Something seems amiss with those benchmarks.
The STE ET one shows AES 1.4 and the H4 ET one show AES 3.2. Might be co-incidence but it's all the drawing routines that are slow.
As for the cache, I'm not sure what the 536 does, but in theory there's no reason cache can't work on the ST-RAM too.
Here's an example of my card working normally (a word or by read retrieves a longword every time and caches it all):-
- stram_normal.png (52.2 KiB) Viewed 3602 times
And here with CIIN tied to GND -- presumably the same as your jumper (a word or byte read only uses one cycle):-
- stram_ciin_low.png (38.62 KiB) Viewed 3602 times
It could be if you've got some DTACK confusion, slowing down AS/DTACK by disabling cache (it'll only make one bus cycle if it only needs a word or byte rather than retrieve a longword every time and cache it) is enough to tip the balance?
BW
DFB1 Open source 50MHz 030 and TT-RAM accelerator for the Falcon
DSTB1 Open source 16Mhz 68k and AltRAM accelerator for the ST
Smalliermouse ST-optimised USB mouse adapter based on SmallyMouse2
FrontBench The Frontier: Elite 2 intro as a benchmark