|
SETI is heavily dependant on a few things. A fast CPU to process the potentially devastating results (what if you find alien life?), a chipset that allows the CPU to work to its full potential and interface with the final important ingredient in a quick SETI system, memory bandwidth.
I've ranted on about AMD's poor show in not moving their high end CPU's to 166MHz front side bus and allow them to make use of all the extra bandwidth provided by these new chipsets like KT333. As you no doubt know, an Athlon processor at 133MHz front side bus can only make use of around 2.1GB/sec of memory bandwidth and KT333 in 333MHz memory mode offers 2.7GB/sec, a defecit of 0.6GB/sec. A move to 166MHz front side bus would let the CPU eat the defecit and boost performance across the board.
However that's not the case at the moment so the XP2200 is somewhat crippled. So with KT333 behind the CPU and feeding it all the bandwidth it can eat, what's the performance like when doing a task like SETI and how does it compare to the P4?
There's not much in it between KT333/XP2200 as seen on the Gigabyte and the SOYO review board. Indeed, the graph shows no tangible difference and the P4 takes the lead here. All in all, strong performance from the SOYO, highlighting that current KT333 solutions are within touching difference of each other.
Onto LAME. LAME is almost entirely CPU limited so look for an identical set of numbers to the Gigabyte and note the comparison between P4 2.4B and XP2200.
Fancy that, identical, winning performance from the SOYO and Gigabyte boards and a full 30+ second advantage over the P4 solution. Given the results from the XP2200 review, I bet you couldn't have predicted that! Thank god we weren't expecting anything different :P
Joking aside, we can see that doing media work (remember that LAME is an MP3 encoder), KT333 on the SOYO is a good solution when paired to a recent Athlon XP, allowing it to overtake Intel's multimedia champion, the Pentium 4, in this test at least.
So what about another test that enjoys memory bandwidth and raw CPU performance? As we run out of tests for you to look at, let's have a look at Pifast.
|
