I've spent a productive couple of hours tonight tuning my DRAM settings. I'd already done a gross adjustment the other night, knocking the Active Precharge setting down from 15 to 12 and kicking the clock rate up from 266MHz to 533MHz, and that was nice and stable. This evening, it was time to have a play with the other settings: RAS-to-CAS Delay, CAS Latency, and RAS Precharge. The SPD timings were a conservative 5-5-5-15, which was now set to 5-5-5-12, and I wanted to see how far I could go.
Just for a chuckle, and before doing anything else, I wondered what would happen if I kicked the DRAM clock up to the max - 667MHz. The answer was; not a lot. The box stayed stable, but the FSB:DRAM ratio drifted from 1:1 to 5:4. I decided I liked the former better, and dropped back to 533MHz.
Turning to the RAS-to-CAS setting, I dropped that to 4 and got a stable boot. Pushing the envelope by nudging it down to 3 gave me a lockup at POST, so a swift BIOS reset followed and back to 4 we went. Exactly the same pattern emerged for CAS Latency and RAS Precharge - 4 was good, 3 wasn't. Finally, with the timings stable at 4-4-4-12, I recalculated the Active Precharge and set it to 4-4-4-10. I'm now into the third Prime95 Torture Test pass, and everything seems to be looking good so far. Note also that I haven't had to bump the DRAM voltage up either, so under load the rig is peaking at about 55C, which is fine.
That pretty much brings us to the end of the BIOS-tweaking phase - the next and final step is to fire-up RivaTuner to see how far I can stretch the GPU and GDDR clocks to squeeze some extra horsepower out of the graphics card. That will have to wait until tomorrow, as it's getting late.
Just for a chuckle, and before doing anything else, I wondered what would happen if I kicked the DRAM clock up to the max - 667MHz. The answer was; not a lot. The box stayed stable, but the FSB:DRAM ratio drifted from 1:1 to 5:4. I decided I liked the former better, and dropped back to 533MHz.
Turning to the RAS-to-CAS setting, I dropped that to 4 and got a stable boot. Pushing the envelope by nudging it down to 3 gave me a lockup at POST, so a swift BIOS reset followed and back to 4 we went. Exactly the same pattern emerged for CAS Latency and RAS Precharge - 4 was good, 3 wasn't. Finally, with the timings stable at 4-4-4-12, I recalculated the Active Precharge and set it to 4-4-4-10. I'm now into the third Prime95 Torture Test pass, and everything seems to be looking good so far. Note also that I haven't had to bump the DRAM voltage up either, so under load the rig is peaking at about 55C, which is fine.
That pretty much brings us to the end of the BIOS-tweaking phase - the next and final step is to fire-up RivaTuner to see how far I can stretch the GPU and GDDR clocks to squeeze some extra horsepower out of the graphics card. That will have to wait until tomorrow, as it's getting late.
