Q6600 won't overclock

[JonW]

Bonzaduck, I agree. I think 1.45v for the cpu is too much for 3.2ghz. I'm going to read some reviews of his mb and see how much droop it does when it's loaded. Maybe he's trying to compensate for 1.35v-1.40v for the core. But temps at those voltage would make it skyrocket.

Ararat, Just a bit of airflow would do for the nb. For 3.2ghz with your setup, you can entirely run the cpu almost on stock vid. If your mb droops, just compensate it through by bumping it to 1.35v-1.38v and droops to 1.30v-1.34v it should hold 3.2ghz. I still suspect the 4x1gb of ram and the combination of low nb volt that's causing you instabilities though. Some programs will run fine and some won't. I had to dig up my notes from my usb drive to when I prime'd my Q6600 at 4x800. It's alright bro, if you can't provide the screenshot. I wanted to see what your mb sets the ram timings when they are in auto. Do you have your ram set that way from the bios? VTT@400fsb is indeed at 1.30v-1.35v, just like the article I linked you earlier.

400x8@4:5
http://img214.imageshack.us/im.../400fsb333strapnw1.jpg

400x8@1:1
http://img214.imageshack.us/img214/7814/32primene1.jpg

Since you have 4x1gb of ram and P35, your nb volt should be around .20v-.30v@400fsb (rough estimate). That's 1.4v-1.5v.

Q6600 spec:
http://processorfinder.intel.c...tails.aspx?sSpec=SLACR

Edit:
When you stress test always follow this order:
-Memtest86+ or HCI Memtest in Windows (It's best to rule out the ram first)
-Prime95 Blend Test (8kfft-128kfft error=cpu related, the rest of the test are fsb/ram related)
-Games/Encoding etc.(check for quirks)...Final stage.

 

[BonzaiDuck]

Originally posted by: JonW
. . . . . . I still suspect the 4x1gb of ram and the combination of low nb volt that's causing you instabilities though. Some programs will run fine and some won't.

I forgot about that, also.

For OC'ing, you're always better off with a 2x2GB kit than with a 4x1GB kit. Less strain on the chipset, less power consumption. I can see a possible need to loosen timings with the latter setup. Not having worked much with four slots filled, I can only guess about voltage and maximum reasonable speed on the memory bus.

I remember more than 20 years ago, there was a memory-extender ISA card you could get for the old 8088 processor from a company called Tall Tree Systems -- the "JRAM" card. Mostly, it was good for RAM-disk applications, because the addressable memory-space for that processor was limited. with each new processor and bus innovation, more memory seemed better with only the addressing as a limitatiion.

Now I see people putting 8 GB of 4x2GB kits in their systems, but for any number of reasons, what is optimal and practical may differ from the maximum. Filling all the slots is fine if one is not planning to OC, or OC very much.

So I'm sticking to 2x2GB for now. For an XP system, I can live with 2x1GB, but for VISTA 64 -- 2x2GB. For me, I'll try and open up the bottleneck on disk access speed, while socketing enough RAM to provide just a comfortable margin, and hope to use only half the available slots.

G.SKILL DDR2-800 and DDR2-1000 2x2GB kits seem to be going for less than $80 now.

 

[Ararat]

OK, I tried 370x9 and it seemed stable, at 380x9, core #4 failed in less than a minute, the other three a few minutes later (the computer reset). At 375x9 core #3 failed after not too long.

These were all done with the memory multiplier at 2.0, vCore at 1.45, FSB voltage at +0.2, northbridge voltage at +0.2, and memory voltage at +0.1 (timings at 6-6-6-18, just to be sure the DIMMS aren't responsible).

Here's the real kicker, 400x8 seems comfortably stable (I did tell you that this motherboard ran my previous cpu at well above 400), so it would appear that my cpu unfortunately isn't capable of getting above 3.3ish. Will this improve with time? (I remember my first C2D - an e6300 - was pretty stubborn initially, but over time i managed to get higher clock speeds out of it, be it due to the thermal paste or some other reason I don't know about).

I seem to have a major vdroop as well. My default vCore is 1.30v, and I have it set to 1.45. But under load (prime95), my core voltage is reading as 1.328 (sometimes momentarily jumping to 1.344, but not often) in cpu-z.

Ideas? btw, it's saturday here, so I can now post any screenshots you think are necessary.

 

[BonzaiDuck]

Originally posted by: Ararat
OK, I tried 370x9 and it seemed stable, at 380x9, core #4 failed in less than a minute, the other three a few minutes later (the computer reset). At 375x9 core #3 failed after not too long.

These were all done with the memory multiplier at 2.0, vCore at 1.45, FSB voltage at +0.2, northbridge voltage at +0.2, and memory voltage at +0.1 (timings at 6-6-6-18, just to be sure the DIMMS aren't responsible).

Here's the real kicker, 400x8 seems comfortably stable (I did tell you that this motherboard ran my previous cpu at well above 400), so it would appear that my cpu unfortunately isn't capable of getting above 3.3ish. Will this improve with time? (I remember my first C2D - an e6300 - was pretty stubborn initially, but over time i managed to get higher clock speeds out of it, be it due to the thermal paste or some other reason I don't know about).

I seem to have a major vdroop as well. My default vCore is 1.30v, and I have it set to 1.45. But under load (prime95), my core voltage is reading as 1.328 (sometimes momentarily jumping to 1.344, but not often) in cpu-z.

Ideas? btw, it's saturday here, so I can now post any screenshots you think are necessary.

Again -- I'm a very cautious over-clocker -- everyone here will tell you that. In hindsight, I've had two motherboards die in about four years' time; in one case, it seemed to be a memory controller (Northbridge) -- in the other case, most likely the Northbridge -- processors and memory proven good.

Unless you take measurements with a multi-tester, the difference between "set" voltage, "reported" or "monitored" voltage and "real" voltage is speculative. With my 680i motherboard, early BIOS versions made it possible to set the VCORE to 1.5+V, but the monitored value would only show as 1.44V. I think the i4Memory.com folks had used a multi-tester -- can't remember for sure. Eventually, BIOS revision made those two voltages more consistent.

You didn't say what your idle voltage is. The jump you see in PRIME95 may show part or all of a "vDroop," because the CPU-intensity of the program's iterations changes between the test sequences.

If you can trust the reported voltage value, then you could probably increase the VCORE again so that the reported value at idle is close to 1.36 or 1.37.

IF . . . I say IF . . . the reported value is close to the "real" voltage, then I think it's still good for 24/7 operation. My B3 stepping reports 1.41V at idle. This is with the 680i board and after BIOS revision at least a year after the board's release -- in which one of the revisions noted a fix to the "set" versus "reported" discrepancy.

I'm pretty sure that there were some threads here in fall, 2007 reporting people's results for the G0 stepping.

I thought I saw people reporting stable operation at 3.4 on air-cooling. But if I myself had such a processor, and that was the limit, I'd find the settings, keep the voltage settings, and drop the speed back 100 Mhz. Same if I found that 3.3 was as far as I could go and I'd chosen my own "voltage red-line" limit -- a self-imposed limitation.

 

[Ararat]

Idle voltage alternates between 1.392 and 1.376. I think that's a pretty significant droop to 1.328 under load, but you're probably right, I'm getting a little greedy here, and it's better safe than sorry. I want this thing to last until I can get a fast Nehalem for about what I payed for this one.

Unfortunately I don't have the electrical knowledge to take a multitester reading from it. I guess I'll have to make do with 3.3GHz, especially as temperatures might become an issue if I raise the voltage further (currently 62-65 C after a couple of hours). At that speed it should be great with multithreaded apps (though I am disappointed with the H.264 performance, given that it's supposed to support up to 8 threads), but will seem a tad slow with those that support only 2 threads.

If the motherboard is the precarious part in your experience, I guess it's a good thing my FSB is slowish at 371. I think I'm going to try to lower the FSB and NB voltages and see if it remains stable.