I'm trying to reach a stable overclock of 3.0 Ghz, or even 3.2 Ghz with a Q6600 G0 on an EVGA nVidia SLI 680i A1 motherboard (System info at the end of this post). This is only my second overclock I've ever performed and I seem to be having alot of problems. I can't seem to get this thing stable, or even predictable enough so I can get some hint on which direction to go. I was hoping someone might have some insight into what I'm doing wrong or what I might try.

My System:
Intel Q6600 G0 (VID: 1.3250)
Thermalright Ultra-120 Extreme & Scythe S-Flex 120mm
EVGA nvidia 680i SLI A1
4 x 2048 (8 GB) Mushkin DDR2-800 4-4-4-12 (2.000V - 2.1000V)
2 x EVGA 8800 Ultra in SLI
1 x 150 Gb WD Raptor
2 x 320 Gb Seagate 7200.10

FSB: 1333
Mult: x9
Memory: (Linked/Synced).
C1E Enhanced State: OFF
CPU Thermal: OFF

I've tried it with all the voltage settings on BIOS Auto:

CPU: 1.39
FSB: 1.4
Memory: 2.10 (Manufacturer setting)
SPP (NBridge): 1.25
MCP (SBridge): 1.500
SPP <> MCP: 1.20

It boots into windows fine. It seems stable. You go to run Prime95 and as soon as the dialog box comes up and you choose Large FFT's and click 'ok'...it displays the first line of text for each worker thread...and the system reboots itself.

I tried bumping voltages. I've tried bumping the CPU, in steps, all the way up to 1.49375 (in BIOS - comes out about 1.44 in CPUID). Absolutely NOTHING changed...it didn't run in Prime95 one iota longer. As soon as you click 'ok' and P95 starts the first test...the system reboots.

So I thought it must be a motherboard voltage. So I started increasing those in steps, one at a time. This is where I am confused...I expected to see it start to get MORE stable...or to see some sign of the changes I'm making....something I could 'read' to figure out what to do next. The system go LESS stable with increased voltages.

CPU: 1.49375 (In BIOS, 1.44 in CPUID)
FSB: 1.4
Memory: 2.100
SPP: 1.45
MCP: 1.550
SPP <> MCP: 1.35

Once I increased motherboard voltages...I either couldn't boot into XP at all and the system would just reboot itself attempting to load windows....or...it would JUST get windows up, BSOD, then reboot.

Odd thing is...if I reset ALL the above voltages (Including CPU) to what the BIOS sets on Auto...it will boot into windows and stay there stable...UNTIL I run P95 and then...it reboots at the EXACT same place every time...right as it starts its first test. If I leave ALL the voltages at "Auto" and then JUST bump SPP (SBridge) a nudge to 1.40...suddenly it won't boot into windows anymore. At 1333 FSB and above...it just seems that if I add ANYTHING to SPP (1.25, 1.40, whatever)...suddenly it won't boot into windows or gets BSOD's.

If I leave SPP at Auto and bump SPP <> MCP to 1.25....it will boot into windows but reboots as soon as you start a P95 test. If I leave everything at auto and just bump the MCP to 1.525 it will boot into windows but reboot as soon as I try running P95. Minute I bump the SPP though...it gets goofy. Sometimes it won't boot into windows, sometimes it will, sometimes I get BSOD.

So...I decided to try 3.2 Ghz...400 x 8. Left all the Voltages on Auto and it booted into windows just fine. I tried P95...Yay! It ran fine. I figured I'd hit some sort of wall. I'd run out of time fiddling with it (Wife getting angry) so I shut it down and figured I'd continue later. I come back later...i didn't change a THING...and suddenly its right back to the SAME behavior I was experiencing earlier.

Increasing CPU voltage didn't make P95 run even one second longer. Increasing other motherboard voltages seemed to just make things unstable.

My temps are WELL within range....whether you read the temps from CoreTemp, Real Temp. Low thirties.

So..just to get some perspective I tried just a very mild overclock to 2.67 Ghz.

"Auto Voltage" all the way around in BIOS and P95 runs nice and stable. I let it run 8 hours or so...Temps under load stayed in the low forties. No problems at all.

I thought maybe it was my SLI setup making things unstable so I disabled SLI in the nvidia panel. That didn't help. I've got 8 gigs of RAM in it (4 x 2048 Muskin 4-4-4-12 DDR2-800) so maybe having all 4 memory slots full is making it very angry at me? It overclocks fine to 2.67. Before work this morning...I just ran a very quick 4 hours of Memtest86+. I didn't do one stick at a time..I did all 4 sticks at once and I didn't have time to complete more than two passes....but at that point...no immediate errors.

Maybe I just have a voltage hungry CPU and all I need to do is add more voltage...however....how far is it reasonable for me to go? I don't wanna fry the thing. I'm already at 1.48750 or 1.49375 (as reported by BIOS), 1.44 as reported by CPUID.

Ugh.

Help? Thoughts? Idea of what I should try next or what might be wrong?

Unfortunately, those boards are very difficult to over clock with q6x00s.

You appear to be doing everything fine, however, do you know if you have a "front side bus to north bridge strap" in the Bios? I have noticed that the chipsets do not seem to like the way the memory is calculated via the system divider and timer, so sometimes selecting a strap manually will help, If you can, try both the 400 Mhz and the 333 Mhz straps when you get the automatic errors.

So at 1.49375, it drops to 1.44 when you enter windows, and what does it drop to when you run Prime? Do you get a split second to see CPUz reflect a lowered core voltage before it re boots?

If that doesn't work, could you run Core Temp, List your VID here, and then go into the bios and manually enter this at the Bios VCore, and use whatever settings actually let you boot and run fine. Disable speed step!

After you enter the number in the Bios, boot into windows and list the idle voltage with CPUz, then run prime small ffts on all 4 cores and after 1 minute, note the lowest level you see the core voltage flicker down to.

You can also try 401 FSB (or 1604) instead of 400 FSB and see if its not liking the way the strap is calculating the memory speed.

I am currently battling it out with a 780i that does the same thing... only its small fft stable, just as soon as you start a Blend test, it bombs out instantly. Though my memory is running at a speed and with timings that I have proven run fine.

Good Luck!


--Lupi

I may have missed it, but did you update your BIOS? That may fix it.

Also another possibility is mobo heat. The NB on the 680i chipset get's stinking hot when you overclock. I'm "stable" so far at 3.3ghz 9x multi. Much past that my mobo temps start to rise exponentially. I was running 3.45ghz for a month or two, but since the weather has warmed up here it was causing things to get unstable.

I was running 1.4v BIOS 1.312 Full load at 3.45ghz. Just to give a point of reference. You should be able to do 3.0ghz with no more then a couple voltage bumps.

First- Unlink the memory/fsb

That is the number one reason these boards are easy to oc, no fooling with the mem dividers.

2nd- loosen the mem timings to 5-5-5-15 and go back to one pair of memory to get a stable oc, then add it back in and take timings down.

3rd- P31 or 32 is the earliest bios that plays nice to Quads.

Follow this guide before you attempt again, after updating the bios-

http://gear.ign.com/articles/747/747671p2.html

Thanks for the help Lupiron

I'm fairly new around here...and when I was searching the forums for answers to my dilemma, or similar situations, I noticed you had offered help in overclocking to quite a few folks. Thanks alot for taking the time to read my thread and offering some help

Alright...I have some more information. Its actually more information than I expected to get so I'm hoping you might be able to draw some good conclusions from it and help me to get this thing stable.

First off...Lucuis had asked if I updated my BIOS...so I'll answer that one first. Yup...I've been on the newest version (P32 I believe) for a while.

So here goes...

1) First off, before I got your responses, I was trying to relax my memory timings and try that. I set those back from 4-4-4-12 to 5-5-5-15 and then 5-5-5-18 and that didn't help at all.

2) I read your post and set my FSB back to 1333. The first time I did that...I got the results I explained with my original post...where I could get into windows but whenever I attempted to run P95...it just rebooted. This time...I couldn't even get it into windows...at all. Every time windows would START to load....and you'd see the XP splash screen...it would just reboot...or BSOD. I left ALL the voltages on Auto and it just would NOT load windows.

3) After lots of reading about 'holes' (I'm still not entirely SURE what they are...but I have an idea)...I decided to try something:

I knew I could set QDR 1185 (296 x 9) for 2.67 Ghz at Auto voltage and run perfectly stable.

So I tried setting FSB to 1200 (300 x 9 = 2.7 Ghz) ...omg...it was SO unstable I almost couldn't get into the BIOS. Same happened when I tried setting 2900...it was HORRIBLY unstable.

Now, the interesting thing is that I JUST finished a successful overclock on my wife's PC. Hers was a B3 stepping Q6600 on a 780i board. Hers had the same 'hole'. between 2.67 Ghz and 2.9 Ghz...it was so unstable I could barely get into the BIOS. However, for hers, it ran nice and stable at 1333 (333 x 9 = 3.0 Ghz).

Mine however, the Q6600 G0 on a 680i was, like hers, horribly unstable starting at FSB 1200. However...it SEEMS like there is some sort of stability 'curve' around some sort of 'hole' and mine is just worse than hers. For mine, at FSB 1333 I could ALWAYS get into BIOS, I could 'sometimes' get into Windows, and I couldn't run ANY P95 testing.

So...I thought...if the 'hole' is bigger...maybe I could get past it (If indeed this IS the 'hole' I've read about). So I set Multiplier to 8 and FSB to 1600 (400 x 8 = 3.2 Ghz).

JUST like last time I did this...it went into windows easily...and it actually let me run P95 Large FFT's for about 20 minutes! However, instead of 'failing' in P95...it just 'rebooted' windows when it croaked again. At least this short run in P95 before my crash and reboot allowed me to gather the data you asked for:

Voltages were ALL on auto:

4) So I tried 3.2 Ghz (3195) with Mult as 9 (355 x 9) instead of 8. and I got identical results to above. I also tried 3150 (350 x 9) and also got identical results.

5) So I bumped it UP further. I tried 3375 (375 x 9)...and somehow...it SEEMED 'more' stable (although, at this point, I've done this SOOO many times it could be my imagination). At this setting, it made it into windows every time, and I was able to run P95 for about 15 minutes. The voltages for this run were identical to above.

The system was still 'unstable' enough however....that it was JUST sitting in windows doing nothing while I was typing this...and it rebooted. I got it to boot right back into windows though.

I apologize, but unfortunately I don't know what a 'strap' is. I've seen the term used before...but I don't quite understand it. Needless to say, I don't see anything named 'strap' in my BIOS.

I've got my memory "Linked & Synced" to the FSB...and it calculates the speed itself. I've tried it 'unlinked' as well.

Thanks Lupi and Lucuis

You do not wanna use large ffts, you wanna use Small FFTs for CPU stability, and Blend for ram stability. Using large mixes them both, and clouds the actual cause.

Small will also tell us the true VDroop. Next time you get a chance, try running small ffts and seeing what the idle voltage goes down too. And remember what you have it on in the Bios.

I am currently doing the same thing you need to do. I just picked up my P5N72T 780i chipset. It works great on Auto and stock. However, 1333 and on up is totally unstable. No posts, some posts, BSODs while loading windows, freezes while loading windows...

All the way up to 425 (1700 QDR). At that FSB rating it boots into windows fine, and has been running small ffts for over 2 hours. However, it took me 35 hours to find stability!

It's the way it calculates the memory final speed... Look for the TRD timing in your memory timing setup in the Bios and set it at 7 if its in there.

The other guy seems to think that "unlinked" means it doesnt calculate a ratio? of course it does. Same deal on my "seamless" strap change for my P5N-d. 400 FSB doesnt post at all. but 401 does... Thats why the strap thing would have helped.

Instead of moving it around by 1 point to see if it works, you just change from strap 400 Mhz, to 333, and it re calculates the divider. As in, at 400 Mhz FSB strap to NB, and 400 FSB, it comes out to 800 Mhz. HOWEVER.. when I change from strap 400 to 333, I get 801 as the final memory speed, and that also works as good as adding the 1 point, or taking 1 FSB point away until its happy.

Like I said, for me, Prime small ffts to check cpu stability was fine! As soon as I touched Blend test, INSTANT freeze and reboot. It's the memory.

But you'll need to play with yours. My P5n72T wouldn't do much of anything below this 425 FSB. Nothing at easier FSBs.

So re do the tests with Small FFTs, or just blend for ram, and try and get the VDroop from windows Idle in CPUz to fully loaded on Primes small FFTs and note the lowest it drops to!

--Lupi

Oh, does your board have FSB termination voltage? Starts at 1.20 and goes up to 1.50... And do you have something called CPU GTL Vref?

--Lupi

Hi guys, thanks again for wading through my really long posts and trying to help me. I need all the help I can get. I swear...I'm going to stress my own hair out trying to figure this out

Okies...onwards with the info you guys requested:

Yup...tried that...no luck

I'm using P32.

hehe...thats one of the guides I originally read to prep me for this...unfortunately...the author didn't put in a page entitled "When everything goes to hell in a handbasket"

Alrighty Lupiron....as requested....I've got it up and chugging away with Small FFT's and I have the data you requested. So far, the most 'stable' of the settings I've had so far has been the 375 x 9...so thats the one I used to get the data you requested.

I ran Small FFT's in P95 and my system lasted maybe 10 minutes...and then did its typical 'reboot unexpectedly' thing. It went long enough for me to get the numbers you wanted though.

Ugh. I'm about at my width's end. I can't imagine spending ANOTHER 35 hours....lol. After reading all the bad things about the 680i I have today in researching this...I'm half tempted to smash the thing and go get a 780i. My wife's overclock with her Q6600 on her 780i was WAY easier than this. I just hate the idea of starting ALL over again. I'd really like to find some miracle to get this 680i a stable acceptable OC Ya know what they say...spit in one hand and wish in the other

For the heck of it...I tried 1700 (425 x 8) as you did...but no luck for me. It was actually less stable.

I didn't see any "TRD". I have the following:

lol...I have no idea what you just said Went over my head like a bad joke

As for the results of the small FFTs. They were about the same (Although I did notice that they generate a heckuva lot more heat). Same thing though...it runs for 10 minutes or so...and then reboots.

My 'record' so far is with the above listed FSB (1500), I got it to run 25 minutes with Large FFTs.

But at least I was able to get ya the small FFT test results

Nope...nuttin' like that.

Thanks again for your help Lupi It takes alot of time to think through and type up troubleshooting steps like that...and its greatly appreciated!