Hmm... low-level RAM timings... I'll give that a shot soon.
Here's some interesting data points for you.
While troubleshooting S3 power-saving issues, I gave up on getting S3 to work w/ > 266 FSB and simply decided to isolate the factor that was causing the resume/boot issues.
To begin with, I do get the double-boot reset to stock FSB sometimes.
C2D E4400 L2 stepping rated for 1.325v
GA-P35-DS3R rev. 1.0 F11 BIOS
HP DDR2-667 RAM rated for 5-5-5-15 @ 1.8v
Here's a quick overview of what happened on my rig:
200 * 10 = 2000 mhz @ 1.325v (all stock, no issues)
266 * 10 = 2660 mhz @ 1.350v (my 24/7 o/c, no issues)
300 * 10 = 3000 mhz @ 1.500v (S3 fails 100% of time, o/c occasionally resets on cold boot)
(Note: do NOT set your vcore to 1.5v unless you know what you are doing!!)
Now, as I said, I wanted to figure out why that 300 * 10 @ 1.500v fails S3, so I tried a variety of odd settings to figure out whether it was the FSB clock rate, CPU clock rate, the CPU voltage, or some wierd BIOS issue (like that Legacy USB disable trick, which doesn't appear to make any difference on my rig).
Here are my notes:
2.66 GHz --> works fine
3.00 GHz --> suspends but won't resume
|
+--> resume failure: tries once, hanges on 2nd attempt; hold power switch off for 4 seconds, which leads to multiple failed cold boots (no user intervention); o/c gets disabled
|
+--> disabled C1E/EIST --> S3 fails
|
+--> increased voltage to 1.525v --> same problem
+--> decreased voltage to 1.475v --> (not tried)
|
+--> tried 300 * 6 @ 1.350v --> works fine
+--> tried 300 * 6 @ 1.500v --> S3 fails! (surprising!!)
+--> tried 200 * 6 @ 1.500v --> POST fails, had to unplug and reset CMOS (even stranger!!)
+--> tried 333 * 6 @ 1.350v --> S3 fails! (also odd...)!
The double-POST issue is the BIOS's failsafe mode kicking in because the board has failed to boot and POST (probably with a preprogrammed timeout). The second boot disables the o/c so that you don't have to open the case and reset the CMOS manually.
Whenever your o/c settings are stable enough to start POST but not stable enough to get past the drive detection, Gigabyte's failsafe will get "stuck" and not allow any o/c settings to stick -- not even with
Idontcare's double-save method -- until the CMOS jumper is used for ~2 seconds with the PSU unplugged from the wall. I hit this with once as detailed above. I've also hit this with 2.9 GHz @ 1.325v, and with certain memory settings.
The CMOS jumper location also sucks. I must remove my 8800GTS to short the jumper. Surely, Gigabyte could have put this at a more convenient location!
Anyway, the fascinating part is that the voltage and FSB appear to contribute to the problem. The board boots consistently (but not 100% of the time) with 300 * 10 @ 1.5v, but 200 * 6 @ 1.5v fails to boot every time. Yet 300 * 6 @ 1.350v works fine, even with S3 mode. Either there is such a thing as too much voltage, or my system board has some power regulation problems.
I'll see if manually setting the memory timings fixes the S3 failure w/ 333 * 6 @ 1.350v. I also remember getting general stability problems with 333 * 8 @ 1.350v whereas 266 * 10 @ 1.350v works fine 100% of the time. That's two problems at 333 FSB, occurring regardless of voltage or CPU multiplier. So it could be memory instability.