Access speed for EPROM on expansion card

[mark_k]

On many expansion cards, the EPROM(s) have an access time of 200 or 250ns according to text printed on the chips.

For fast RAM (DRAM), 150ns or faster chips are needed for no-wait-state operation. Does the same apply to expansion card ROMs? Or, since they work more like SRAM instead of DRAM, is a 200ns (say) EPROM access time not directly equivalent to DRAM with the same access time?

[voxel]

expansion cards roms are read once at boot and their content shadowed in memory, so no need to use lightning fast roms ^^)

(if I remember well... ;-) ) ^^)

[mark_k]

That's only true for cards with byte-wide ROM. On cards with word-wide ROM (typically indicated by there being two physical EPROM chips) the driver can run directly from ROM.

Some cards with word-wide ROM (just the first few from amiga.resource.cx): Alcomp SCSI card, Combitec AutoBoot-Karte, Commodore A2090, Commodore A2091, Elaborate Bytes/bsc A.L.F. 2, Gigatron Arriba HD, GVP Series I cards, HK Computer Vector Falcon 8000, ...

SetCPU's CARDROM option can be used to remap word-wide board ROM code to 32-bit RAM on accelerated machines.

It might be possible to use memspeed to see if there's any difference between reading Zorro II fast RAM and a card's ROM. Anyone want to try?

The reason I'm curious is, I've created a word-wide (dual-ROM) version of gvpscsi.device for GVP Series I cards which runs the driver directly from ROM. I don't have real hardware to check whether that would run any slower than the original driver which copies itself to RAM though.

[pandy71]

Memory cycle in Amiga is approx 280ns - add (i.e. subtract) to this address decode and buffer delay (something around 20 - 35ns) i.e. anything faster than 250ns should be sufficient with DTACK grounded (0 WS).

[hooverphonique]

also, the logic on the board is designed for some specific rom access time. just changing the roms will not make the logic access it any faster..