some update
now manage to get same 4.5g prime stable with lower vcore, 1.195v with higher dram speed, 2133MHz @ 10-12-10-27, 1T, 1.525vdimm.

why not prime 95 28.5? actually whats the correct stressing tool for haswell? its kinda hard to pass stability test on haswell with 28.5

Right now I really have no idea which is the best stressing app for Haswell/DC etc.
Newer Prime95 version might require higher vCore but it doesn't seem to stress other stuffs such as IMC, RAM etc.
LinX is good because from the GFlops values you would know that your OC is not "underperformed" than it should be but until now I'm still not sure whether the current version shows the correct GFlops. It may not stress the HT correctly.
I believe it's been debated until now. Some say the best stressing tool is the one which is causing the vCore to drop to the lowest point under load. Hmm...

Hi bros,

Just wondering.. you guys use offset or fixed vcore method for 24/7 usage?
I'm planning to get at least 4.7g or 4.8g for 24/7 use. I don't really want to manually feed 1.25-1.27v static vcore for a prolonged period as it might just kill the proc prematurely.

any ideas ?

yup.
over the years i had many so call prime stable, linx stable etc all fail when i do long encodes..

the issue is always not the vcore or cpu on high clocks.

as cpu goes up, other factors come in and mainly its pll, switching frequency etc etc especially with high dram speeds, low timings, low imc timings

overvolting vcore will mask this instability but true colors will be shown in encoding etc especially with avx instructions etc.
or even chrome on its mem paging and swapping.

hence i just run a few test but more importantly is always imc/ram stability rather than cpu vcore.
and always check scaling. for me its 3dmark11 physx. each multi must increase atleast 200-250 points and must be consistent after 10 runs. if not than theres some issue.
gsan
so 3dmark physx for 4790k should be.
4.5ghz 13000, 4.6ghz 13250, 4.7 13500, 4.8 13750 etc.

it will be sweet to get a 14k on everyday running comp. estimated 4.9ghz. but i forsee not possible atm with the current cooling or just some limitation somewhere. it could be even the bios was designed for a specific type of cpu scaling voltage etc. high dram speeds will show more instability and not necessary its always vcore. hence the initial guides of haswell are all wrong. owikh test may not be so vcore prone but its more stressfull as its stressing all the other voltages together. heat and thermal limits also test cpu leakages etc and hence with the inbuilt IMC u may see other failures that are not vcore related.

its very vague but ultimately all test can only be perform on better cooling. Even phase has its own issue with haswell.

This post has been edited by cstkl1: Aug 1 2014, 02:30 PM

Easy......(bootable)



This post has been edited by stasio: Aug 2 2014, 03:38 AM

24/7 usage mean full load all the time? if your proc need 1.25-1.27v for 4.7/4.8g stable, there is no other way to lower down the vcore.

Hi guys. i'm new here. Making new friends sharing the same interest. I'm in the level of beginners of overclocking..
And all experts advice is very much appreciated.
Currently using MSI's Z97 board and pentium G3258 and i'm indeed surprise how well it performs. For only two cores can performs very well when overclock over 4GHz.
Especially in gaming.
I'm looking forward for my 4790k to arrive and see how well the devil's canyon performs against haswell's 4770k.

i can only manage to overclock the G3258 to 4.3GHz on air. i'll try to push harder and see how it goes. Thanks.

My G3258 ES is able to overclock to 4.6GHz only, stable @ 1.488v!
4.7 onwards is unbootable
Max clock is 4.8 in OS but at excessive voltage



Anyway it inherits the Haswell’s strong IMC
2800c9 is easy with Samsung kit



Things going nuts at 4.6GHz when I had to bump the VCCIN up to 2.2v jn order to pass x264 benchmark lol

Lemon chip is everywhere especially the ES lol

This post has been edited by owikh84: Aug 4 2014, 10:41 AM

Maximus Vi Price all dropped at czone

go grab ppl. M6e now Rm1099

out now is M7F at RM1720.

cstkl1

i've been having 0x101 BSOD at 4.7ghz @ 1.295v running Cinebench R15
it passed AIDA for 30 minutes tho

what do u think is the culprit?

Tried upping the VCCIN already? 1.9v etc

already at 1.9v and i start from 1.7v

Read on OCN and hwbot......"Suggestions for DC overclocking"==> LOWER VCCIN....

http://forum.hwbot.org/showthread.php?t=108373

Edit:

4.9GHz with 1.50 VCCIN.......



This post has been edited by stasio: Aug 6 2014, 06:52 PM

Just curious, would that 1.5v VCCIN stable for X264 benchmark

hilmiangah

i've been having 0x101 BSOD at 4.7ghz @ 1.295v running Cinebench R15
it passed AIDA for 30 minutes tho

what do u think is the culprit?
Change the CPU pwm frequency and dram to be the same, try 300,300 and keep increasing till u see a difference in stability. Retest with 3dmark11 performance Physx. Lower the PPL voltage to 1v.
4.7ghz physx score should be 13500 atleast.

but hilmi i can only stipulate until my wc setup is done. So far got everything except for tubing and fans. Still hitting my head sometimes as could have bought two more titan blacks for all this wc gear.
whats ure cpu vid?? (asus to check, go into bios, load optimize, save and reboot. enter into bios and see whats the vcore is running at)
up to 4.7ghz a decent 4790k.. the variance on benching and actual stable voltage can be up to 0.07v

stasio


i do that on one 4770k when i first got it. but after it died.. i didnt try that anymore.

hmm intel must have a reason for the +0.4v difference.

but its a good idea to find out why after 1.3v onwards the issue seems to be vccin related always. when i reduce my x-talk i get better stability at 5ghz. and when i say reduce.. its a ridiculous number that a 4770k would most probably die.

also this suggestion .. like a hole.. like how pll voltage between 1.25v to vccin .. its unstable.
hmm come to think of it.. pll and vccin.....i think the vccin lowering it makes the pll stable. theory

This post has been edited by cstkl1: Aug 6 2014, 09:14 PM

cstkl1

Here's the vid from initial installation



will try out your suggestion and report back here soon

eh bro.. ure gonna need more than 1.28v.
i estimate u need 1.175v for 4.4ghz ( btw check using hwinfo64 on the vid for all core later on after u set it manually)
so by that estimate.. 1.225 for 4.5ghz and 1.275v for 4.6ghz and 4.7 should be 1.325v.

4.7 in DC.. can run at lower voltage .. infact DC only shows its true colors of difficult clocking from 4.8 onwards.

Turbo (4.4 GHz) must be enabled, then you will know your real VID.

This post has been edited by stasio: Aug 7 2014, 12:52 PM

Turbo is enabled