Benchmarks of 733 MHz and 800 MHz Intel PIII CPUs (using an Intel 815E motherboard) show that there is roughly a 33 MFlops increase in performance associated with a 66 MHz increase in CPU clock speed (for large array sizes), or equivalently 0.5 MFlops/MHz. We assume that this holds as well for the AMD Athlons (without doing the explicit benchmarks).
| speed(MHz) | Cost($) | MHz/$ | MFlops | MFlops/$ |
|---|---|---|---|---|
| 700 | 216 | 3.24 | 140* | 0.648* |
| 750 | 247 | 3.04 | 165* | 0.668* |
| 800 | 281 | 2.85 | 190 | 0.676 |
| 850 | 372 | 2.28 | 215* | 0.578* |
| speed(MHz) | Cost($) | MHz/$ | MFlops | MFlops/$ |
|---|---|---|---|---|
| 700 | 144 | 4.86 | 50* | 0.347* |
| 750 | 164 | 4.57 | 75* | 0.457* |
| 800 | 187 | 4.28 | 100* | 0.535* |
| 850 | 220 | 3.86 | 125* | 0.568* |
| 900 | 279 | 3.23 | 150 | 0.538 |
| 950 | 381 | 2.49 | 175* | 0.459* |
| 1000 | 460 | 2.17 | 200* | 0.435* |
*= Extrapolated from measured data using the 0.5 MFlops/MHz "rule".