it is...
two of the four channels have 2 slots each and the other two slots have only one slot each. because of the serial nature of the fb-dimm architecture this setup should work fine (i'm using 6x 1gb modules right now). however as more modules are pluged into one channel, you mem latencies are a bit higher but also you mem throughput raises.
as you can see this is not an optimal setup as 2 of the four channels have higher latencies but also higher throughput and the other 2 channels have lower latencies but also lower throughput.
at the end i think it doesn't matter much for real world applications imho...
if anyone is interested in plain numbers let me know which benchmark i should run.
ps: anything like sisoft sandra or everest seems not adequate for testing ram performance on a quad-channel mainboard from what i've seen. for example sandra gives me roughly half the mem throughput on vista64 like it does on vista32 on the exactly same system without any changes to the settings in bios nor the configuration. anyways the measured ram-performance is quite low no matter what slot-configuration i use (2/4-dual/4-quad/6).





Reply With Quote

Bookmarks