First attempt at haswell oc 4770K

[ratjacket]

This seems to easy, I must be doing something wrong

i7-4770K L316B317

GA-Z87X-OC

NH-C14 cooler

first oc test

45 multi
set vcore to 1.210
VRIN LLC to extreme
VIN on auto which is 1.8

disabled HT for stress test
disabled turbo, oc didn't work with turbo on auto
iGPU disabled

all power saving states etc on

Ram on xmp2

is this correct CPUID HW monitor doesn't seem to display correct VIN it gives the same high number at stock

during stress test cpu stayed stable at 4.50GHz

not 100% stable tested but seems like I have a lot more headroom

how high would you guys let the temps go for 24/7 oc but with power saving states enabled

 

[ratjacket]

hmmm 4.6 is a lot more difficult to achieve

all settings the same except vcore 1.250 & 46 multi

and aida64 BSOD after 5mins

do I have to bump the VIN higher than stock 1.8 when at 1.25 vcore or is my chip just going to be a 4.5GHz for practical 24/7 OC

temps are rising fast the more vcore I push

1.254 failed even faster :/ next step 1.26 I think I may just end at 4.5GHz at this rate

at any rate il run aida overnight to see if 4.5 is truly stable

1.270 survived 5mins max temp 80C not sure if its worth it for 100MHz

what do you guys think? 4.50 or go for broke and push 4.70 (not sure my cooling will cope would probably need 1.4 vcore)

maybe I should pop the lid

 

[ratjacket]

I just noticed my uncore seems to be auto overclocking to 4000

so I set ring voltage to 1.15

and now I am getting better results

4.6GHz @ 1.24 don't thick its stable but it hasn't BSOD on adia64 yet

 

[JimmiG]

Don't forget to run Aida64 with only FPU checked to verify temps.

 

[ratjacket]

would this generate worse case scenario

I notice that even Adia64 (CPU,FPU,Cache) with HT disabled still puts more vcore than i have asigned in BIOS

4.60 GHz @ 1.240 vcore survived 2hours of Adia64 but over night while i slept it BSOD so i guess i need to just bump it a fraction more for 4.60 stable

how high Vcore would you guys go safely for 24/7?
I ask this because I could just stick with the lower vcore (1.210) 4.5GHz or try push for 4.7

With FPU mode only how high would you let max temp hit?
Considering when i had the stock fan and settings (3.7 turbo) the cpu hit 100C and throtled within 1min of adia64

Does FPU only mode stop the AVX overvolting?

It use to be alot easier when you could just prime95 for 24-48hours and be done (and only have to worry about vdroop) :/

Has anyone else got a GA board and noticed the Uncore at 4000 even when set the Uncore @ stock once you oc the cpu multi

 

[ShintaiDK]

With AVX2 load the vcore gets bumped woth 100mv. That might be what you see.

 

[ratjacket]

yes i suspect this is so, so in the real world is this a valid estimation of max temps?

does FPU only mode stop the AVX2 overvolting?

 

[ChronoReverse]

Try the newest Prime95 (with AVX) in smallFFT mode.

In some of my more adventurous undervolting while overclocking attempts on my i7-4770k, AIDI64 would be safe as long as I want and it would also pass IBT. But Prime95AVX would soon reveal the problems even if temps don't hit 90c.

 

[ratjacket]

I remeber reading that prime 95 was unsafe to use with haswell

is the new prime95AVX safe to use with haswell?

I guess it overvolts aswell

can it be used with HT enabled?

 

[ChronoReverse]

It can always be used, it's just that it represents the highest possible load you'll see and "unrealistic" in that you'll almost see as heavy a load in practical use.

When grinding AVX, the voltage is automatically bumped up a notch though.

 

[ratjacket]

Sweet so i can go back to prime95

Do i need to disable Hyperthreading while stability testing, i dont think it loaded the cores correctly in adia but it might on prime?

Small FFT for 24 hours followed by blend for 24 hours to almost ensure 100% stability?

 

[ChronoReverse]

Might as well run AIDA64 too because it tests some things besides just the CPU. You can let Prime95 load up the virtual cores too. There's no reason not to if you want to simply max loads.

 

[tweakboy]

You need 1.30v for stability and never crash.

 

[ratjacket]

I was doing something wrong

p95v279 small fft

much harder on cpu upto 1.260 vcore on 4.5GHz slowly increasing in length as I bump

101 bsods
last one was longest and eneded in 9c BSOD

are you sure prime is safe

1.27 vcore made it to 30mins before I stopped it, getting pretty hot but today is worse case very hot day and I still need to get 2 x more case fans

maybe I have to go to 4.40 boooooooo

 

[ChronoReverse]

The question you have to ask yourself is how certain you want to be that your system is stable. Even though Prime95AVX pushes your system to the limit, it's still within specifications (if you're on a stock CPU).

So if your OC isn't prime stable, then you have to decide whether "everything else mostly stable" is good enough.

 

[ratjacket]

good point but I have always been a fan of prime on older systems

weird 1.22 vcore 44 doesn't boot 45 does

 

[Kenmitch]

Why did you pay xtra for the 4770k if your going for overclock stability with HT disabled? I'd think it would be best to have it enabled during stability testing if you plan on running it with HT enabled.

 

[ratjacket]

the reason i disabled HT was when i was first testing the system with the stock cooler adia would only load two cores becuase of hyperthreading this resulted in the the system turboing to 3.9 and overheating if i disabled hyper threading it would load all four cores and only turbo to 3.7

I am now running Prime with HT enabled

On another note the GA-z87x-oc bios is buggy

after messing around with 45 multi if i try drop it to 44 it wont except this in windows i have to reload defaults reboot to bios then set the multi vcore etc

it might be the fact it trys to boot 44 with too much volts but that doesnt make sense

anyway it was anoying until i found out how to circumvent it

now i have droped the ram xmp and let it ran at low speed deafult, droped my vcore to 1.250 multi 44 ring to 1.150 (casue the stupid bios ocs the uncore to 4000) vrin to 1.8 (default) and are stress testing prime at home while i am at work small fft

this was after i gave up on 4.5 i was at 1.28 vcore 1.9 vrin 1.150 ring and ram at 1600 and it only done 1 hour of prime small fft before it restarted :/ (97 was the hottest core max temp before that happend it was at 85ish before it restarted)

if i can stablise 4.4 i might try 4.5 again with some more juice to the vring incase its the uncore

 

[ratjacket]

that went better 9hrs stable small FFT

time to tweak some more

I will try set the uncore so it doesn't auto oc, add see if my fail 4.5 were to do with buggy bios

wow that was insane turning uncore to 36 and multi to 45 ht on made prime have a massive jump in temps weird I didn't think GHz made a big diff on haswell for temps



confirmed bios buggy as
maybe I need to reflash this will be why the guy sold the mobo 2nd hand so quickly

need to reload a profile or defaults bfore changing any setting or changes are erratic

 

[ratjacket]

hmmm maybe I should smash this mobo

drastically reduced voltage down to 1.210, uncore turned down to 36 so it doesn't auto oc, ram set low on default

massively increased temps, all be it its hotter in the day now but a reduction in 0.4 in vcore should more than compensate for 5C ambient differences

what do you guys think

should I go punch Gigabyte in the head

 

[Headfoot]

set settings to stock and flash the newest revision of the bios, see if that helps.

 

[ratjacket]

thats the plan thanks

 

[JimmiG]

hmmm maybe I should smash this mobo

drastically reduced voltage down to 1.210, uncore turned down to 36 so it doesn't auto oc, ram set low on default

massively increased temps, all be it its hotter in the day now but a reduction in 0.4 in vcore should more than compensate for 5C ambient differences

what do you guys think

should I go punch Gigabyte in the head

In your Prime95 shot above this post it says FFT Length 560K, in the second one it says 12K. This makes a massive difference to the temp.

Personally I don't care about the temps any more. Since I run Adaptive Vcore, it shoots off into the mid 1.3x range when running AVX-heavy loads, causing temps close to throttling even after delid. However in games the temp is in the mid-high 60's C.

 

[Headfoot]

thats the plan thanks

Yeah sorry if it seems duplicative. The reason I call it out again is because I've read elsewhere on the great interwebz that Gigabyte in particular had some pretty crappy early releases bioses, worse than other manufacturers.

 

[felang]

If you ever want to enable HT, then you have to stress test with HT enabled. It increases temps and needs higher volts