DDR vs RDRAM Mathematics

The following is my mathematic analysis of why I went DDR:
---------------------------------------------------------------------------
1st- the Rambus numbers:
PC800 bandwidth/channel=1.6Gigs/second.
However, many primo modules will reach PC1066 and yield:
PC1066 bandwidth/channel=2.1Gigs/second.
Total for 2 channels PC1066 Rambus= 4.2Gigs/second.
---------------------------------------------------------------------------
2nd- the DDR numbers:
DDR@166Mhz bandwidth/channel=2.7Gigs/second.
However, many primo modules will reach 184Mhz and yield:
DDR@184Mhz bandwidth/channel=3.0Gigs/second.
When Intel's "Granite Bay" 2 channel DDR chipset is released (Q3/2002):
Total for 2 channels 184Mhz DDR= 6.0Gigs/second!!!
----------------------------------------------------------------------------
My KingmaxPC2700 will run at 200Mhz. Even if PC1200 RDRAM comes
along it still will only yield 4.8Gigs/second for 2 channels. And the RDRAM
RIM4200 modules soon to be released are just (2)PC1066s mechanically
connected into a single module, no different from the present situation.
Granted,for the moment, hand-picked RDRAM modules at PC1066 certainly
are the bandwidth king. But when 2 channel DDR motherboards are common
later this year, won't DDR be fastest at 6 Gigs/second?

thanks for the info interesting read

how are you able to get your chip up to 2.8, stable? i have ddr, but when i get the new p4, i will get a rambus board, because the ddr cant keep up with the chip.

My ABIT SD7-533 has a 4-5-2-1 ratio that runs the following:
140fsb/175DDR/70agp/35pci. My 2.0a CPU runs at [email protected], [email protected], or [email protected].(I can boot [email protected], but not stable). My KingmaxPC2700: [email protected]. The SD7 has a 3-6-2-1 ratio that I can run the DDR at 200Mhz and keep all else in spec, but not overclocked...Mark

Just for a comparision, here is Rambus 1066 v DDR400. Rambus takes it quite easily, and as rambus latency decreases the higher it runs, and DDR latency increases when you run it asyncronously, then Rambus is better.

Hey, Anonicle! I benchmarked my KingmaxPC2700 DDR memory using Sandra2002 and got the following#s: 2569/2566@175Mhz(DDR350), and 2471/2473@200Mhz(DDR400). Why were my #s lower at the higher Mhz? In the 1st case the CPU was at 2.8Ghz. In the 2nd cast it was at 2.0Ghz (due to divider limitations). Also why were your DDR400 #s so much better than mine? ...Mark

Maybe running the ram asynconously affects the performance of it. Your also using the SIS chipset, i am using the i845. Still, running yours at DDR400 i would of expected over 3000/3000. What timings do you use?

Anonicle, my timings are 2.5-3-3,but since they remained constant over the adjustments I discount that as an issue. I looked at the Aceshardware article where they kept the memory speed constant,increased fsb,then re-benched the memory scores. They took a 100Mhz fsb with PC800 rambus(perfectly matched with both at 3.2Gigs/sec theoretically)and then raised the fsb while keeping the ram at PC800. A 33% increase in fsb frequency resulted in an 8% memory score increase(for an efficiency of about 25%). My guess on the math of my situation is as follows: Startpoint:2471 sandra2002 bench with a 100Mhz fsb/200Mhz DDR for a theoretically exact bus speed match. Next: ram frequency decreased to 175Mhz: 175/200= -12.5% benchmark decrease(linearity assumed). Next: fsb increased from 100Mhz to 140Mhz= +40% increase. But while an fsb increase at Aceshardware yielded a 25% efficiency in improving the memory benches, my chipset is different. An efficiency of exactly 47% calculates to raise the benchmark to what I got: 2569. All of this translates into two conclusions: #1 that raising the ram frequency linearly raises memory benchmarks, #2 that raising the fsb from an equalized bandwidth situation only raises memory benches half as much at best. Still, it shows me that the best way to get system speed is to keep both ram and fsb as high as simutaneously possible, which is back to the beginning of my adjustments...Mark

Why do I get only 1900 scores for my memory with Crucial2100 DDR? I have a P4 Northwood 2.0 at default settings and cas timings at 2, also I have a Soyo P41 Fire Dragon mobo.