Accelerator hardware theory

[Gulliver]

I have an A600 and a spare 68030 micro, so i though i could figure if it is feasable building an accelerator card for it.
Please do correct my assumptions and clarify my doubts if possible!

Q: How do i stop the 68000 from working?
A: Using the HLT pin i can let the other micro get control

Q: How do i synchronize the accelerator micro at eg 25mhz with the 7 mhz A600 bus? Can i overclock safely the A600 bus?
A:

Q: 68000 is 16 bit externally, 68030 is 32 bits. How do i join data signals?
A: I guess i shoul join the 16 bit 68000 data to the LSB of the 68030. Dont know!

Anyway its a hobby project, and to keep it small and simple, and cheap, i wont put ram or a rtc on it.

[trekiej]

I wonder if wait states need to be included.

[Gulliver]

I thought abou waitstates but i found several designs claimed to have Zero wait states! How? I dont know!

So, ATM the only solution regarding bus synchronization would be to acces the A600 bus according to its 7mhz clock rate thus providing a great performance penalty not to use the accelerators speed.

[Matt_H]

Not including any fast RAM is also going to be a big performance
penalty. Either way, this is a massive project.

Something I thought about a few years ago was desoldering the
A600's 68000, adding a socket, and dropping in a pin-compatible 68010.
Negligible speed improvement, but it would have a VBR which is helpful
for WHDLoad.

I'm hoping Jens Schoenfeld brings his A600 accelerator into production
soon.

[Gulliver]

I am also waitin for Schoenfeld accelerator, but anyway i would like to know how to build one.
Of course an accelerator without mem is a huge penalty! But i wanted to do one thing at a time, first learn how to properly accelerate an Amiga and then add all other beauties such as Fastram, RTC, Local bus expansion etc.
It is easier doing it little by little, than attempting to do it in one shot.

Yep 68010 is better than 68000, but 68030 is far better!

[billt]

For ideas of what needs done, read up on teh Frances project.

ftp.funet.fi  /pub/amiga/hardware/accelerators  

[Gulliver]

Thanks for pointing me that out billt!
Any suggestions, ideas, criticism?

From an electrical engineering point of view, it should be relatively easy to adapt Frances to run with a 68030 instead of a 68020.

[Oli_hd]

Hi

Q: How do i stop the 68000 from working?  
A: Using the HLT pin i can let the other micro get control

Nope, A: request the bus (BR) wait till its granted (BG)
and send a BGACK, the 68000 doesnt do anything and you have the bus.
The Haynie archive has the equations for the fastslot found on the A3/4000 which is:
CBR = ’b’1;
BOSS = BG30 # BOSS & !poweron_reset;
BG = local_card_bg;
BG.oe = BOSS;
BOSS is a private BGAck, CBR is just BR.

Q: How do i synchronize the accelerator micro at eg 25mhz with the 7 mhz A600 bus? Can i overclock safely the A600 bus?  A:

The bus remains unchanced, the main difference is you have to clock the interupts and control signals in. If you have memory or other stuff on board you have to have bus buffers between the card and the Amiga which are turned off whenthe board is talking to something on the same board.

Q: 68000 is 16 bit externally, 68030 is 32 bits. How do i join data signals?  A: I guess i shoul join the 16 bit 68000 data to the LSB of the 68030. Dont know!

Have a look at the A2630 user manual or service manual, both have the schematics for the card, an A500/A600 card would be the same! (The PAL code for the card is available on the Haynie archive)

[Doobrey]

Gulliver wrote:
I thought abou waitstates but i found several designs claimed to have Zero wait states! How? I dont know!

 They're talking about read/write access cycles to fast mem not needing wait states.
 Any read/writes to chip mem or motherboard chip registers will still be controlled by the custom chips, so you're stuck with their timing as to when the cycles are terminated.

You might wanna think about making the CPU clock a multiple of the A600 bus to make syncing easier.
IIRC,you're not gonna be able to make use of the 030's cache burst , cos it needs a 32bit data bus to be able to do it.

BTW  if haven't already got it, get the  MC68030 User Manual from Freescale and read chapter 7 (Bus operation). The bit about Dynamic Bus Sizing answers your earlier question about hooking up the 030 to a 16 bit data bus.
 There's also a section on howto use a 030 in a 020 design (such as the Frances board others have suggested)

[Gulliver]

Thank you guys, I will defiately take your recomendations, those were the kind of answers i was hoping to get.
Big thanks! :-)

[alexh]

On the net there is also the PAK68k which was a DIY 030 (?) accelerator for the Atari which went into production (and also worked with the Amiga).

[billt]

Any suggestions, ideas, criticism?

I don't know much about Frances, never had one. From memory, which could be rusty since I hadn't seen anything about Frances for at least a decade now, thre was a separate memory card module for it, and they did something weird with a couple address bits that prevented some things from working normally, maybe Mac emulation had problems with it. But it's a start for what you're interested in, and I'd recommend googling for more info about it. There may be some stuff for it on aminet as well, but I just grabbed the first link that came up for you. I think I came across a bare PCB for it in a pile of stuff I got somewhere, the CPU module only I think, but don't know where it might be now. It was the long skinny 68000 style though, not the squar one I think the a600 has.

[Olecranon]

On the subject of the Schoenfeld A600 accelerator.  My question is why bother?  Why not do an A500 accelerator instead?  First, you have the numbers.  There must be tens of thousands more A500's in the world than A600's.  Second, you have more room inside the case and you dont have to build something that is going to pop off the top of the chip everytime the system is moved.

I would think a nice 020/030 accelerator with simm module socket and simple IDE controller would sell pretty well I would think.  A500's are cheap to buy.

[alexh]

Olecranon wrote:
On the subject of the Schoenfeld A600 accelerator.  My question is why bother?  Why not do an A500 accelerator instead?

The size of the A600 is it's best feature. It is small, easily tidied away, wife "acceptable".

Couple that with it's built-in IDE interface (for WHDLoad), PCMCIA interface (for networking) and video modulator for direct hookup to the TV makes the A600 better than an A500 hands down.

you dont have to build something that is going to pop off the top of the chip everytime the system is moved.

The Schoenfeld A600 accelerator screws in place, no "popping off" when moved.

I would think a nice 020/030 accelerator with simm module socket and simple IDE controller would sell pretty well I would think.

If you made one, you'd never make your money back.

A500's are cheap to buy.

A600's are cheap to buy.