For DRAM ratios and whatnot - there is another thread on that... look for y2viks.
As to the effect of FSB on Athlon XPs, here's my take.
Using an XP2400 on an NF7 v2, with timings at 2-2-2-11 on WinXP; benched with Aquamark, Sandra 2k4 and Pifast. The unbuffered memory bench in Sandra can be activated when looking at the 'options' tick-box screen (the icon on the bottom left when bringing the benchmark screen up).
The Athlon XP was benched at 133x15, 166x12 and 200x10 to investigate the effects of increased CPU and system bandwidth on everyday performance.
Aquamark first.
It would appear that - despite the silly scale on the graph - front side bus rate does have an effect on the CPU's athleticism... albeit rather small.
PiFast next...
PiFast appears to rely on system bandwidth more heavily than Aquamark (perhaps Aquamark is more GFX dependent?) with reductions of over five seconds just by increasing the front side bus - remember, the CPU clock rate has not changed (ok, fine - it's 1992MHz or whatever... the BIOS picks it out as 2GHz).
Let's see what Sandra has to say for herself...
Oooh... do I detect a decrease in useable memory bandwidth? As can be seen, decreasing the FSB of the CPU reduces the amount of data it gets to allocate to the RAM meaning the CPU is sitting, twiddling its thumbs waiting for the RAM to catch up.
Despite all the CPU clocks being (effectively) the same, there is obviously extra peformance to be had by increasing the front side bus of the processor. This is especially pertinent for those of you playing with locked multipliers - you see? All is not lost! Concerning the PiFast results, too - although there is only a very small drop in Aquamark performance with decreasing FSB, PiFast is raw calculation, meaning that DivX-ing (for whatever reason) and mp3-ing will be faster the more FSB (memory and system bandwidth) you have.
The moral of the story? Reduce that multiplier and ramp up the FSB for a better wired life.
As to the effect of FSB on Athlon XPs, here's my take.
Using an XP2400 on an NF7 v2, with timings at 2-2-2-11 on WinXP; benched with Aquamark, Sandra 2k4 and Pifast. The unbuffered memory bench in Sandra can be activated when looking at the 'options' tick-box screen (the icon on the bottom left when bringing the benchmark screen up).
The Athlon XP was benched at 133x15, 166x12 and 200x10 to investigate the effects of increased CPU and system bandwidth on everyday performance.
Aquamark first.
It would appear that - despite the silly scale on the graph - front side bus rate does have an effect on the CPU's athleticism... albeit rather small.
PiFast next...
PiFast appears to rely on system bandwidth more heavily than Aquamark (perhaps Aquamark is more GFX dependent?) with reductions of over five seconds just by increasing the front side bus - remember, the CPU clock rate has not changed (ok, fine - it's 1992MHz or whatever... the BIOS picks it out as 2GHz).
Let's see what Sandra has to say for herself...
Oooh... do I detect a decrease in useable memory bandwidth? As can be seen, decreasing the FSB of the CPU reduces the amount of data it gets to allocate to the RAM meaning the CPU is sitting, twiddling its thumbs waiting for the RAM to catch up.
Despite all the CPU clocks being (effectively) the same, there is obviously extra peformance to be had by increasing the front side bus of the processor. This is especially pertinent for those of you playing with locked multipliers - you see? All is not lost! Concerning the PiFast results, too - although there is only a very small drop in Aquamark performance with decreasing FSB, PiFast is raw calculation, meaning that DivX-ing (for whatever reason) and mp3-ing will be faster the more FSB (memory and system bandwidth) you have.
The moral of the story? Reduce that multiplier and ramp up the FSB for a better wired life.