alexh wrote: 22 Mar 2023 09:58
Is HW byte swap noticeably faster?
Using PP's AHPT94 program which purports to test various drive speeds, I've tested logical drive transfer rate for a PC-formatted SD card running:
1) under DFB1 with 50MHz CPU:
1a) Under EmuTOS (TTRAM), logical (software) byteswapping: 2652 kB/s
1b) Under EmuTOS (TTRAM), hardware (twisted cable) byteswapping: 3799 kB/s
1c) Under TOS404 (STRAM*)+PP's PC driver, logical byteswapping: 1926 kB/s
1d) Under TOS404 (STRAM*)+PP's Twisted PC driver, hardware byteswapping: 2509 kB/s
1b:a = 43% improvement
1d:c = 30% improvement
2) Stock Falcon
2a) Under EmuTOS (PRG), logical b/s: 1196 kB/s
2b) Under EmuTOS (PRG), hardware b/s: 2209 kB/s
2c) Under TOS404+PP's PC driver, logical b/s: 1430 kB/s
2d) Under TOS404+PP's PC driver, hardware b/a: 2592 kB/s
2b:a = 85% improvement
2d:c = 81% improvement
Consequently and unsurprisingly the increase in performance is noticeable with this distinction even more obvious as the underlying processor (doing the logical byte swap) runs more slowly.
Given the 68k is even less performant in these operations than the 030, tested here, I'd say Christian's smartswap for his ST IDE boards offers a very significant bonus. [edit: I see Cyprian has quoted the actual speed diff -- very good indeed].
All this, of course, presupposes you want to have a PC-formatted drive on your Atari IDE bus. If you don't, these solutions offer no speed benefit.
BW
* PP's free PC driver doesn't seem to like TT-RAM. All these run with ST-RAM only. The machine locks up with TT-RAM enabled trying to read from the hard disc.