench...

if you run into BSOD:
0x124 Weha uncorrectable blabla -> usually more Vcore is needed
0x101 Clock Watchdog blabla -> usually to high or low input voltage
these are recommendations to watch on and not to be taken as hammered in stone

Cheers

Thanks, I got it pretty stable now, but core2 and 4 on HW monitor shows a large temp gap between the other cores which got me worry as they go up to 9 degree higher than the others sometimes. Realbench didnt appear to have similar issue.

I also found this which was an interesting read.

http://tatoismus.net/asus-x99-deluxe-i7-5930k-prime95/

[Chino]

Lelouchvi wrote:
Thanks, I got it pretty stable now, but core2 and 4 on HW monitor shows a large temp gap between the other cores which got me worry as they go up to 9 degree higher than the others sometimes. Realbench didnt appear to have similar issue.

The difference in temperature from one core to the next depends on your mount.

[ERIU-OC]

Lelouchvi wrote:
Thanks, I got it pretty stable now, but core2 and 4 on HW monitor shows a large temp gap between the other cores which got me worry as they go up to 9 degree higher than the others sometimes. Realbench didnt appear to have similar issue.

I also found this which was an interesting read.

http://tatoismus.net/asus-x99-deluxe-i7-5930k-prime95/

differences from 6c up to 10c are not unusual and have been seen many times.
on my chip I have 6c to 8c difference aswell, depending on how I mount the block and what thermal compound I use...

[Lelouchvi]

I see, is just that when i use realbench the higher temp cores are different to the ones compare with prime95

[ERIU-OC]

well realbench is a completely different thing compared to prime95.
realbench put's crossloads on your chip and bus's to simulate more realistic loads like in normal use but more extreme
because of this cross loads the temperature behavior is a complete different thing there.
Prime lets your chip do the exactly same thing over and over again especially when running one FFT length. the only thing that changes is where the data fits in (L1-L2-L3 Cache, memory) with the different fft length.
Hope this clears it up a bit, for you...

[Lelouchvi]

ERIU-OC wrote:
well realbench is a completely different thing compared to prime95.
realbench put's crossloads on your chip and bus's to simulate more realistic loads like in normal use but more extreme
because of this cross loads the temperature behavior is a complete different thing there.
Prime lets your chip do the exactly same thing over and over again especially when running one FFT length. the only thing that changes is where the data fits in (L1-L2-L3 Cache, memory) with the different fft length.
Hope this clears it up a bit, for you...

In a way it does, hahaha was reading this link here for other settings.Good to know that it is normal, and I didnt mount the block weirdly. it seems this guide is for extreme LN2 clocking, I follow some initial settings like cpu spread spectrum, enable and disabling tabs etc on the first chapter. but didnt set any manual voltages. What do you think?
useful insight or just up to user perference?

http://pan.baidu.com/s/1gdejLuz

[ERIU-OC]

it comes down to what you wanne do with your chip.
in the guide you posted are many useful things to look on and the guide is mainly meant as a guide line for those who want to squeeze the last drop of blood out of their chip.
if you want a 24/7 system with some benefits of the features this platform and chips have, you probably don't want to set everything possible to the max or extreme and disable everything that has a function depending on the current cpu load requirements.

maybe this is an example:
your chip is the driver of a car on the road, he can drive faster or slower depending on the traffic, he can choose a higher gear on a given speed to save fuel, or he can choose lower gear to have quick access to acceleration depending on the current situation.

what the guide does is to put the driver in a car on a single lane racing track, giving him only one specific gear , no break and the gas pedal nailed to the bottom. so the only choice is to drive in the hope of being able to handle the speed...

with that being said I mean user preference starts when you choose what to do with your chip.

Cheers

[Lelouchvi]

ERIU-OC wrote:
it comes down to what you wanne do with your chip.
in the guide you posted are many useful things to look on and the guide is mainly meant as a guide line for those who want to squeeze the last drop of blood out of their chip.
if you want a 24/7 system with some benefits of the features this platform and chips have, you probably don't want to set everything possible to the max or extreme and disable everything that has a function depending on the current cpu load requirements.

maybe this is an example:
your chip is the driver of a car on the road, he can drive faster or slower depending on the traffic, he can choose a higher gear on a given speed to save fuel, or he can choose lower gear to have quick access to acceleration depending on the current situation.

what the guide does is to put the driver in a car on a single lane racing track, giving him only one specific gear , no break and the gas pedal nailed to the bottom. so the only choice is to drive in the hope of being able to handle the speed...

with that being said I mean user preference starts when you choose what to do with your chip.

Cheers

Thanks for the explanation =p

I like this car metaphor hahaha as I drive myself, it makes alot of sense. hows your built coming? complete yet?

[LiveOrDie]

here what my chip can do 4.4ghz @ 1.26v or 4.5Ghz @ 1.31v ill be staying with the 4.4Ghz overclock.

8 hour pass LinX 0.6.4 .