4.2 to 4.4 ?

Associate
Joined
26 Oct 2009
Posts
93
Location
London
ok ive got my rig nice and stable at 4.2 now.. prime overnight, linx 20 passes on max etc..

the temps peak at about 65 degrees (watercooling) and ive got the vcore set to 1.36v (in windows i get 1.32v, drops to 1.30v under load)

i feel it could go to 4.4, but ive tried raising the vcore all the way up to 1.435v (1.4v idle in windows, drops to 1.39v under load, and also raised the qpi voltage to 1.4v (currently on 1.35v at 4.2)

the temps start creeping into the high 70's under load, (46 - 50 degrees idle eek!)

and its not IBT stable, i either get a checksum mismatch on the second round or a bluescreen.

any advice on how to get it this last step?

i dont wanna keep cranking up the vcore and getting silly temps if i can tweak something else.

other non standard voltages i have already are:

ioh core 1.88v

cpu pll 1.2v

ram 1.66v
 
I suspect you've got ioh core and pll muddled up there, as ioh core starts at 1.1 and pll at 1.8.

I think the issue lies with uncore ratio or with clock skews, throwing more voltage at it isn't getting me very far either.

I'm currently trying to force 210 stable by throwing voltage at it. Not great success, though I'm in windows it's definitely not stable. Do you have turbo turned on/off? If it's off I believe you'll have a lot more luck with a lower bsck and it turned on.

edit: got 210x20 stable for 15 passes of ibt. Flicked turbo on and trying for 4.4. If ibt passes I'll probably leave f@h running, ibt + f@h stable would do for the time being. I can't offer much advice beyond saying that vcore appears to have been the key, along with making sure the ram is running well under spec.

edit2: Now appear to be ibt stable at 4.4, using turbo. Plan is to leave it folding at this for a while to stave off Dekez, then try for 4.4 without turbo. I've set 1.4V vcore in the bios, seeing 1.33V under load. QPI set to 1.315, no idea what it's actually at after vdrop. Temperatures hit 65, hottest I've seen them yet.
 
Last edited:
ahh yes i expect i got them muddled.. i was writing a little while after my last bios visit.

my watercooling consists of an ek supreme cpu block, a black ice triple rad, 3x noctua 120mm fans and a ddc 18w pump and res combo.

yes i have turbo on, had it on since i got it.. trying to get 210 bsck and it boots fine into windows at that, but just craps out when i stress test (but pretty quickly so its definitely not what id call stable.

where is a good place to find out about i7 clock skews? i could drop the uncore ratio, but then id have to drop my ram down to 1200, which kinda sucks.. ill try that tonight though.

any other pointers?
 
IOH core? is that vtt? If it's vtt intel recommends around a .5v differnece between this and vdimm, although if you need less don't raise it to make this difference. It's about stabilty iirc

I have an IOH voltage is bios, can't recall exact name but default if 1.1v. Recommondation over at xs is this needs to be upped to 1.24v for 200bclk. I've got away with it at 1.2v.

This a gigabyte?
 
yeah its a gigtabyte ud5 mobo..

running my IOH at 1.22v

vtt is qpi voltage i believe.. mine is at 1.35v and my vdimm is at 1.66v, so currently a
.31v difference.
 
lol i think we are all getting confused now. do you mean ioh or pll voltage? my pll is already up from 1.8v default to 1.88v my IOH is up from 1.1 to 1.25v

thanks for the advice though, ill try that tonight when i get home.
 
lol i think we are all getting confused now. do you mean ioh or pll voltage? my pll is already up from 1.8v default to 1.88v my IOH is up from 1.1 to 1.25v

thanks for the advice though, ill try that tonight when i get home.

if that was to me i mean CPU PLL mate intel max rec is 1.88, auto is 1.8 i use 1.3 :)
 
hmm i also read on another forum after this somone who dropped their pll voltage rather than raising it... this is the first times ive heard of people dropping a voltage to get a higher clock... is this something ive missed?
 
Nah, on P45 I had to decrease a few of the voltages as fsb went higher. Decreasing PLL to 1.3 hangs the bios for me, I can't set it below 1.6. Dropping it to 1.3 has helped a fair few people out though, I should try 1.6 and 1.7.

Still can't find any information on the clock skews online.
 
hmm yes ive tried dropping the pll to 1.3.. i also tried a cpu clock skew of 100, but to be honest it didnt seem to make a blind bit of difference, positive or negative.. man each chip really has its own little personality no?

ive given up trying to get from 4.2 - 4.4 for now, the performance difference just isnt worth it if i have to crank the voltages/temps so much.. im after a fast workstation not a benchmark winner.

just a bit frustrated that going from the noctua air cooler to a pretty beefy watercooling setup has netted me only a 200 mhz extra clock. (its noisier too...) maybe i should be braver with my voltages.. just when i see the idle hitting 47 degrees and load hitting 77 on such a powerfull cooling setup.. i get nervous about longevity..
 
How long has the loop been running? I'm seeing much better temperatures than that with a similar system. I'm wondering if there's air in yours, or if the block mounting is a bit off. It'll only be louder if you're running all the fans at 12V, once stable you can experiment with slowing them down to reduce noise.

I think the problem is turbo. The instability is too odd, with too little information available through Google. Mine is so close to stability, yet anything on the voltages doesn't help. Turning off turbo does though. So I'm going to aim for 220x20 rather than 210x21, if I see the same issue with turbo off I will revise this belief.

You can be pretty aggressive with voltages if it's cold. I'm quite happy pushing 1.425V at 60 degrees. However I'd be a lot less keen at near 80 degrees. Until you've found out what's going wrong in the cooling loop I'd leave the voltages about the same as on air. Up to you though.
 
to be honest actually i noticed the temps are a little lower these last few days... i tried at 4.4 again this eve (after saying id given up..huh) and it was idling at 38 and going to 69 under load..
i actually managed this eve to get it quick IBT test (standard but with 15 threads) stable, when i dropped the qpi multi to 13x (from memory) and the ram to 6x... but then the ram is around 1250 or something, which annoys the hell out of me.. and a quick test render in mental ray showed it was only 2.1 percent faster than 4.2 ghz with the ram at full whack. probably not worth the extra strain on the chip..

i then thought.. hey, if its stable if i drop the ram, maybe i can speed the ram up a bit by going to 4.6...lol.

it boots at 4.6 with 1.425vcore...sometimes.. and crashed out of ibt immediately.. so ive gone back to nice stable cool 4.2.. at 1.38v

idles at 35 load is 65...

u still think this is too high for my cooling system? dont think i have air in the loop.. its been about 1 month .. no bubbles.. ran the loop outside for an hour.. radiator at all angles, and i installed a backplate and clamped the ek block down scarily tight.. just a drop of AS5 .. i am running the 3 noctuas at max rpm at the mo so i can slow em down once im happy with the setup..


the only possible weak point of my setup is the rad.. its a 6 odd year old Hardware Labs Black Ice Pro III from ocuk that i salvaged from my old overclocked fx-55 setup.. its a triple rad, and i used to have it cooling a 150w cpu and a 6800 ultra gpu (quite a power hungry card iirc.) now its only cooling a 150w cpu, but i guess things have moved on in 6 years!

let me know how you get on with/without turbo.. cheers!
 
you wont even notice the difference, i would be happy at 4.2ghz which im currently at, benchmarked at 4.5ghz and didnt see a massive boost, in the end i went back to 4.2ghz knowing my voltages and temps are safe ;)
 
oh just out of interest JonJ.. you say you are happily at 1.425 and 60 degress... is that 1.425 in bios, or under load in windows? i ask because to get 1.425 in windows id have to set it to like 1.46 or 7 in bios...
 
Back
Top Bottom