Memory Benchmarks
A quick word on the performance to expect from KT880. We've already mentioned that the DualStream64 memory controller setup makes a lot more sense with a processor that can take advantage of it. KT880 was never going to be much faster than KT600, if any faster at all. So we're just looking to see if it can keep pace with nForce2 Ultra 400, nothing more.Our first set of tests are memory influenced. That is, changing your memory configuration will have a noticable effect on the results of the tests. Pifast to kick things off, it relies on low latency main memory accesses and memory bandwidth to run quickly. Remember, memory latencies are 2-4-4-8 on each board.
A quickfire initial win for the nForce2 Ultra 400 board. Sciencemark should show us why. Bandwidth first.
Hmm, nearly 93% efficient when streaming memory from controller to CPU and a win over nForce2 Ultra 400 in the bandwidth stakes, at the same memory timings. Latency must be longer surely, for the Pifast result to make sense.
Not so, according to Sciencemark at least, fetching data from main memory is no slower than nForce2 Ultra 400 at the same memory timings. Something else explains the performance delta in that case. While it might appear to have the same latency as far as Sciencemark is concerned, it can't be the general case, either measurable by software or otherwise. Internal chipset timings seem to be the likely candidate. Let's see what KT880 makes of our other system and CPU tests.
