Possible NEW USB solutions for the Amiga

[platon42]

That is a surprise to hear. My Thylacine card seems very stable and works with printers, mice, keyboards and thumb drives. In fact the Thylacine.device lists Chris Hodges as having put the final fixes to it in 2004.  I've read where a lot of people used it with Epson scanners and loved it ( my Epson scanner is still packed from a recent move).  Could it be a problem with USB 2.0 devices?  The USB hard drives and 100 MBit/s NIC's might be slowed, but do you think there is data corruption with them?  I've only had positive experiences, what details have you had that show it is "rubbish?"  Could you list the references to those comments to pin down the problems?

http://www.cypress.com/?docID=26413

Carefully look at Errata number 1 (I've got no idea if the SMD hub chip crasbe is going to glue onto the board his this "problem", but if it has, then no mice or keyboards will be working) and number 6 (not that the other ones aren't bad, but the last one would make me refrain from using any MSD devices with it!).

[platon42]

I would guess that wouldn't pass initial tests, so he wouldn't use that chipset.  Something to be aware of tho....

AFAIK he has not planned to do too many iterations.

It sounds like that's just something to be aware of for the USB stack (don't use "auto-increment").  I don't see how this affects me at all..
The performance of USB (especially over clockport) is probably going to be throttled by many other things before this would cause an issue..
And remember, this is supposed to be low cost, so slower isn't a big deal.  People who need more speed can buy the more expensive faster cards.  ;-)

This is not a clockport card, but a Zorro card. 95% of the CPU time is spent in the copyloop for copying data to and from the USB chip for any card (except for the Deneb in DMA mode, where, hence, the Buster will do the copying) -- at least with bulk transfers involved. The current driver implementation of the Thylacine uses the auto-increment mode. The workaround (in the HCI driver, not the USB stack BTW) as stated in the errata would AT LEAST halve the speed of the data transfer on an already slow 8 bit interface. In my eyes it would render the mass storage use case useless.

So, I don't see any major issues holding this back..

Sometimes I wish the Amiga users wouldn't be wearing those reality-warping glasses.

Looking forward to a possible clockport version.

desiv

The very first Subway prototype used the SL811HS chipset. After experimenting with it in March 2001, it was found completely inadequate and unreliable and was abandoned in favour of the UHC124 chipset (both used on the production version of the Subway and Highway).