Recently I tried OCing it and I just haven't hit a snag yet... now I am up to a 3.10 Ghz OC and reduced my core voltage in the BIOS to 1.18 V (below stock!). I run the memory at 1:1, so now it is like DDR2-690 or something.
Prime 95 runs error-free for 6 - 7 hours, temperatures settling at 51 C (just about the same across 4 cores) (room temperature in New Jersey in the summer .. maybe 26 C?).
1. Can I reduce the voltage further and further? Any danger from undervolting? Will there be thermal or other gains?
2. At some low voltage, if Prime 95 can run error free for say 15 hours but produces an error at 23 hours, do I have to worry about it? When is normal operation ever like 23 hours of Prime 95? Or is this the wrong way to think about it?
Oh, in case people want to know, the temperatures above are reported by lm-sensors, a common linux utility. The BIOS reading is typically 3 - 4 C lower but then of course, the CPU cools a bit by the time I reboot to check the BIOS temperature.
the rest of my config (that I'm really happy with -- thanks to Tom's Hardware!):
OCZ Stealth X Stream 600 W PSU (got a good deal on this.. too much power for me)
XFX GeForce 8600 G? - with quiet Zalman fan ( don't game.. don't remember if it was GS/GT/..)
Western Digital 640 GB HDD -- best purchase. Quiet, cool and really fast (large cache - 16 MB, just two 320 GB platters)
Everything went into an older 2006 version of the popular Cooler Master Centurion 5 case. Added in 3 Scythe S-FLEX fans -- really quiet but with good airflow.
The PSU has a large 12" fan on it too that sucks hot air from the CPU -- most pleasant surprise in this building this computer.
Another nice surprise was the ease of mounting the Xigmatek S1283 heatsink using the mounting bracket (purchased separately).. just screw it in and tighten.
Remember that Intel sets the default voltage for each CPU based on extensive testing during manufacturing. If your specific CPU tested 100% at a lower voltage than the current VID voltage, it would have been given a lower VID voltage value. Thus, running at lower-than-VID voltage settings is a crapshoot. The lower you go, the greater the chance that during your testing you'll run into the problem(s) Intel found during manufacturing. If you're unlucky, the problem won't show up until after your testing phase, when you *think* the CPU is working fine. If you're really lucky, the problem will only affect some CPU feature that you happen to never use, perhaps something like virtualization extensions, and you'll never see it and will live happily ever after.
Of course they do! They have to make sure it will last with ZERO problems for at least 3 years straight. Once out of warranty, who cares, but until then, I am sure they have some head room to be sure its stable 24/7 for 3 years, no matter the temps and voltage within spec.
Under volting is interesting, because all boards are different. If your board has a vdrop and droop of .05 total, and I have .1 Then you will be able to "under clock" it below me, and make it look odd.
But in reality, it's made to go almost .2 under your VID and opporate.
So under clocking is in itself, and odd thing to do, as there are no way to establish if you are under clocking, or it just looks that way!