cant get a stable overclock.

[skuzzzzy]

ok so another user on overclock.net forums has the same motherboard and cpu as me, so he's been trying to help me oc my cpu to 4ghz. i cannot seem to get it stable, within 1:40-1:50 running linx with max memory it errors and stops.

ive been using higher vcore then him because i was going to work my way down to see what the lowest vcore i could use that was stable.

heres some pictures of his settings then ill post what i used, so maybe you can help me figure out what to try.

http://img13.imageshack.us/i/dscf3757o.jpg/
http://img841.imageshack.us/i/dscf3758.jpg/
http://img237.imageshack.us/i/dscf3760.jpg/
http://img210.imageshack.us/i/dscf3762v.jpg/
http://img185.imageshack.us/i/dscf3763.jpg/
http://img830.imageshack.us/i/linxmaxtemp.jpg/

i used same exact settings but i used, 1.38v on dram due to me having 1.35v ram (when i enter 1.35v to switchs to 1.38), and thats not the culprit because when i reset bios to stock and ram to 1.38 it ran linx fine.

i also use 1.32v vcore, 1.30v cpu vtt, everything else is identical. after that i tried "high vdroop", 1.35v vcore, 1.32v vtt and after loading windows i opened linx and bsod'd. after that i reset bios and still kept bsoding after starting the computer with default settings. after clearing cmos it's working again, so thats where i am at the moment.


my specs are,
intel i7 860 cpu
msi p55 gd80 motherboard
g.skill eco series 1600mhz ram

 

[ModestGamer]

every cpu is a bit different and even small differences in batch consistency of the silcon can impact vcore demands for a particular clock speed. it just varys. 2 chips made on the same wafer or from the same batch of metrials should be the same but 2 chips made from 2 differing batchs of material could be just a bit different.

 

[skuzzzzy]

every cpu is a bit different and even small differences in batch consistency of the silcon can impact vcore demands for a particular clock speed. it just varys. 2 chips made on the same wafer or from the same batch of metrials should be the same but 2 chips made from 2 differing batchs of material could be just a bit different.

i realize this, why im using higher vcore then him. just trying to figure out what to try next to make the 4ghz clock stable.

 

[ModestGamer]

maybe more cooling and if the chip can tolerate it, a higher vcore. it also just might not be able to do it.

 

[skuzzzzy]

maybe more cooling and if the chip can tolerate it, a higher vcore. it also just might not be able to do it.

only reachs 75-79c, running prime95/linx using 1.32v vcore.
im using a silver arrow heatsink and a corsair hx850 psu, so it should be cooling or power.

 

[ModestGamer]

only reachs 75-79c, running prime95/linx using 1.32v vcore.
im using a silver arrow heatsink and a corsair hx850 psu, so it should be cooling or power.

push up the vcore a bit more, also make sure that the ram can handle the fsb

 

[skuzzzzy]

push up the vcore a bit more, also make sure that the ram can handle the fsb

what should i put the vcore at? and not sure how to know if the ram can handle the fsb? first time overclocking.

 

[ModestGamer]

small upward movements and keep a eye on temps. you'll wanna look at the chip datasheet for the maximum voltage.

 

[john3850]

http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-lynnfield

Read more

 

[skuzzzzy]

as of right now, im using memory ratio 5, makes memory run at 1674mhz. 167 base clock for 3.5ghz. 1.23v vcore, and pll is on auto (1.8v), vtt is on auto (1.126v), dram is on auto (1.38v).

passed 10 runs of linx with max memory selected and max temp was 70c.

what should i do next for trying to work my way up? leave pll/vtt still on auto or raise them up? if so raise to what?

raise base clock while using same vcore or raise both? heres a picture when i ran my first run on linx so you can see load vcore etc.

http://img838.imageshack.us/img838/7340/87056539.png

 

[fleflikr]

your vtt will probably need to be increased considerably. no further than 1.4. out of spec but more than likely needed. also hyper-threading enabled is questionable but you may get lucky.

 

[MrK6]

Going over 1.21VTT will significantly shorten the life of your chip, I wouldn't advise it, and you shouldn't need to. The settings I use to get my system at 4.0GHz is 190x21, Vcore 1.344V (1.2V stock + 0.1875V dynamic with LLC off), VTT 1.21V, and CPU PLL 1.8V. I'm not sure if MSI boards allow it, but see if you can get the 22x multiplier working by enabling Turbo (and disable the C3/C6 states so that it doesn't use anything higher than 22x) that way you can use a lower BCLK for 4GHz.

 

[skuzzzzy]

http://img716.imageshack.us/img716/8913/38v2.png

using,
1.28 vcore
1.270 vtt
1.81 pll


ill lower the vtt to 1,.21 but any idea whats makin it unstable at 3.8?


also did a prime 95 test,
http://img215.imageshack.us/img215/8945/38v3.png

ran that using the prime 95 the one that doesnt test ram i believe, not sure if it was ran long enough, but passed first 10 tests.

 

[skuzzzzy]

http://img96.imageshack.us/img96/9201/38v4.png

set vcore to 1.29 then it changed to 1.287
set vtt to 1.21 because you suggested it, afterwards it changed to 1.206.
cpu pll 1.81
set dram to 1.4 but then it changed to 1.455.
and rams speed is at 1448mhz. qpi i think it was called is set on 16.

 

[MrK6]

http://img96.imageshack.us/img96/9201/38v4.png

set vcore to 1.29 then it changed to 1.287
set vtt to 1.21 because you suggested it, afterwards it changed to 1.206.
cpu pll 1.81
set dram to 1.4 but then it changed to 1.455.
and rams speed is at 1448mhz. qpi i think it was called is set on 16.

Well, as you can see, lowering your VTT didn't hurt your stability anymore and you'll also extend the life of your CPU . I would just raise the Vcore more. I'm at 1.360V under load according to CPU-Z @ 4GHz, so you're well below where I am. I'd also keep your RAM multiplier low like you are until you get the CPU stable. Also make sure your QPI stays below ~6.4GHz to keep everything stable. Keep using LinX to test, you should be able to do at least 3 hours or so.

 

[skuzzzzy]

Well, as you can see, lowering your VTT didn't hurt your stability anymore and you'll also extend the life of your CPU . I would just raise the Vcore more. I'm at 1.360V under load according to CPU-Z @ 4GHz, so you're well below where I am. I'd also keep your RAM multiplier low like you are until you get the CPU stable. Also make sure your QPI stays below ~6.4GHz to keep everything stable. Keep using LinX to test, you should be able to do at least 3 hours or so.

another user at tomshardware suggested i run memtest overnight tonight just to make everythings ok with ram just incase its the culprit.

 

[MrK6]

another user at tomshardware suggested i run memtest overnight tonight just to make everythings ok with ram just incase its the culprit.

You can if you want, and it wouldn't hurt, but if your RAM is that bad you wouldn't be booting into Windows. RAM typically errors out very quickly when you push it, and has a very small range of stable->failure, unlike CPUs (IMO). A 30 minute, single pass of memtest is all you really need to verify whether the sticks are bad or not.

Like I said, the easiest way to stablize your overclock is to boost the Vcore. Your temps look fine, so boost it to at least 1.35V, boot up at 4.0GHz, and see how long you last in LinX. This is just taking an educated guess, getting a result so you can see the system's standing, and then adjust accordingly. I call it "guess and check overclocking," and it's typically the fastest and easiest way to find a stable clock.

 

[skuzzzzy]

You can if you want, and it wouldn't hurt, but if your RAM is that bad you wouldn't be booting into Windows. RAM typically errors out very quickly when you push it, and has a very small range of stable->failure, unlike CPUs (IMO). A 30 minute, single pass of memtest is all you really need to verify whether the sticks are bad or not.

Like I said, the easiest way to stablize your overclock is to boost the Vcore. Your temps look fine, so boost it to at least 1.35V, boot up at 4.0GHz, and see how long you last in LinX. This is just taking an educated guess, getting a result so you can see the system's standing, and then adjust accordingly. I call it "guess and check overclocking," and it's typically the fastest and easiest way to find a stable clock.

set vcore to 1.35 and it wouldnt boot, cept restarting when mb logo page came on, tried again at 1.34v and same problem.

 

[ModestGamer]

set vcore to 1.35 and it wouldnt boot, cept restarting when mb logo page came on, tried again at 1.34v and same problem.

power supply maybe suspect or the MB is.

 

[BTRY B 529th FA BN]

Yeah, your current has to be a steady flow with low ripple; check Johnnyguru for reviews, and its' either your RAM or your MB that is ... lets just say my guess that is causing it.

 

[skuzzzzy]

past 7 passes on memtest.

 

[BTRY B 529th FA BN]

try memtest hci instead