Question on Amiga HD controllers

[Pat the Cat]

I know you can get IDE-SD card and various forms of compact flash adaptors for the Amiga.

I was wondering what the limitations of IDE and SCSI controllers are with the Amiga? I would guess the low end would be a raw A600 IDE interface, then a A1200 with some fast RAM, then the A4000 IDE controller, and then third party and Commodore SCSI controllers, for transfer speed.

It's a complicated issue, but I was looking at what the max read and write speeds of controllers are, with a view of what is sensible to attempt when hacking SSD drives onto them.

There is a reason for exploring this, in that Amigas with MMUs can use paged hard drive space as fast RAM - in theory, if your controller can handle the speed, you could run MMU friendly apps of huge complexity, at pretty good execution speeds.

Or maybe the controllers themselves are too limited, and a reasonably quick SD card, says 40MB/S, is plenty good enough anyway and there is no point reinventing the wheel.

On the other hand, M-SATA to ZIF PATA convertors are pretty cheap, and it should be possible to hack a ZIF PATA onto an Amiga. Which might be worth it on a real A4000, but I'm a bit doubtful about what else.

An M-Sata to SCSI 50 pin Single ended could give quite a big boost to SCSI equipped Amigas. But haven't seen any kit that does that.

[Pat the Cat]

That's better than what I expected. Paged Ram from a single ended 50 pin working around 20MB/S is still very useful for very large applications, and that's what most Amiga SCSI controllers are limited to. I have no idea on that front, but it sounds right for SCSI-2.

I guess for the real faster interfaces you would need a Zorro III based Amiga and a whizzy SCSI controller. IIRC, some of the 3/4000 accelerator cards had such a beast fitted too. Could be wrong about that, but I guess a 32 bit DMA friendly card, and later Buster and DMAC chips might let you use some of the faster types of SCSI.

A lot of modern accelerators have way faster RAM access than that... so unless you have a very big application, requiring hundreds of software components, it's just plain unnecessary.

What applications was I thinking of? Well, smart robotics, actually. Cognition research would be closer.