e8400 and Ga-ep45-ds3r voltage/stability issue

So this chip has been driving me nuts for about 3 months now.

First lets start with a screeny.

This is a shot of the stock settings


This is a shot of the current (8.5 X 471 for 4ghz)


I've run memtest so rule out any ram issue. I've also run virus scans with 3 different scanners.

This chip/board drove me nuts because to get any sort of stability I had to turn off vista superfetch, readyboost, and windows search. Took me months to get this in my head that having any of those on results in an insta blue screen of death of prime. Fine. Then I hard to turn off bit execute and virtualization technology to not get blue screens of deaths. Took months as well to figure this out.

Anyway with my current setup I can prime for 8+ hours and hit temps of high 50s and idle is 29-31. But look at that vcore! Actually, on idle after drop, cpuz and ntune report its only 1.45 and then during load it falls to 1.408-1.424.

What bugs me is that going down the next degradation of vcore, (which is like .05) results in a 3 second orthos freeze, BSOD, and restart. I mean bam, 3 seconds, freeze, blue screen, restart and error report. WTF!

Anyway, yeah that cpu pll is ridiculous. Stock is 1.5 and i have it at 1.15 and still stable. My cpu vtt could go as low as 1.34 but for some reason it could prime for 7 hours at 1.34 but then wouldn't boot up

Please help unless you think I'm worrying over nothing. I mean 1.45 vcore, pretty good stability (no crashes or bluescreens), and decent temps... but man you know I'm just not happy...

well hate to say, it seems you expected a bit to much from the E8400. you may still get one that does 4ghz but this doenst look to be one of them. three are other voltage settings that prolly need to be messed with but i dont seem them in that list. imma have to go find the manual then read a review... be back in a while...

CPU PLL is 1.5v stock. Why do you have it set so low at 1.15v?

I usually use 1.5-1.6v for CPU PLL on 45nm.

well i don't have to set it so low. i just kept dropping it and it'd prime for like 5 hours without crashing. actually at 1.15 i haven't tested the spot below it and it shill hasn't crashed.

@Evilsizer
Its weird. This chip actuallywill boot at like 490 fsb and will run around in windows at that speed. It just won't prime.
What also bothers me is that before I reinstalled windows, I had it at 1.4625 vcore as opposed to 1.4875 and i used a 1.6 cpu vtt and 1.69 cpu pll and it was capable of priming over night, but now if i used my old settings it crashes. I really don't have any idea where the instability is coming form and I'm hoping you guys can give me some advice.

the only thing i will say is dropping the cpu pll dropped like 10 degrees off my idle. i really appreciate the help

again, before i could prime with 1.4625 (actual I don't remember, but it was probably about 1.43-1.44), a 1.6 cpu pll, and a 1.6 vtt.

Now those are completely unstable (as in 3 second crash in prime). All i did was reinstall windows and maybe play with sleep settings on my bios. What happened?

if it is crashing so soon in prime its the cpu needs more voltage. another thing could be higher temps due to bad contact on the cpu's IHS. sorry i have been a busy and not able to look up that stuff yet.

could you list out the rest of the parts your using please?

ga-ep45-ds3r

my ram is split. 2 gigs of ocz reaper 6400 and 2 gigs of ocz platinum 6400 something or other. I synced the timings though by slightly dropping that for the ocz reaper (5-5-5-15 for all sticks).

i'm running an e8400
a pny gefore 9600gt but its not oced and my pcie is locked.

not sure what else is relevant. latest bios from gigabyte.

if it was high temps wouldn't i know? I mean, i monitor my temps constantly and again its simply going from 1.4875 to 1.4815 that causes an immediate freeze but only in prime and not normal ops. its bothersome because i previously ran it much lower at 1.4625 (the actual was closer to 1.4) stably and i changed OS and bam, massive instablity and suddenly the need for like next to no voltage for cpu pll...

Seems increasing the cpu ref to 0.86 lets me drop the vcore by to 1.43 (1.37 actual vcore). Working on getting longer stability runs but man, the difference one little change makes.

Edit: wrong! still can't go down even a single degradation without a 3 second crash...

1.6 vtt, and 1.69 pll!!!

Maybe you've damaged the chip?

Try going in both directions w/ GTL. More is not always better.

gtl is what? cpu reference? or cpu vtt?

Sunburn74 said:

gtl is what? cpu reference? or cpu vtt?

Click to expand...

this will explain alot...
http://www.ocforums.com/showthread.php?t=515316

Ok. First of all, thanks all you guys for your great help.
One line by brolloks and evil sizer in a couple of threads turned everything around.

I haven't had any cpu problems, the issue has been an mch or possibly ram (though I doubt the ram) issue. My cpu was not properly communicating with the NB and I waswas beating it severely overvolting the chip.

After playing with the voltage settings for cpu ref and nb ref, I've found a pair of settings that work, that let me drop my cpu settings to more or less what everyone else is using. Great!!! Here's a screeny

Btw, for those who have this board graphics booster is on turbo, i'm using the 333mhz strap, and on standard performance. Note that I'm running it at 3.9 ghz using a 435x9 setting and can prime blend for 7+ hours and so on.



Here's the problem. The chip will boot as high as 4.2 ghz and will run most apps. The board is ridiculous letting me function in windows with settings like 8X525 so I know both the board and the chip have much headroom left to expose.

But I'm not using such high fsbs because my problem is with the mch and northbridge settings. Running prime causes a crash in 8 seconds (always 8 seconds) and by crash I mean a lock, BSOD, and restart.

You guys are pros. Can you give tips on stabilizing my mch refs and cpu refs so that i can slowly start increasing my fsbs. Going even 10 fsb more gives and instant 8 second BSOD, but at where I'm sitting I can prime for 8 hours with temps in mid 40s...

Changing the FSB speed or voltages might mean that you need to re-try all the GTL settings. What works best at 490FSB, may not be the best at 500FSB.

How is your RAM set right now? If you could post CPU-Z screenshots (CPU & Memory tabs) as well it would help more.

Its hard to know where your issue is b/c you are kinda going at it all-at-once. Its best to try and isolate the CPU, FSB, and RAM in turn to figure out where the problem might be.

FSB - keep RAM 1:1 and loose, and drop CPU multi - test w/ OCCT2.0 custom 2hr RAM test

RAM - keep FSB at a known good setting like 333 or 400, keep CPU multi low enough so CPU is not stressed, and use mem dividers to push RAM - test w/ OCCT2.0 custom 2hr RAM test, and memtest86+

CPU - keep RAM 1:1 and loose, and keep multi at max - test w/ P95 small fft

By knowing the limits of each component, and what it takes to get there you'll be able to push your system more intelligently. You won't always know which setting to try next, but you will at least know which group of settings to try.

jason has already covered alot.. with the cpu failing so shortly indicates to me a cpu voltage problem,meaning it wants more to be 4.2ghz stable. i dont like ocing by windows utils, i might to see what happens on what votlage. im old school you could say i mainly stick to the bios... things i would try first, make sure voltage is reading 1.45 or slightly under. i wouldnt run over 1.45 and would tell anyone to for 24/7 unless they were on a phase unit.

now if im understanding that bios correctly, cpu reference is cpu gtl. now your running it higher then i would for a 45nm cpu. that would be part of the problem for starters, dont use more then .67. now what i want you to do it actually use less then that start at .63 or the closest setting you can. now do this in bios cause you want both cpu references set to the same setting. once you have done that use the current settings you are using now. if it still fails to fast try lowering the cpu v and/or cpu pll voltage. as you can in some cases by adding to much voltage actually becausing the problem.

thanks for the response. I'm not ocing in windows. easy tune is just the fastest way to show you the critical voltage settings with a single screenshot.

Right now my ram timings are set to auto. The timings reported at 5-5-5-15. I'd use manual but there've been reports of instability with manual settings and to tell the truth, I've experienced that myself with this board. I'm also running a 333 strap with a 2x multiplier ( so less than 1:1)

I don't have a cpu gtl setting. I just have a cpu ref setting that seems to auto adjust whenever I change the cpu vtt. Even if I manually punch in a voltage, it autoadjusts. It's really very frustrating actually. Still you're saying my cpu ref might stand to be lowered. if I lower my cpu ref, won't my mch ref need to be lowered as well?

no mch ref wont, your talking about two different devices. have you updated the bios yet? actually since your under 500fsb use 266 strap.

use the 266 strap? by the way, what exactly does the mch strap do? I was under the impression it affected the mch internal clock or something.

here's a screeny of my ram timings. seems I was wrong and instead it was running at 5-6-6-17. SHould I change it?

the mch strap effects the sub timings which effects ocing. though you dont need 333fsb strap as your under 500mhz fsb. p45 is different then p35 but on p35 the 333 strap was auto forced on. yea using auto could cause problems with ocing. which is why we always suggest using manual timings.

Evil, I think the 0.76 you see in the pic is an actual voltage whereas the 0.63 you are talking about is a percentage.

.76/1.2 = .63