More Speed = More Performance? Fast DDR Memory from TwinMOS and Buffalo Technology Put to the Test

Relationships And Ratios

Now all these subsystems running at different speeds have to do is work together. When we used the 333 MHz FSB clock speed for the Athlon XP 3000+ with DDR333 memory in the test, we discovered that data can be exchanged one-to-one between the processor and RAM every clock cycle. If you use DDR266 or DDR400, however, the FSB and main memory no longer run synchronously. So combining fast FSB with slower RAM, or vice versa, will increase the amount of management to be done and create wait states. We call this relationship the FSB:RAM ratio. Recent experience has shown time and again that this ratio is a big factor in overall performance, since only rarely does a faster clocked RAM improve performance.

The Test System: Athlon XP With nForce2 Chipset

NVIDIA's nForce2 chipset is an excellent foundaiton for the test runs because it offers a large spectrum of FSB:RAM ratios:

Swipe to scroll horizontally
Clock RatioMemory Clock (based on FSB333/ 166 MHz)
2:183 MHz (DDR166)
5:3100 MHz (DDR200)
3:2111 MHz (DDR222)
4:3125 MHz (DDR250)
5:4133 MHz (DDR266) 1
6:5139 MHz (DDR278)
1:1166 MHz (DDR333) 1
5:6200 MHz (DDR400) 1
4:5208 MHz (DDR416)
3:4222 MHz (DDR444) 1
2:3250 MHz (DDR500) ²
3:5278 MHz (DDR556) ²
1:2334 MHz (DDR668) ²

1: Used in this test

2: Failed to work at all

All clock speeds listed in this table are the result of an underlying FSB clock speed of 333 MHz. If you want to determine possible clock speeds with FSB266, simply calculate the ratio in the left-hand column.

Sadly, owing to the clock speed ratios supplied by the nForce2 chipset, the 466 MHz speed couldn't be set. As the table shows, we had to stick within the bounds of 444 or 500 MHz. The latter, unfortunately, no longer operated stably.