CompuServe Messages

Hal, No, swap space should have played no part at all…the 3000 UX used in the test had 9 Meg of RAM. The NeXTstation had 16 Meg of RAM. Both should have been sufficient. Neither used and external processor cache. Not sure how much internal cache the '030 has, but the '040 has an 8k (4k instruction, 4k data) internal cache. I seriously doubt that the differences between Mach and SVR4 are enought to account for the differences. However, I do *suspect* that since the NeXT is about the only machine using Mach…and there's no ABI compliance requirement…that the Mach running on the NeXT is specifically tuned to the hardware. More efficient compilers, kernel, libraries, etc. can all make a big difference. For example, the HP9000/425…also running a 25MHz 68040 (same as the NeXTStation) about doubled most of the CPU intensive benchmarks over the NeXTStation. If Commodore could make their hardware/software as efficient HP's, they may have looked significantly better. Don