I disagree, I get 90fps in 1024*768* 32bit colour in HQ mode in Quake III Arena on a 1.4 GHz Athlon. That is also what the average GTS gets at the same res. On a clock for clock basis, DDR RAM is less efficient than SDR RAM, meaning DDR's average bandwidth is farther from it's maximum bandiwidth than SDR is from it's maximum bandwidth. This is clearly illustrated in SiSoft Sandra's system RAM benchmark where a KT133A chipset with PC133 RAM is very close in performance to a DDR chipset. The benchmark suggests that DDR is approximately 40% efficient and SDR is 60% efficent. I believe this is a cause of latency and unequal timings due to the doubling nature of DDR. What I mean by that is that the Motherboard/Graphics Board is no longer synchronized with the RAM and monitoring Asynchronous RAM for feedback adds a toll to latency.
Now I'm not a RAM expert but from what I've read, I believe what I'm saying sounds logical. Again, I stress I'm not trying to be biased but how else can a MX tie a GTS? BTW, I have tried several different tests on both MX and GTS boards and for all but the most demanding tests, the number of rendering pipelines and core speed have almost no effect on overall performance. So in that respect, I agree RAM performance is vital however, DDR is not the ultimate solution as I said before. If I were offered a 166 MHz DDR (333MHz effective) board and a 333 MHz SDR GTS (if it existed), I would definitely choose the latter as it's performance would probably be near or even at the GeForce2 Ultra level.
If you're interested, here are my system specs:
AMD Athlon 1.2 GHz (AXIA) @ 1.4 GHz (30c idle, 40c load)
GlobalWin WBK-38
1 5000 RPM Exhaust fan + 1 5000 RPM Slot fan beside Graphics card
Abit KT7A RAID (140MHz)
256 MB Micron PC133 CL2
40 GB Western Digital 7200 RPM Hard Drive
Asus v7100 Pro @ 260/260 <- the ultimate MX board (but I'm probably getting a GeForce3 Ultra or Radeon2 or Kyro III when they come out)
ATI TV-Wonder
12x Sony DVD ROM
8x4x32x HP CD-Writer
Network Card
19" NEC Monitor