The review is very useful because it shows the performance(slight) difference between different RAM timings.

However, I think you guys should've added Corsair TWINX modules too.
They're recomended settings are 2-2-2-6-1
I have 2 x 512MB TWINX modules running @ 203MHz(2-2-2-6-1) on an Asus A7N8X with an o/c AMD 2600+

I found the review informative. I was considering to go with 200 MHz fsb. After reading the article I decided to wait a cople of months. On the other hand, there is a misinterpreted result on page 3. The articel compares the memory bandwith of various memory sub systems. It compares the performance growth in relative percentage. The increase from 266 to 333 Mhz and from 333 to 400 Mhz is linear thus the results should be measured in linear scale not relative percentage. Page 3 shows that going up to 400 Mhz from 333 gives more bandwith than going 266 to 333 in terms of increased bandwith. The article implies somehow the opposite though the numbers are evident. Thanks,

Interesting results.

The only thing you didn't investigate was how the 333MHz - 400MHz bus performance varied with relation to overall processor speeds. I believe that as the CPU gets faster, the performance difference between the 2 bus speeds should increase, but how much is the question.

Well I dont know about the boards you´v tested but I´m runing my Epox 8RDA 215Mhz fsb in syncronous mode with 2 samsung 256 pc2700 memory in dual mode 3-3-3-2.5 timings.

All good, thanks for the article

Somewhere in my mind I already knew this all would happen with AMD. Not only did their products on their two latest roadmaps got delayed but it also was immediately visible that AMD wouldn't keep up with the '400MHz and beyond' products.

I'll be fasinated how AMD will succeed in regaining their reputation and money they lost because of dissapearing in Intel's dustcloud.
With the introduction of the 64-bit processors by AMD it looked like AMD would regain everything within a short period but with Intel's 800MHz monster and 600MHz memory within two years the attention of people is again sliding towards their competitor.

After reading this article I agree that 400MHz memory is not a product anymore for tweakers and oc'er but building a fast 400MHz machine has become a new challenge for tweakers.

Surely the mobo manufacturers are the key players in this. Will they really want to put out a whole new set of boards, when power users will all be chasing Opteron based systems? The middle of the road crowd is still happy with their 266 or 333, and once word gets round that 400 is sometimes slower than 333, I can't see much demand for the public. Also, AMD won't have the money or inclination to push Athlon, the Opteron PR machine will need a lot of money and effort to beat Intel FUD.

Will Smith

When is 400 sometimes slower than 333? Provided the fsb and memory are running at 400 and same timings are used? I've never seen that. Granted, when the fsb and memory are not syncronized you get into latency issues and there is really no perfomance boost when running memory at 400 and fsb at 333(or 266 for that matter.) And of course, sometimes bandwidth isn't a big issue for a particular app, and timings are more important. In those cases you see cas2 333 ram beating cas2.5 or cas3 400, is that what you are referring to?

hello there i have a athlonxp +1800 with a via kt400 mb (GA-7VA) from gigabyte technoloies
i would like to know whether this mb supports fsb 400 barton core processors
pls could so you send me a reply as soon as possible

Do you know of a formula to estimate DDR memory access bandwidth performance?

Thank you very much.