First post, by noshutdown
board in question is a pcchips board with unknown model and ali m1429 chipset, its identical to this abit 386at3:
https://stason.org/TULARC/pc/motherboards/A/A … 86-386-AT3.html
as the m1429 is a 486era chipset and the board supports up to 192mb of ram(even though only 32mb would be cacheable), i had high expectation on it, and it did perform superbly without cache: cachechk reports ram speed at 53us/kb compared to most typical boards at 67us/kb, thats 2 clocks faster per read, and its also noticeable faster than other boards in 3dbench, pcpbench, norton si and doom.
when cache is enabled, cachechk and speedsys results are still good: cachechk reports 40us/kb for cache which is typical for 386dx-40 boards, and cache-miss dram acces is 67us/kb, which indicates there is a 2-clock penalty over the uncached dram, which is also typical, and its still faster than my lenovo um82c491 board which is 80us/kb for cache-miss dram. speedsys reports 31mb/s for cache which is typical, and 19mb/s for dram which is also quite fast. however it never did well in other benchmarks:
norton sysinfo at 42.6pts
3dbench 14.4fps
pcpbench 3.6fps
doom 7.7fps
these scores are achieved with write-back mode, which would only cache for 16mb of ram, write-through mode would slow things down further by about 5%.
since this board performed so fast in cachechk and speedsys, there must be something wrong with it to be so slow in other benchmarks.