Poor Q6600 Overclocking

[BonzaiDuck]

Sorry I misinterpreted your "iconic-appearance," Ry . . !

I'd investigate that PSU question a bit more.

Right now, I'm using a 700W Seasonic M12. This may be more than I need at the moment, but I'm running four Seagate 7200.10's in RAID5. I'm still just a tad worried that adding a second 8800 GTS to the mix may be like the U-boat in "Das Boot" trying to slip through Gibralter. Probably not, but it could come close, if only for the Extreme Outervision Calculator, which -- as I understand -- over-estimates power requirements.

OC'ing is going to increase the power-draw, as you seem well to know. And PSU's degrade a bit over time -- when they also may not fully live up to the factory-spec, or the 12V-rail(s) may not have the needed muscle as compared to a different PSU sporting the same overall wattage rating.

I'm up top 3.25 Ghz and stress-testing now, and glad because I wanted to get the FSB back to 1,440 and get some mileage out of the RAM -- which gives some nice, tight latencies at that speed. this has pushed my thermals up to a Core 0 (always the hottest) reading of about 62 to 63C at room-ambient currently 75.5F.

 

[aguilpa1]

Is he sure he doesn't have the AR Evga 680i SLI instead of the A1?

Reason being I couldn't get mine to overclock worth squat past 1200FSB, max 1220FSB regardless of voltage till I did some digging and found out mine was an AR not A1. very unstable.

I just got the new board installed late, last night, its already OCed to 1333FSB 3.0Ghz, stock v. I will start more serious oc tonight.

 

[aka1nas]

Originally posted by: airhendrix13
Well, it seems that my CPU absolutely refuses to go above 3ghz, I can only max out at 2.9ghz. I'm also thinking that my PSU might not be feeding it enough power because it is only a 420 watt PSU I believe. I don't believe it's cooling issues due to the fact that at 2.9ghz, my CPU only reaches 58C and my case temps around 45C after running Orthos for 10 minutes. I might still try your ducting mod though just to knock a few degrees off.

Thanks for your help, I'll let you know if anything changes.

Ryan

Make sure you are using Core Temp to check your CPU temps. Speedfan and several other common programs are still reading 15C too low on Quads.

I'm putting together a system with the same board(T1 bundle) and CPU right now, and it looks like it's FSB walling past 333FSB(1333QDR). It's passing prime at 3Ghz right now, but I can't take the FSB much higher, even if I drop the multiplier all the way down to 6x. Supposedly the newer P30 bios that Gary reviewed is better for Quads and might be worth trying for you if you don't already have it.

 

[BonzaiDuck]

ala1nas --

I'd like to think you could tweak the VTT, NB, or even Memory voltages and get past that.

I don't want to sneer at either you or airhendrix. But imagine my thoughts about some others here who've thumbed their nose at my Striker Extreme board. I was beginning almost to believe it.

I'm testing at 3.2 Ghz now, and 356 CPU-FSB. I've got the VCORE at 1.4125V. I'd set it as high as 3.24 Ghz and 360 -- even testing at 362 -- but it just keeps wanting more voltage and I think the wise thing to do is accept the limits of the B3 stepping with this voltage and lower speed.

Also -- check the manufacturer web-site for BIOS revisions. Whether or not they're new to the business as opposed to Intel -- and so they even admit -- recent BIOS revisions may be catching up to the Kentsfield. People may doubt it, but I've noticed some improvements that weren't so subtle that they failed to get my notice or surprise me a bit.

the fact is -- I wasn't able to get beyond 2.92 Ghz using an 8 multiplier. Then I flashed the BIOS to a July '07 revision. And suddenly, it was pretty darn easy. In fact, I'm going back to that 2.98 setting to tweak it so I can get 3.0 Ghz at multiplier 8.

 

[Markfw]

First BIG problem. Under load, your vcore needs to be NOT under 1.328, and you are at 1.28 ??? Not good. Up that baby big time. Keep upping in bios until under load it never drops under 1.328. And what PSU do you have, that can be a big problem also, as these babies need a lot of power.

 

[PCTC2]

what revision of the board do you have? AR/A2 and TR/T2 revisions do not support Quad-OCing, even after BIOS updates. Also, check your BIOS for vDroop. The eVGA board is vicious for vDroop problems. With my vCore set at 1.34 in the BIOS, actual vCore is 1.28. Also, update to the P35 BIOS and try bumping it higher (400*8) b/c there can be nasty FSB holes as well.

EDIT: If you bump to 400*8, you might need to bump your vSPP and vFSB as well.

 

[BonzaiDuck]

I've got a discussion topic I'll put up as a separate thread, relating to "FSB holes" and possibly linked items that one would think are separate.

On the matter of vdroop, whichever one of you said that is likely correct. There is this inherent fear of going too high, or thinking that "set" voltages will either match monitored values or "real" values. I think there is a tendency to improve the coincidence of these things with further BIOS revisions.

Because the C2Q has four cores -- essentially packing two E6600's into one package -- I was much more reticent about upping the VCORE on the C2Q than I was, say, on my old Northwood 3.0C.

But with either the old Northwood or this year's Kentsfields, the retail box has a voltage maximum specification. On the Kentsfield B3 stepping, it is 1.35V, and that may also hold true with the G0 stepping, but I cannot at this time confirm it (I don't have one.)

At my first crack at the B3 Q6600, and in order to get to 3 Ghz FSB 1,334 DDR2 667, I nominally chose to volt the processor at 1.325V -- below the box maximum and just at the upper limit of what (I thought someone said) was the normal operating range. That setting was not too high, but it was higher than necesssary, and I trimmed it to 1.31875V. It would droop to 1.29V -- (emphasis) -- ON MY MOTHERBOARD.

To go from 3 Ghz to 3.2, I had to bump up to at least 1.4125V, and to get even temporary stability for 3.24+ for more than an hour with 2xORTHOS, I had to push it up to 1.425 and beyond. [I've since realized that there may be another reason for this, but these processors have their limits. They're all a little different. Even so, someone with water-cooling has volted theirs to 1.45V.]

Now I see that the load voltage droop takes the 1.4125 value down to maybe 1.38V, and that's only some 2% above the retail box maximum. Only problem there is that more voltage means more heat, and there will be a point where the TCase temperature is pushing closer to 65C.

Just keep in mind this is all about probability. "What is the probability that a voltage only 2% above the retail box maximum will damage my processor?" The answer is: "Not nearly as great as a voltage that is 10% above the retail box maximum."

And the "set" voltage may give you a slightly lower "monitored" voltage reading at idle, while the load droop will push it close or under that maximum spec. I think the load value is the value relevant to our probabilistic worries.

 

[Markfw]

on vdroop. On my ASUS P5K@3.2, I set the bios to 1.4750, it boots at 1.392 (or a little more) and under 100% load its 1.328. And anytime its unstable, is when the vcore dropped below 1.328, so I would go back to bios, and up it one more notch.

 

[BoboKatt]

We've still not figured out what version of the 680i he has. If it's the early one AR like mine, as others have said no amount of fiddling will get you a decent OC using a QUAD. Period. End of discussion.

If it's a newer revision then disregard.

 

[BonzaiDuck]

Like all the earlier patterns in the motherboard and chipset market, BIOS revisions are likely in the pipeline to fix little imperfections in the P5K. So be on the lookout for new BIOS revisions promising to correct discrepancies in VCORE settings and VCORE sensor readings. I would, however, wait several months until a cumulative BIOS version provides a lot more fixes. In the meantime, be careful about over-volting.

What Markfw900 describes is the same pattern that emerged with the release of last year's Striker Extreme and P5N32-SLI boards. Those things have now been fixed on my Striker, and we're headed toward a full year since that board's release. But confidence about VCORE settings and readings was hit or miss before those BIOS revisions were made. Anyone accustomed to better -- and we all probably are -- would've called it "borky."

How time flies . . . .

Per BoboKat -- It still may be a matter of BIOS revision, but then again, maybe not. But given last year's release-date for the board and other 680i boards, I'd be stunned if they had to revise the hardware in addition to the firmware to fix it, or if the deficient production runs were still being sold off. Easily possible worst-case scenario, however . . . .

I recall reading a letter on a Australian web-site with a lot of Striker Extreme enthusiasts, letter having been reprinted from some publication -- it was an interview with an nVidia CEO about glitches with the 680i chipset and boards. Apparently, it was not a problem with my board per the executive's remarks.

 

[eelw]

Originally posted by: airhendrix13
I'm also thinking that my PSU might not be feeding it enough power because it is only a 420 watt PSU I believe.

With my Q6600, at bootup, it peaks at 275W. At idle, it consumes 180W. Under CPU full load, 255W. While running 3DMark with my X1900, it hits 355W. Currently, I have 4 HDs connected to my system. Running a 450W Coolmax PSU with my quad OCed to 3.1 Vcore 1.4V.

So I don't believe your PSU is holding you back.