Benchmarks I
Let's look at the unbuffered memory benchmark of the KT600 first.
SANDRA seems to like the KT600 from the start. Its performance is on a par with the nForce2 Ultra 400 running in single channel memory mode. SANDRA isn't the definitive answer when it comes to evaluating true latency and bandwidth properties. We'd prefer a more real-world benchmark for that. So let's run Pifast.
Ouch, what's happening here ?. The KT600 was close to SC nF2 in SANDRA's eyes. It comfortably distanced here. This is no one-off anomaly. We tried almost every conceivable combination of timings and settings in BIOS to make it quicken up a little. No joy. The time represents 2-6-2-2 (1T command). System performance was Normal. There's little explanation other than it's simply slower here. We also tried a couple of different Hyperion drivers (4.46 and 4.48) to no avail.
WAV crunching now.
Performance parity is restored via the use of a almost completely CPU-bound benchmark.
SETI, on the other hand, is anything but a pure CPU benchmark. Time is quoted in hours, minutes and seconds. A typical 0.417WU.
The interaction of the CPU, memory controller and system RAM doesn't appear to be quite as good as the nForce2's. It's all a little perplexing when trying to explain why.
