Toshiba Gary

[mark_k]

Apparently there are some software-visible/measurable differences between the original Toshiba-made Gary chip and the later MOS one. This newsgroup posting by Dave Haynie from 1993 says:

That's the Toshiba version.  The MOS version speeds up access to ROM and the
"I/O" space of Zorro II (eg, it fixes a bug in the original Toshiba-foundried
design).  Along with the Gary change, you'll probably need to add a couple of
pullup resistors to the A2000 motherboard for optimal performance.  When they
did the MOS version, they changed a few active outputs to open-drain, so speed
critical outputs (pretty much just VPA* and DTACK* as I recall, but it's been
years) need more pull to +5V with the MOS part, since the chip doesn't provide
it.

Does the Toshiba Gary treat accesses to Zorro II I/O space ($Exxxxx) and ROM ($Fxxxxx) the same as Chip RAM, in other words like slow RAM? Or is access slowed but differently or not as much?

CrossMAC comes with a GaryTest program which can detect the difference. Reading/writing Mac 800K disks via A-Max II+/IV is very timing-sensitive and I guess it can be adversely affected with Toshiba-made Gary.

GaryTest needs an A-Max II+ or A-Max IV board. But for real-hardware testing I could provide a patched version which doesn't, using either Kickstart ROM or a fixed address in Zorro I/O space instead. (It times a MOVEP loop reading a location in chip RAM, first with all DMA disabled then with DMA enabled and a blit running. Then it does the same but with the MOVEP loop reading an address in A-Max board space. With Toshiba Gary the active blit must cause contention slowing the Zorro I/O space accesses.)

Worth adding another option to advanced chipset settings for that?

[Toni Wilen]

This requires either having the chip and checking cycle access timing or someone else (who knows what he is doing) confirming it. Software results are useless.

It does sound like it does not consider ROM and Z2 IO as "fast ram"-like.

[mark_k]

platatomi posted about his rev. 4 A500 recently, that has a Toshiba Gary.

[mark_k]

Another possible thing to check with Toshiba Gary chip, would be the both-CIAs-selected-at-same-time behaviour. Same as MOS Gary or different?

There are two Toshiba-Gary machine variants: original (includes slower Zorro I/O and ROM access) or with the ROM access fix (still with slower Zorro I/O).