bytemark score relative power per Ghz Manufacturer model CPU clk bus clk L1cache L2cache integer fp integer fp integer fp Intel 486DX2 66 33 8K 256K 0.428 0.255 1.000 1.000 15.15 15.15 IBM 486BLDX3 75 25 16K 0.461 0.267 1.077 1.047 14.36 13.96 Cyrix 486DX4 100 33 8K 256K 0.599 0.336 1.400 1.318 14.00 13.18 AMD 5x86 133 33 16K 256K 0.856 0.494 2.000 1.937 15.04 14.57 Cyrix 5x86 120 60 16K 256K 0.962 0.675 2.248 2.647 18.73 22.06 Intel Pentium 90 60 16K 1.000 1.000 2.336 3.922 25.96 43.57 Cyrix MediaGX-S 180 60 16K 1.180 0.772 2.757 3.027 15.32 16.82 IDT Winchip 200 66 64K 512K 1.562 0.877 3.650 3.439 18.25 17.20 Intel PentiumMMX 166 66 32K 256K 1.910 2.110 4.463 8.275 26.88 49.85 Cyrix 6x86L 150 75 16K 512K 2.007 0.936 4.689 3.671 31.26 24.47 Cyrix/IBM 6x86MX 225 75 64K 512K 3.060 1.690 7.150 6.627 31.78 29.46 Intel Pentium-M 300 50 64K 2048K 4.818 5.505 11.257 21.588 37.52 71.96 AMD K6-2 380 95 64K 512K 5.460 3.910 12.757 15.333 33.57 40.35 Intel Celeron 400 66 32K 128K 5.770 6.560 13.481 25.725 33.70 64.31 Bytemark v2 crashes on a P3-600 (Pentium-M was underclocked for the test) Motorola 68020/882 14/20 14 256B 0.036 0.011 0.084 0.043 6.00 3.07 Motorola 68030/882 24/24 24 512B 0.058 0.018 0.136 0.071 5.67 2.96 Motorola 68040 40 40 8K 0.229 0.036 0.535 0.141 13.38 3.53 Motorola 68060 50 50 16K 0.588 0.281 1.374 1.102 27.48 22.04 Motorola PowerPC 604e 180 60 16K 2.709 1.296 6.329 5.082 35.16 28.23 My experience would suggest that the 020, 030, and 040 CPUs (I've never used an 060) are better than what these results indicate when compared to an 80x86. Also, the 020 and 030 probably both had 100ns DRAM which would be more of a bottleneck in the case of the higher clocked 030 (and more relevant to these CPUs since the cache is small).