i7 920 problems, unstable, need help!

[Zonnza]

I have a 920 D0 stepping and cant seem to get a stack 4.0ghz
I just got my megahalem rev b with 2 S-flex SFF21F fans running at 100%


I went thry matts thread and some other threads and started at:
BCLK 200
20x multiplier
1.25v cpu vcore
1.65v vDIMM
1.25 QPI
C1E enabled
speedstep disabled
mem ratio is 2:8 at 7-7-7-24 1t command rate, running 1600mhz


I turn on realtemp, linX and cpu-z.
Idle temps were highest at 40 and lowest at 38

I ran linX, using 2gb ram, comp crashed within 15 seconds, load temps got to 65 before it crashed.

Next I brought my cpu Vcore to 1.75v, ran linX, so far stable... 5m into test; highest temp atm on full load 100% is 69-73

So now first question, my idle temps are high 30s, low 40s... good or bad? If this OC is stable, should I lower my cpu Vcore or up my GHZ to 4.2? I dont think ill want to go over 4.2ghz. What should I let my max temps get to before I should stop my OCing?

UPDATE: LinX at 10m so far, stable; gonna take wait another 50m and see how it goes.
UPDATE 2: LinX 20m stable and going.

Question id like to figure out fast: My fan controls in my bios, I have a CPU fan controller, a power fan controller and chassis fan controller... I had to plug in one of my 2 fans on my megahalem into the CPU fan header that is next to the ram... What is that fan header suppose to be in bios? Power fan or chassis fan?



Gonna post this and edit it as time goes on...

 

[simviatoR]

I have a 920 D0 stepping and cant seem to get a stack 4.0ghz
I just got my megahalem rev b with 2 S-flex SFF21F fans running at 100%


I went thry matts thread and some other threads and started at:
BCLK 200
20x multiplier
1.25v cpu vcore
1.65v vDIMM
1.25 QPI
C1E enabled
speedstep disabled
mem ratio is 2:8 at 7-7-7-24 1t command rate, running 1600mhz


I turn on realtemp, linX and cpu-z.
Idle temps were highest at 40 and lowest at 38

I ran linX, using 2gb ram, comp crashed within 15 seconds, load temps got to 65 before it crashed.

Next I brought my cpu Vcore to 1.75v, ran linX, so far stable... 5m into test; highest temp atm on full load 100% is 69-73

So now first question, my idle temps are high 30s, low 40s... good or bad? If this OC is stable, should I lower my cpu Vcore or up my GHZ to 4.2? I dont think ill want to go over 4.2ghz. What should I let my max temps get to before I should stop my OCing?

UPDATE: LinX at 10m so far, stable; gonna take wait another 50m and see how it goes.
UPDATE 2: LinX 20m stable and going.

Question id like to figure out fast: My fan controls in my bios, I have a CPU fan controller, a power fan controller and chassis fan controller... I had to plug in one of my 2 fans on my megahalem into the CPU fan header that is next to the ram... What is that fan header suppose to be in bios? Power fan or chassis fan?



Gonna post this and edit it as time goes on...

Wait... what? 1.75v? That's ridiculous! I run 4.2ghz @ 1.317 on my 920.

 

[Zonnza]

sorry, ment 1.27v

atm its stable at 1.27v with all other settings, i dont think ill push it over 4ghz, dont really need to


but my CPU fans give me a humming or buzzing like noise which i never heard from a fan before... is this normal or should i be worried about it vibrating against something?

 

[simviatoR]

I would expect at least 1.29-1.305 for 4.0ghz - for me it was achieved easily by having it at 20x200, 2:8 memory @ 1602mhz, and vcore @ 1.305... all the rest on auto.

 

[Zonnza]

i kept crashing initially because my vdimm was suppose to be 1.65v for my ram, and i remember that u had to keep qpi voltage within .5v of vdimm or you will not get good results; i had my vdimm at 1.65v but my qpi at 1.1v and my tests would fail within 15sec so

 

[mattspalace]

D0's have typically been stable at 4GHz with about 1.25v, so you may have gotten one that is a little worse than the others, but still can't complain about a 1GHz overclock, eh?

I would disable C1E as well. Everything else looks good to go.

 

[antipesto93]

my #3849B runs stable 4.0 @ 1.3v

 

[mattspalace]

my #3849B runs stable 4.0 @ 1.3v

Hmm..maybe I actually got one of the better ones then!

 

[Daddyjaxx]

Try the 21x multiplier or Turbo on. It takes less voltage.

 

[antipesto93]

Hmm..maybe I actually got one of the better ones then!

yeah - batch numbers can make quite abit of difference

i originalyl ran it on 1.25v @ 4.0ghz.....and it could pass hours of Linpack stress testing

but it would have generall instability problems untill i added a whole more 0.05v

 

[Nebulous]

Dam. I've seen several "cherry D0's" have this same problem. Makes me even happier my C0 is a tank. 20x200 with 1.36v w/Turbo on, ram @ 1.65v and evrything else on auto and not 1 single hiccup since the day I built it.

That board you have should be slapping that D0 of yours all around. Have the latest bios?

 

[Zonnza]

I use the shipped bios with the board I got, mainly because from reading the evga forums that it is the best bios for OCing stated by a few people there.

I know my board should be able to easily take this chip.

I started to do something different, I went with relying on the voltage that my PC health in bios showed and the cpu-z and evga e-leet tunning utility showed instead of the voltage in the options: What I mean is, whatever voltage I set in the bios, it will show a lower voltage in cpu-z/evga eleet/pchealth-in-bios. An example would be when I set my voltage to 1.29375v in bios, when i restart and go back into bios, i check PC health and it shows at 1.28v, I go into windows and open cpu-z and evga e-leet and they both show 1.282v.

With that voltage thing, I knew that when I set it to 1.25 in bios, it was actually around 1.23-1.24v instead. Which would make sense that linX would crash within 15 seconds when I was at 4ghz then.

Now Im running the 1.282v and 4ghz with 1.25 qpi and 1.65 ram voltages set in bios (may be lower/higher) but everything runs stable. The new question is: Is it worth it to try to OC further? In my opinion, first I need to solve this issue: http://www.ocforums.com/showthread.php?t=627043

That is a link to my other thread in cooling, Im trying to solve an issue Im having with my megahalem cooler and this odd noise which I cannot stand having, after some time I get a massive headache so this is an unacceptable issue.


My other issue is if I applied my thermal paste correctly? I had to slide the cooler a bit to get the mounting systems screws aligned to secure it... I think it screwed with the TIM and now my idle temps are high? I have no idea where they should be but with my "idle" status (load < 10%) I get temps that are 42,43,42,40 and min temps that realtemp records over time has been 40,41,39,38.

I honestly have no idea what a "idle" temp should be, but my max temps usually don't even hit 75 on any core.

 

[Neuromancer]

Why CPUZ and such show lower voltage is because of vdrop and vdroop

drop occurs at idle droop occurs at load. Many board have LLC (load line calibration) settings in BIOS to help prevent that.

If you do not have that option, just give it a little extra voltage in the bios.

I played around with running low volts for 4200 MHz on my D0, personally the fact that I had to "tweak it" to get it to run at much lower than 1.3, just easier to give it the 1.3 it wants.

I do not normally run that low a speed. (i7 is NOT a desktop PC at all IMHO) so right now its all about benching at 4.6+ GHz and 1.35volts aint going to get you there

 

[Zonnza]

Ok so I ran some new tests. Turned on all the programs I use for testing and let them sit there: Prime95, Real Temp, CPU-z, Evga e-leet. After 10 minutes of "idle", minimum temps on real temp showed 40,40,39,37 (computer has been on since 11am and it is 8pm). My "idle" temps were 41,42,40,39 on average after those 10 minutes.

Then I ran prime95 for 10 minutes, at 10 minutes it was on test 6 and almost done. Here is a screenshot: http://img197.imageshack.us/img197/1551/tempstest1.png (excuse my background )


Take note: I have side of case OPEN and only 1 fan(fan1 on speedfan) on my megahalem [other fan taken off and put to the outside of the case but still connected and running (fan 2 in speedfan)]
So to sum things up: 10 min of testing for idle and load
"idle" <10% load temps: 41,42,40,39 ish
min temps: 40,40,39,37
"idle" computer temp (displayed on lcd on mobo): 25

"load" 100% temps: 70,70,67,67
"load" max temps: 71,73,68,68
"load" computer temp (displayed on lcd on mobo): 55


EDIT: @Neuromancer; i was wondering why my voltage to my cpu on cpu-z and evga e-leet went from 1.282v to 1.247v DURING the "load" testing. Should I keep it like this or how do I get rid of it?

 

[CompuTamer]

Dam. I've seen several "cherry D0's" have this same problem. Makes me even happier my C0 is a tank. 20x200 with 1.36v w/Turbo on, ram @ 1.65v and evrything else on auto and not 1 single hiccup since the day I built it.

That board you have should be slapping that D0 of yours all around. Have the latest bios?

You, sir, are lucky.
My C0 is horrible. 1.43 and i THINK that it's stable under linx now... at 3.95

That extra 50MHz is bugging me, but i can't quite make it up that high thanks to my heat wall.

 

[mattspalace]

Zonnza - you are running both fans on your megahalems, correct?

Still having issues, or is everything good now? If you have loadline calibration setting in your bios, you can turn it on. If not, then just account for it when your setting the voltage, i.e. if you want 1.25v but you have vdroop, then set 1.27v and it should be about right - like Neuromancer mentioned.

Also, you should have idle temps in mid-30's and load temps around 70c, but this all depends on ambient room temp, the fans on your cooler, and the voltage you're giving to the CPU.

Did you figure out the fan noise issue? I'd figure out which fan is making the racket and replace it with a new one.

 

[Neuromancer]

If your case is setup right, taking the side panel off should increase temps. I noticed on my NZXT Zero case that just cracking the sidepanel a half an inch raised CPU temps about 5C. But it assumes a good air flow design (Used to high speed yates as rear exhaust, 1 HS yate for front, and 2 low speed in the side panel providing DAC to the ram and CPU intake area and the GPU/SB area.


Never the less those temps are pretty decent.