|
3DMark to start proceedings. Radeon 9700's in both setups for a fair fight, lets see how the XP2700+ gets on with its new found memory bandwidth and lots-o-MHzTM. Will it get close to 2.8 and RDRAM and keep an eye on the performance so you know where XP2800+ would sit.
An early victory for the 2.8 but don't count the XP2700+ and nForce2 out just yet. The P4 takes the victory in the Low Detail tests since it's all about system throughput when paired with an R9700 Pro and RDRAM buys you that, especially in PC1066 mode with a 2.8Ghz CPU behind it.
XP2800+ with its few more MHz would close the gap but the 2.8 would still edge it out.
So what about Quake 3 in Fastest configuration? 512x384 and 16-bit colour and vertex lighting will bring out the bandwidth in any system. RDRAM and the 2.8 should walk away with this, by design the memory bandwidth and CPU bandwidth are both a lot higher on that system. Here's the graph.
I re-ran this on the XP2700+ about 10 times to verify and checked that the configs used on both systems were identical but the system just doesn't have the throughput the P4 system does.
Still, not shabby performance and with an R9700 it translates into 1600x1200x32, 4xFSAA and aniso filtering being massively usable.
Serious Sam 2 next, lets see if the same pattern shows up.
Another win for the 2.8 but this time it's a bit closer. You can see that certain tests highlight certain strengths in the test configuration. XP2800 would more than likely go neck and neck with the 2.8 and the scores would be very tight.
As far as nForce2 helping performance along, check out Tech Report's latency analysis that they do with ScienceMark and Linpack to help you see why there's some new found performance over a VIA based chipset running at the same settings. It's no coincidence that NVIDIA and nForce2 are AMD's chipset of choice for these new processors.
Finally a chipset for Socket A that holds a candle to an RDRAM box given a fast CPU.
|
