i7: Help me get to 4GHz

[esdxx]

Hi. I know how this sound like... another n00b who havent read any guides or did not make any effort to study things before asking. I assure you this is not the case. I have been experimenting for last week and trying to understand everything about my bios settings. Some things are clear to me, but some are unfortunately still beyond my understanding. Ok, i dont want you to read wall of text, so i'll get right to the point...

Currently i am running stable at 3.8GHz (181x21) with 1.268V vcore.


Anything beyond that will give me 124 BSOD. I tried 190x21 even with 1.32V vcore, still BSODing. I am starting to believe my chip is horrible and its borderline is at 3.8GHz. I tried setting BIOS according to few guides, but that didnt work for me.

My other settings are at:
RAM Multiplier: 6.0
QPI Link: x36
PCIE: 100

Rest is pretty much on auto (obviously i tried different settings, but nothing resulted in stable system).

Details of my rig - i7 920 D0 (# 3911A318), Gigabyte X58-UD5 (F8h BIOS), 6GB RAM OCZ DDR3 1600MHz PC12800 CL7-7-7-24 OCZ Platinum Edition Low Voltage XTC

[Sam_oslo]

What's your current Vdimm and VVT/QPI?

124 BSOD wouldn't usually be related to Core/Vcore, but most probably related to IMC, Uncore or QPI. This could be related to RAM sticks too.

You can try to play with different VVT-volt (chose different VVT and run some PrimeBlend) to start with. You should also use 1.65v Vdimm with relax RAM-timing to eliminate problems from the RAM itself.

[oggie]

I have batch 3907A360 and gotta say i'm finding it very hard to get it above 4.0GHz ( I need 1.31v to get it stable), strange thing is my cpu's default vcore is 1.275v which is crazy high. Think 39XX chips are really bad compare to old 38XX, wonder if Intel have done this on purpose close to release of i5's

[esdxx]

Forgot to mentoin, vdimm is at 1.65V. I have tried VTT voltage to the max (1.34V i believe, but that didnt help). I have read that my RAMs are not the best for overclocking, but for moderate overclock (4GHz) it should be more than enough.

RAM timings are at 7-7-7-19, i'll try 8-8-8-24.


oggie: my default vcore is also 1.275v. it is fairly dissapointing, after reading all those forum posts aka "4GHz on default voltage - no effort needed"

[esdxx]

Ok.. This is what i did, it would give me either 124 BSOD, 101 BSOD or it wont post at all. I went from 1.30V vcore to 1.36V.

vcore: 1.36V
vtt: 1.33V
vdimm: 1.64V
qpi link / uncore: lowest possible
ram multiplier: 6.0
timings: 8-8-8-24
cpu pll: 1.88V
ICH/IOH: 1.2V

BCLK: 191x21

Those settings are already ridicluously high, so i dont see how it can work with lower settings... What is interesting tho, it runs rock solid stable @ 3.8GHz (the screen i posted above). Looks like a hole or something (you remember FSB holes? - are those present with i7 generations too?)

[Splave]

Those settings arent that high friend

CPU pll can go back to stock 1.8 some even lower it to 1.5.
Bump the vcore to 1.375 and vtt to 1.375 (ive had the best luck with the 2 synced with d0's)

[esdxx]

Those settings arent that high friend

CPU pll can go back to stock 1.8 some even lower it to 1.5.
Bump the vcore to 1.375 and vtt to 1.375 (ive had the best luck with the 2 synced with d0's)

I will try it up to 1.4V, but to be honest, i already getting above 70C temps on 3.8GHz, so i probably wont be able to run 4GHz 24/7.

[v0dka]

Have you tried different combo of multi and blck like 20*200 or 19*210?

Keep uncore / nb speed lower than the QPI link (8:9 or so).

And you need a better cooler / fan. If that voltage reading is correct (1.184v) the temps are really high even for the summer. On the bright side, it seems that you dont need 1.275v stock to run @ 3.8 which is good. Maybe reseat the cooler or lap the base & cpu?

Where did you get that 1.275v is stock voltage btw?

[esdxx]

Have you tried different combo of multi and blck like 20*200 or 19*210?

Keep uncore / nb speed lower than the QPI link (8:9 or so).

And you need a better cooler / fan. If that voltage reading is correct (1.184v) the temps are really high even for the summer. On the bright side, it seems that you dont need 1.275v stock to run @ 3.8 which is good. Maybe reseat the cooler or lap the base & cpu?

Where did you get that 1.275v is stock voltage btw?

When i set everything on default in BIOS, the vcore number is 1.275V (atleast thats what it shows in BIOS), so i figured it is stock voltage... maybe i was wrong.

I reseated the cooler couple of times, but i believe it is my room and lack of airflow in it. I had high temps on my q6600 too (with same cooler). I have not tried to lap it, but i might look into it later. I am using artcic silver 5 by the way..

I tried 19x210 (or even lower values), but it would freeze during windows startup no matter what. It was like week ago though, so i might try again.

[EDIT]
With really high vcore of 1.4V i have trouble even boot into windows. Tried 19x210, 20x200 and 21x191. Usually 124 BSOD. VTT is max recommended for my mobo - 1.33V, VDIMM is 1.64V, QPI Link x36, Uncore on auto (it wont boot on x16 - lowest one). RAM timings 8-8-8-24 2. It is interesting, becouse i can actually boot in windows with lower voltage, like i am now, with 1.29V and 21x191... i guess i will BSOD after i run prime though.

[v0dka]

Maybe it froze because of too high uncore and not enough QPI. Or maybe you need more VTT for this to work.

If you want to isolate the temperature problem, open up the case and get a huge room fan blowing right onto the mobo from close by. Looking at the fact that higher vcore only gives more problems, I say temperature is definately an issue.

Also, leave latency settings for memory on auto and see what happens.

[esdxx]

Hmm, as i wrote in previous post

It is interesting, becouse i can actually boot in windows with lower voltage, like i am now, with 1.29V and 21x191... i guess i will BSOD after i run prime though.

Well i ran prime small fft and its running for 30 mins now. 4GHz (21x191), 1.29V, 1.64V RAM, 1.34V VTT, also i set uncore to second lowest, so its slightly more than 1:1. HT is off tho', should i call it progress anyway or its to be expected that HT off will result in more stable OC? Temps are higher 60's almost reaching 70C.

http://i25.tinypic.com/ir3vvd.jpg

The temps might be a problem i agree, but my ambients are usually above 25C during summer, no AC. I was thinking about replacing my Noctua with Megahalems or TRUE, but i am afraid it will make no difference.

[esdxx]

Ok, so i tried with HT ON and it works! I dont know, if this is possible thanks to new BIOS, or one of the things you guys suggested me (regarding uncore and qpi). Either way, many thanks for your replies. I also bumped VTT above recommended number, 1.35V as for now. Maybe i could go lower with VTT and even vcore.

[cobra_kai]

I see you already have it working, but my experience was that it was much easier to get a 4 GHz OC with the 21 multi than 20. When I did 20x200 I needed about 1.36 Vcore for it to be stable, and right now I am at 21x191 and I am at 1.28 Vcore and still working my way down. I have watercooling, so the temps might not be comparable, but I was priming in the upper seventies using the 20x multi and I am now priming in the upper sixties. I have HT on btw.

[DavyBoy]

Hey man,

Your batch is pretty close to my new 920, I have 3910A***, and I have got it stable @ 3.8GHz also, and have spent the last week trying to get 4GHz stable, but it just isn't happening, I have tested right up to 1.35v Vcore, and 1.4v QPI, but it is temps that are giving me the worst problems with this chip..

My last chip was a 3851A, and it would do 4GHz without any troubles at all, and I even got it verified @ 4.4GHz, and the temps on that chip were excellent...

I will be getting this processor changed in the next few days though, so hopefully the next one isn't a dog like this one..

I get Stop Code 0x00000101 everytime I try running LinX when I have it clocked any higher than 3.8GHz..

The thing that bothers me is it will do 3.8GHz with really low Vcore and QPI voltage than my last chip needed, 1.200v Vcore, and 1.260v QPI is all that this chip needs, whereas my last one needed 1.264v Vcore and 1.355v QPI for it to be stable, however, even with more voltage going through it, my last chip ran significantly cooler than this replacement one...

I suspect the IHS is bad on this chip, and that is what is causing my high temps, as I have tried 3 different coolers, and I have remounted and reapplied TIM loads of times to see if that will sort these temps out, but nothing I do makes a blind bit of difference to temps, even using the same HSF I had on my previous chip the temps are still insanely high.

I get higher idle temps with this new chip at stock frequency than I did with my last one overclocked to 4.2GHz!!!!

[esdxx]

Well, i thought it is working, but after installing new version of LinX, that statement is unfortunately false.


My chip also running pretty hot, but i think it has something to do with ambient temp and airflow in my room, which is terrible. My previous chip - q6600 was running also quite hot. My Noctua has push/pull with stock fans, i might replace them with higher-CFM fans later.

When it comes to stability, i dont think my cpu is the issue. I can run small fft test for hours without problems, but LinX with all memory used gives me BSOD after like 5-10mins. I heard thats RAM/QPI/Uncore related. My RAM timings are currently at 8-8-8-24-2CT, voltage 1.64V (should i try 1.66V as suggested in some quides?), CPU voltage is 1.29V (1.248V in CPUZ under load) and QPI is 1.35V. Memory is running at 1146MHz, Uncore is slightly higher, than 2x of RAM clock.

Oh btw. BSOD code is 09C.