Overclocking voltage worries

[TheChuckster]

First of all, my system specs are:
Antec True Power II 550W
Intel Bad Axe Rev 306 using the latest stock BIOS
OCZ Special Ops Edition DDR2 667 memory (rated at 4-4-4-12 and 1.9v but lifetime warranty up to 2.2v)
Core 2 Duo E6600 (2.4 GHz stock)
Zalman 9500 Copper fan with Arctic Silver 5
eVGA GeForce 7950GT
Zalman VGA Copper fan
http://thechuckster.homelinux.com/~chuck/build1.jpg

To overclock my CPU to 3.1 GHz, I went into the Advanced Chipset menu of the BIOS and set the host burn-in to positive 30%. I left PCI and PCI Express burn-ins to default. Then I bumped up the MCH voltage to 1.525v and the FSB voltage to 1.271v. For 1:1 ratio, I set the memory to 553 MHz 5-5-5-15 at 2.2v. My BIOS does not have a strap setting (is that what the burn-in does?), no power slope (supposedly this automatically ramps up the CPU voltage when under load), or CPU voltage (is that what the FSB voltage override does?) settings. Of course, I disabled the Intel Speed Step feature, and I set all the fans to the max. It was Prime95 stable for a 3 hour test, and my CPU is now running at 3.1 GHz.

Then, I brought down the memory to 4-4-4-12 @ 2.2v and Prime95 tested overnight. Stable, and the 3D Mark results are wonderfully tempting.
http://thechuckster.homelinux.com/~chuck/overclock.png
http://thechuckster.homelinux....uck/3dmark_results.png

Temperatures seem a little uncomfortably high (word around the street is the board sensors are 10C too high... can someone verify?). The idle was 40C and load at 50C at stock frequencies... now its 43C idle and 58C load.

So I took down the MCH voltage to stock, and ran Prime95 for three hours... still stable.
Then, I brought the memory back down to 1.9v. I'm running Prime95 right now... it's been almost an hour, and things seem stable. Now I just have the FSB voltage at 1.271v.

My question is... are these voltages reasonable for the 3.1 GHz overclock? I'm completely new to overclocking, so I have no idea beyond here. I can just keep listing all of my questions: Should I bring all of the lowered voltages back up as a safety net? Was I overvolting the system to begin with? Should I bring down the FSB voltage and run it at stock as well? Am I undervolting the system for my 3.1 GHz clock speed? Is one instance of Prime95 an adequate stability test? How long should I run it? What I don't want to do is kill my system (or any components)... is this overclocking a bad idea then? Should I go back to stock speeds? How much of a lifetime should I expect of my PC? Am I doing something extremely stupid that I haven't picked up on? How unstable is my PC going to be? Will I fry anything?

I would greatly appreciate it if an experienced OCer could fill me in on all of the details. I don't plan on maxing my system out... I just want a stable speed boost for day-to-day computing that won't introduce risk of frying things. How much risk is there at this point? What can I do to correct that risk?

 

[BTRY B 529th FA BN]

you can always expect a shorter life span of your components when you over-volt above factory recommendations

 

[TC91]

the e6600's have a tjunction of 85c, so anything under 65c (preferably 60c) is good for the cpu temps. one instance of prime95 is definately not an adequte cpu test. you have to stress all of your cores when overclocking. i prefer doing 24hrs of dual prime 95 (for dual cores of course). for the lifespan of the cpu, as long as you can cool it properly and keep the voltages close to stock voltages, then the lifespan will not be shortened by much if at all. the mch voltage @1.525v looks a bit high imo. i dont know what you are using for your cpu core voltage, but that is a bit more important than the fsb voltage. other the the mch voltages, your other volatges sound reasonable. you might be better off running the cpu @ 3ghz (333x9) if you are too worreid about component lifespan.

 

[BonzaiDuck]

Ditto on the importance of the CPU core V. I'm a bit stunned that your BIOS doesn't have a setting for it.

With my 680i board, I've found two OC settings for my Q6600: 3.0 Ghz and 3.2 Ghz. The VCORE for the lower setting is 1.318V, and for the higher setting it's 1.412V. The retail-box maximum is 1.35V. Voltage droop under load brings the higher-setting's sensor-reading down to 1.37.

I'm still experimenting with stability while tweaking the SpeedStep settings in BIOS and the Control-Panel "Power-Options." Things had seemed rock-solid until yesterday, when I changed the Power-options from "Always On" to "Max Battery," and got a BSOD. This is the first and only BSOD I've ever had with this system and with the 3.2 setting. I'd discovered that "Always On" overrides the SpeedStep and holds the speed at 3.2.

I may just content myself with leaving SpeedStep off and using "Standby" to reduce power-consumption when I'm not at the machine.

 

[TC91]

BonzaiDuck, if you turn on c1e and leave speedstep off, your cpu will still reduce in speed and maybe voltage. i do this for my e6600, and it is completely stable, just not sure if the voltage is actually decreasing as cpuz doesnt show it. i initially had speedstep on, but windows kept reporting my cpu speed @ 3.39ghz and when i had it @ 3.6ghz, it showed it as 3.59ghz which annoyed me, and i turned it off, while leaving c1e on. c1e on my setup appears to do the same thing as speedstep, at least when i look at cpuz.

 

[BonzaiDuck]

Interesting -- to have your remarks here, TC91.

I've noticed that Everest Ultimate's "CPU_Stress-Test" doesn't play well with SpeedStep. I think it even gave me some trouble with ORTHOS. For some time, it appeared that Speedstep was completely stable with my configuration:

Q6600 (B3) @ 3.2Ghz, 1.413V-core
Striker Extreme, BIOS v 1305
3 GB w (2x1GB + 2x512MB) Crucial Ballistix DDR2-1000 @ (DDR) 712 Mhz, 3,4,4,8, 2T
BFG (nVidia 8800 GTS 640MB @ 576(core)/1800 (mem)
3Ware 9650SE 4-port PCI-E (x4) RAID5 w 4x Seagate 320GB 7200.10 SATA2

Given your uncertainties, perhaps we should all look into this a bit further. Your information definitely adds something here.

I'm especially interested in your 8800GTS over-clock. With ATI-Tool, I was able to confirm that my core will go as far as 620 and the memory to 1030 or (DDR) 2060. The VGA cooling seems to keep the GPU temperature below 54C under load, and the GPU-Diode below 63C at 75F -room-ambient. So I could push it a little higher, I think.

The troubles with SpeedStep confused the stability picture just when I was discovering that ATITool is not the best choice for OC'ing the 8800GTS. I was getting crashes and lockups in the middle of gaming; switched to RivaTuner, and those troubles seemed to disappear. I'm not yet ready to flash the VGA BIOS. But with tweaking the Control Panel->Power-Options, I think SpeedStep is still causing some trouble. I've still got it enabled today, but am keeping "Power Options" to the "Always On" profile.

Another among us here had noted that he'd OC'd his system -- similar to ours and with the Q6600 -- with SpeedStep enabled. I think I'll simplify my BIOS settings (disable SpeedStep) for now, and look into these things further.

Your E6600 at 3.4 is pretty good if you've set VCORE to 1.412V. I was able to run mine at that voltage and 3.2Ghz, only after I'd cooled down the 680i chipset. That was before I swapped in the Q6600. And I might have done better with the E6600 by bumping up some other voltages a tad here and there.

 

[VirtualLarry]

I thought that the 680i chipset boards had "issues" with the Q6600.

 

[BonzaiDuck]

See, I started planning this year's system early in '06, and was looking at a P5WDG2-WS-Pro (whew) board with an Intel 975 chipset. Around August of that year, friends were all goo-gah about the Striker Extreme 680i board. there were forum posts here and there with people having trouble with the earlier BIOS revisions. There was controversy and ambiguity -- some good reviews for the board appeared, but some customer reviews at various resellers were pretty nasty, and you had to do your homework to eventually conclude they were enthusiasts who didn't know what they were doing with the board. For a while, I just shined it on, as we used to say in the '60s, and my friends inclined toward 965 boards. I watched certain forums for a while.

Then it started to look much better -- with ASUS coming out with BIOS revisions almost twice a month.

By the time I bought mine -- for use with an E6600 -- I didn't have any trouble with it. People were telling me that ASUS lies -- that it probably wasn't Kentsfield compatible, but it was.

There were all sorts of troubles with the 680i boards generally -- and an nVidia executive was interviewed. But -- keep in mind he was an nVidia executive -- he noted that ASUS 680i boards were an exception. However it might be the case, if people didn't have problems with these boards, they may have eventually experienced troubles from running up the clocks and voltages without proper attention to chipset cooling. But I had that problem nailed down early -- and without a lot of expense.

I never had a problem with my Q6600 B3, and there wouldn't be a problem with the G0 stepping. Then I discovered that ASUS had updated their spec sheet on the Striker 680i, saying it was compatible with 45nm processors -- meaning the Penryns -- which hadn't even been released.

Now I find that the Striker 680i WILL fit a Penryn, but only a Wolfsdale. They haven't certified it for the Yorkfield quad-core, and unless a BIOS revision resolves that difficulty, my best bet to "stay quad" is either a QX6850 or a G0 stepping for Q6600. Or perhaps a Q6700 -- but at least the 6850 runs stock at a 1,333 FSB -- the top stock speed spec for the Striker.

Others have told me to junk it, sell it, and go with a newer Intel board, like the P5K or an Asus Maximus.

I'll just have to save some pennies and wait and see. Meanwhile -- with a 33% over-clock across the board (ha! -- a pun!), and a modest over-clock of my BFG 8800 GTS VGA card, it seems to be humming along just fine.

 

[TC91]

BonzaiDuck, for my 8800gts, i am currently using rivatuner with the startup overclock option. i previously had the fan on my gts set to 100% and the load temps were only 64c, and i thought it could go higher, but i am sure it will need more voltage to go any higher, just like cpu's. i currently have the fan set at 70% for more silence, and a little better cooling over the stock of 60%. i didnt flash the bios, and i dont plan to volt mod it as i think it would be a bit of a hassle. for my e6600, i had it set @ 3.6ghz 1.5125v in bios, 1.48v actual, but i deemed that too much voltage for only 200mhz and i backed down to 3.4ghz (1.4125v in bios, 1.38v actual). i ve seen other e6600's hit 3.6ghz with 1.35vcore too, but i am not sure about how the user stability tested it. for me, my e6600 failed @1.39325v in bios or something close to that after 8hrs of dual prime 95, and @ 1.4v in bios fail in 14hrs of dual prime 95, but 1.4125 passed 24, almost 25hrs of dual prime95.

 

[TC91]

BonzaiDuck, i dont think you should sell your board. its plenty fine, and upgrading to a g0 q6600 wouldnt really be worth it as you may only get 200mhz more out of it, and the qx6850 would likely cost a fortune. if you are a gamer, you would be better off by getting a newer graphics card when you deem yours is insufficient as your b3 q6600 should be fine for quite a while and should not really hold back a newer graphics card too much.