Friday, 9 November 2007

Well, i finally did it!

It's been a goal of mine since March, when i upgraded from my Dual Core Opteron 170 to my current Core2Duo E4300, to crank it to 3.2Ghz and have it stable. So far, my attempts have always stopped short of the mark. The most i've been able to get out of it is 2995 Mhz (333Mhz FSB x 9 ). Despite repeated attempts to get it stable at anything more than that, no matter the memory timings, vCore or vMem - it's been a no-go. I could get it to boot into windows at 340Mhz x 9, but again, Orthos would fail after a few seconds.

Recently though, i've re-visited the idea that its the motherboard itself holding me back, and not the CPU. The board i have at the moment, is an Asus P5B-E, a pretty standard enthusiast board, but no-frills in its execution. I bought it when it first came out, so i ended up with the unfortunate 1.1 PCB revision, so i cant increase vMem past 2.1v and vCore is limited, as well as there being no provision for altering the NB voltage. Shame really. The big problem with the board is that it suffers from HORRENDOUS vDroop.

For those of you that dont know what that is, basically, when you set the voltage in the BIOS, say 1.475v, you should end up at that voltage under idle, but load, you'd drop to around 1.46v. I was having to crank up the voltage in the BIOS to 1.5125v to get it to even BOOT at 3.2Ghz previously, but even then, CPU-Z and the BIOS's own hardware monitor reported the CPU at 1.474v. Under load, this dropped to 1.440v and was all over the shop - not exactly stable at all. Orthos did its usual 'BIG RED FAIL!' of course. So this time i scouted XtremeSystems and OCForums and found a mod for the regular P5B board (non 'E') - however, the P5B-E uses the same VRM Chip (ADP3198). Now, there are two ways of modding the board - one involves some pretty delicate soldering to the SMT resistors, one involves an HB pencil. Now, i dont have a soldering iron even REMOTELY fine enough to try the first method - so i went down the pencil route. It's easier to reverse if things go wrong anyway :)

So, i shade over the resistor until the top is almost totally blackened - making sure the pencil line is actually touching the solder on either end. Measured with a DMM, the resistance drops from 144k Ohm to a mere 60k Ohm. Everything got put back together, and i booted at the same settings as before. For a start, the voltage reported in CPU-Z and the BIOS hardware monitor was a LOT higher - a fair whack closer to what the actual value I *set* in the JumperFree Config.

So, i dropped the voltage down to 1.4750v and booted again. To my surprise once more, the system booted, entered windows, load temps had dropped and CPU-Z was reporting 1.464v as the idle vCore. So, i fired up Orthos, expecting it to immediately skitz at me....and it didn't. I'm still running Orthos - and after just over an hour - the load temp is 69*c (a little hot, but then its warm in here), but theres been NO errors. It looks stable!

"So, whats the load voltage?" you might ask. How does 1.456v sound to you? ;)

More testing is in order, i might even be able to drop the voltage a little bit more...

However, in conclusion - WOO and YAY.