Sometimes a little FSB boost goes a long way — Buzz Bizzell
I am writing this story from my “just saved from a nasty fall off a tall bridge” computer. Saved because the very fault in the motherboard that had me ready to trash it has worked out to be a good thing in terms overclockability and just plain usability.
The story begins with a poor, yet well-intentioned, individual who built a fairly nice box in February of 2001. Yes I said 2001 – not everyone has the scratch to buy a new processor or mobo each time a new one comes out…sorry, I digress. The system is an Asus P3V4X, PIII 700, and 512 RAM, in addition to the normal host of things.
I began to have problems immediately with the P3V4X – the board would not boot at any bus speed with the 700 installed. I tried to run the 550 the 700 was replacing and it would boot only at 366 MHz. Not funny.
I took the board back to the reseller and they were not too happy, but eventually they replaced the board with a new P3V4X (no swapping it for a different board). The replacement board worked fine, except the -5 volt line ran 7% out of specs – a worry but nothing major. After setting up the box and deciding on a cooling scheme, I set the CPU to 933 and forgot it.
The comp ran great for months and I gave it no thought until last November. I started to get strange problems, such as programs which had run flawlessly would now crash with no discernible cause. I suspected a heat problem and opened the case for a good cleaning. Sure enough, there was a layer of dirt thick enough to grow crops in. All cleaned up, I went back to using the comp, but the problems not only persisted but grew worse.
Becoming frustrated, I decided to change from Win98 to W2K. Nope, not the OS either. Two new 80 mm fans: Nope again. Heat sink upgrade: Sorry. Lapping, Arctic Silver, cursing, and madness ensued, all to no avail except to lower system and CPU temps to all-time lows. I was at the end of my patience.
As I sat wondering how much the fine would be for hucking this POS from the Coronado bridge into San Diego Bay, I noticed something in the hardware monitor: The RAM voltage was running at 3.5 volts instead of 3.3 volts.
Could this be important? Could the higher voltage at 133 FSB screw the transient response to the point of data corruption? I thought so, but more importantly, I have nothing to lose since a comp that crashes is worthless.
With no further ado, I pushed up the FSB to 142, the core voltage to 1.95 and began to test the programs that had given me the most trouble.
All passed with flying colors!
Now I have a stable comp again, albeit for an indeterminate amount of time, and a boost of ten frames in Q3 (175 w/gf2 32 meg). Sometimes overclocking is more than a good time.
PS: If anyone wants my config and demo for Q3, I’m happy to share.
Discussion