What would your ideal upgrade be right now for your Amiga(s)

[Belial6]

Well, you know the other problem for high speed USB is that it is very interrupt heavy. When transferring large amounts of data from one high speed USB hard drive to another on my PC, I can track the interrupt overhead by watching /proc/interrupts. Servicing said interrupts ties up the CPU, which on linux shows up as abnormal amounts "%sy" (system) CPU.

So, even assuming we had a fast enough bus to put a high speed USB2 controller on, we'd probably find that the classic amiga hardware can't service the interrupts fast enough to get it going at full speed (high speed ethernet has the same problem too, but the packets are bigger and therefore fewer are generated per unit of data transferred).

I have thought about this, as a side note of thinking about USB Mice and Keyboards for my MiniMig.  I would think that a fairly cheap Arm processor could handle the USB stack and then interface to the Amiga in whatever protocol the Amiga could handle.  The idea being that to the Amiga it looks like a standard disk drive, and to the USB device it would look like an Arm computer.  Using this method you could use 100% of the the processor for the USB stack and interface to the Amiga, as that is the only task the Arm processor would be responsible for.  This would be useful for interfacing USB hard drives, USB keyboards, and USB mice.

[TheBilgeRat]

I have thought about this, as a side note of thinking about USB Mice and Keyboards for my MiniMig.  I would think that a fairly cheap Arm processor could handle the USB stack and then interface to the Amiga in whatever protocol the Amiga could handle.  The idea being that to the Amiga it looks like a standard disk drive, and to the USB device it would look like an Arm computer.  Using this method you could use 100% of the the processor for the USB stack and interface to the Amiga, as that is the only task the Arm processor would be responsible for.  This would be useful for interfacing USB hard drives, USB keyboards, and USB mice.

Like a modified raspberry pi device...  seems imminently doable.

[bloodline]

Like a modified raspberry pi device...  seems imminently doable.

Raspberry Pi, running AROS/UAE... That would be my next Amiga!

[Belial6]

Exactly.  The Rasbarry Pi already supports USB 2.0, so we know that it can handle that.  Add the interfaces for connection to an Amiga (or other retro system) and your ready to go.

[J-Golden]

I'd like everything you could put on a Mediator without the mediator!:afro:

[JimS]

Exactly.  The Rasbarry Pi already supports USB 2.0, so we know that it can handle that.  Add the interfaces for connection to an Amiga (or other retro system) and your ready to go.

I have been thinking about something like this... maybe using the Propeller chip as controller. The trick would be connecting to the Amiga. Do the ARM chips have the old school address & data bus interface to memory, or just SPI? I suppose you could use a sort of clockport style interface and have a service routine on the ARM do the grunt work of reading it then passing it off to or from the I/O devices.

[trip6]

Retromasters UFE SD card floppy emulator so that we can have a 1541 ultimate type device for amiga with keyboard overlay control software for ease of use. I wish this item would get produced, been drooling over it for two years now...

[nicholas]

Raspberry Pi, running AROS/UAE... That would be my next Amiga!

That would be fantastic!

[Moggen]

It's been a month since the last Emumiga update

Oh, sorry about that. I've been quite busy doing other stuff the last couple of months. (Was actually in UK for two weeks recently).
I'm trying to find the same level of inspiration I had two years ago and get somewhere with this project.

[Belial6]

I have been thinking about something like this... maybe using the Propeller chip as controller. The trick would be connecting to the Amiga. Do the ARM chips have the old school address & data bus interface to memory, or just SPI? I suppose you could use a sort of clockport style interface and have a service routine on the ARM do the grunt work of reading it then passing it off to or from the I/O devices.

That's what I was thinking.  The reason that is always given on why these retro systems cannot use USB is because the system (real retro) or the system they are recreating cannot handle the USB load, and don't have the USB software stack to handle it.

By doing all of the USB on the Arm, and just outputting in whatever format is the most convenient for the retro system, the ability of the retro system to handle a USB stack becomes irrelevant.  At the very least, it should be trivial to make a USB gamepad/USB keyboard/USB mouse to 9-pin joystick/PS/2 keyboard/PS/2 mouse adapter.  This would give all of the new retro systems the ability to use new periphrials.

Making data storage and network would be a little tougher, but shouldn't be over the top hard.

[smerf]

Hi,

I am thinking of ripping the guts out of my A4000 Desktop and putting in a Quad Core Intel board then running Cloanto's Amiga Forever on it, then my A4000 will be one of the meanest fastest Amiga's in town.

What an upgrade.


smerf

[platon42]

Well, you know the other problem for high speed USB is that it is very interrupt heavy. When transferring large amounts of data from one high speed USB hard drive to another on my PC, I can track the interrupt overhead by watching /proc/interrupts. Servicing said interrupts ties up the CPU, which on linux shows up as abnormal amounts "%sy" (system) CPU.

So, even assuming we had a fast enough bus to put a high speed USB2 controller on, we'd probably find that the classic amiga hardware can't service the interrupts fast enough to get it going at full speed (high speed ethernet has the same problem too, but the packets are bigger and therefore fewer are generated per unit of data transferred).

You recognize that the Deneb is a High Speed USB 2.0 card available for years now that can have a throughput of over 8.5 MByte/sec on an Amiga Classic system? And about the interrupts: well, it very much depends on the controller and the driver. The Deneb transfers double buffered chunks of 24 KB each per interrupt. That's a lot bigger than your average 1500 Byte Ethernet frames... Even the Subway/Highway does not cause an interrupt for every 64 bytes transferred...

[TheBilgeRat]

You recognize that the Deneb is a High Speed USB 2.0 card available for years now that can have a throughput of over 8.5 MByte/sec on an Amiga Classic system? And about the interrupts: well, it very much depends on the controller and the driver. The Deneb transfers double buffered chunks of 24 KB each per interrupt. That's a lot bigger than your average 1500 Byte Ethernet frames... Even the Subway/Highway does not cause an interrupt for every 64 bytes transferred...

Is this the reason for the quick throughput on ethernet dongles on a Deneb?

[desiv]

Is this the reason for the quick throughput on ethernet dongles on a Deneb?

Shouldn't be as most frames are smaller than 1500 bytes.
If you are transferring big files, yes, but there is a lot of chatter and small stuff traffic...

Possibly better buffering...
A more efficient driver can relay frames quicker also, and any decrease of latency is a huge plus...

desiv
(Hey, I just said "shouldn't be" and then mentioned buffering, which was mentioned above..  D'oh!  In that case, (me actually reading) yes.., but probably not the larger packets unless you are doing lots of file transfers..)

[Karlos]

You recognize that the Deneb is a High Speed USB 2.0 card available for years now that can have a throughput of over 8.5 MByte/sec on an Amiga Classic system?

Which is certainly impressive, but 8.5MB/s, whilst fast for a classic Amiga is not high speed as per USB2 standards for bulk transfers, ~30MB/s is. My reply was on the basis of achieving the latter, ie USB2 transfer rates comparable to what everybody else is used to.

And about the interrupts: well, it very much depends on the controller and the driver. The Deneb transfers double buffered chunks of 24 KB each per interrupt. That's a lot bigger than your average 1500 Byte Ethernet frames... Even the Subway/Highway does not cause an interrupt for every 64 bytes transferred...

Of course, but that's an implementation issue. There's not a lot in the standard that I'm aware of that dictates how much data a hardware device should buffer. I've seen USB2 controllers that will happily generate an interrupt every packet. Likewise, I've seen ethernet adapters that seem to implement a sensible degree of buffering and don't produce quite the same overhead - very important given the speeds they can reach.

[marcfrick2112]

Man, an RTG graphics card that I could afford for my A1200T!

Please!

Sorry, but 2MB CHIP RAM is lousy!!!