Smaller, energy efficient turbo card for A1200

[shoggoth]

Problem is digging up rev. 6 060:s. Czuba sort of vacuum cleaned the world for them.

-- Peter

[keropi]

amigakit.com has new rev 060's...

[DamageX]

What about the external BUS etc. hardware that you cannot emulate?

Before you know it you'd have almost created a 68000 CPU with glue logic just to get the square peg in the round hole!

I don't see where you get this idea that it will take "almost a 68000." Just look at any A500/A600/A1000/A2000 accelerator and you'll see that the 68000 state machine is at most a handful of oldschool programmable logic and TTL chips.

A bus is a bus. Even a modern x86 CPU with hundreds of pins can still be interfaced with the slow 8-bit wide EEPROM that holds the BIOS.

[alexh]

keropi wrote:
amigakit.com has new rev 060's...

Not officially.

The price for Rev6 68060's on ebay from my seller has almost doubled in the last 2 years. From 60 euro to over 90 euro.

[Oli_hd]

've been pondering whether it would be possible to create i.e. a 030/040 compatible but more compact and energy efficient turbo card for the A1200 that would fit nicely inside the standard desktop case?

If you did a board that ran totally at 3.3V then it would use less power. Freescale make a few 3V 68K's... I havent looked but as the 68060's core is 3.3V I bet the IO bank could be run at 3.3V too.
You would just need some buffers that can drive a TTL signal and accept 5V logic levels.
As for sticking an FPGA on board for SCSI or SD, yup it could be done, opencores.org have a few designs as do Xilinx and lattice. (Heck a PIC can interface to an SD card)

As for making a more compact card... well I dunno, unless you can get some of the old 208 pin QFP 68060's that are out of production now the 060 is always going to stand tall.
The rest of the logic would be surface mount and chips in the TSSOP and TQFP packages stand only a few mm off the board. Space wise adding connectors is what takes up the space, mount the memory on to the PCB (8 chips normally) and you would save yourself a bit of space.
You could go mad and do a 8 or 10 layer PCB and make a card that looks more like the Amijoe was, a card full of chips with the tracks routed on the layers below but then your talking about a lot of expense.
I would say KISS, 3.3V the CPU board, mount some SD-Ram on the board itself (available from Digikey) and if you want SD card readers or SCSI why not add a USB host chip and use that? I think the overall cost of such a board would be low, except the 060, but you could sell it with a socket so people could hunt around, ripping Atari's to shreds for the 060 ;-)

[keropi]

well, I got a rev6 060 from amigakit.com and is currently sitting on my 2nd csppc that has socketed oscillators.

[alexh]

Oli_hd wrote:
I would say KISS, 3.3V the CPU board, mount some SD-Ram on the board itself

You'd need an FPGA/CPLD for your 680x0 bus compatible SD-RAM controller. As there are non in existance (except the CT060) you'd have to design one or ask Rodolphe of Czuba tech if you can license his HDL source code.

Oli_hd wrote:
I think the overall cost of such a board would be low, except the 060

It is still too complicated for anyone not in the electronics industry. You yourself tried to do a Coldfire board several times and was never totally successful.
 
Even for someone who is in the electronics industry, regularly makes multilayer PCB's, can get good prices and understands the design constraints, producing volumes of less than 100 would still not be economically viable.

It is unfortunately a non starter :-(

[Oli_hd]

You'd need an FPGA/CPLD for your 680x0 bus compatible SD-RAM controller.

Actually there is, I dont have the spec here at work but give me a couple of hours and I will be home.
That said a CPLD one would be easier, Xilinx give away the VHDL code for use on Xilinx chips. (They have DDR ones too but on a 50Mhz bus thats pointless)

It is still too complicated for anyone not in the electronics industry. You yourself tried to do a Coldfire board several times and was never totally successful.

Well you could use the Haynie archive PAL code from the 3640, stick it in a CPLD (using Abel) and then alter it.
The full design is available, even though the 3640 wasnt the most pimped up CPU card ever made.

producing volumes of less than 100 would still not be economically viable.

Well I ordered a slab 14.5" by 10" yesterday, two layer (yup, not suitable for high speed CPU designs, 4 layers would be needed but..) with solder mask and silk screen, inc postage and all fee's for £70 all in. fit as many designs/boards on as you can.
It aint an expensive thing to try if someone wanted.
plus you can still buy PLCC CPLD's so you dont have to play with tiny SMD things.

[rockape]

Hi Oli,

Just make us a good bit of kit, I`d buy a couple

Regards, Michael

aka rockape

[bloodline]

DamageX wrote:

What about the external BUS etc. hardware that you cannot emulate?

Before you know it you'd have almost created a 68000 CPU with glue logic just to get the square peg in the round hole!

I don't see where you get this idea that it will take "almost a 68000." Just look at any A500/A600/A1000/A2000 accelerator and you'll see that the 68000 state machine is at most a handful of oldschool programmable logic and TTL chips.

AMD CPU's also include an on chip memory controler so support chips could be kept to a minimum.

[alexh]

Oli_hd wrote:
Actually there is [free SDRAM controller] I dont have the spec here at work but give me a couple of hours and I will be home.

There are free opensource SDRAM controllers, but it wont have a 680x0 bus interface. It will have AHB or wishbone or something.

Xilinx give away the VHDL code for use on Xilinx chips. (They have DDR ones too but on a 50Mhz bus thats pointless)

Doing a good quality 680x0 bridge will take some time. Czuba tech is very nice and has posted several FAQ's that will help but still it will take several weeks worth work (a few hours a day) to perfect.

Well you could use the Haynie archive PAL code from the 3640, stick it in a CPLD (using Abel) and then alter it. The full design is available, even though the 3640 wasnt the most pimped up CPU card ever made.

The A3640 is a CRAP card. It doesnt have a RAM interface and it doesnt support 68040 burst mode. Couple with that PALASM isnt the easiest thing to reverse engineer.

Is the PAL code available as source? Is it commented?

[alexh]

DamageX wrote:
I don't see where you get this idea that it will take "almost a 68000." Just look at any A500/A600/A1000/A2000 accelerator and you'll see that the 68000 state machine is at most a handful of oldschool programmable logic and TTL chips.

The logic on most Amiga accelerators is JUST a very simple 68k to DRAM bridge, plus some sort of asynchronous buffer from the higher speed 680x0 bus to the 7MHz Amiga bus. The two are VERY a like (and were designed to be so).

A non 68k bus to Amiga Bus will take a lot more glue logic, speed bridges, buffers etc. No one will ever attempt this. Too much work. No reward.

DamageX wrote:
A bus is a bus.

Spoken by a true non-technical user.

DamageX wrote:
Even a modern x86 CPU with hundreds of pins can still be interfaced with the slow 8-bit wide EEPROM that holds the BIOS.

1) An IO mapped eprom is hardly the same as a multi-master tristate bus but hey...

2) The x86 Northbridge (or is that southbridge?) has like 2,000,000 gates or more to convert between buses and standards. Obviously I am not saying you would need even a fraction of those gates, but it is not going to be a sunday afternoons work.

[bloodline]

alexh wrote:

DamageX wrote:
I don't see where you get this idea that it will take "almost a 68000." Just look at any A500/A600/A1000/A2000 accelerator and you'll see that the 68000 state machine is at most a handful of oldschool programmable logic and TTL chips.

The logic on most Amiga accelerators is JUST a very simple 68k to DRAM bridge, plus some sort of asynchronous buffer from the higher speed 680x0 bus to the 7MHz Amiga bus. The two are VERY a like (and were designed to be so).

A non 68k bus to Amiga Bus will take a lot more glue logic, speed bridges, buffers etc. No one will ever attempt this. Too much work. No reward.

Yeah it's a lot of work... which is why I said that if there was some serious investment, that is the route I would take.

Right now, there is NO chance of any more hardware for classic hardware as any new hardware is going to cost more than it could ever sell.

[Snaggle]

Looking at the course this thread has taken I feel the need to add my opinion on some issues raised. Cos I'm so egotistical that I'm sure you all value that opinion highly.  ;-)

All those people saying that a new accelerator would not be commercially viable, you're damn right. No one could start a project like that expecting to make a nice profit at the end of it.
Then again, I've been scouring various Atari related sites recently and noticed a distinctly non-commercial attitude in many hardware projects.
You'd almost think some of these people were designing upgrades for the love of the platform or something.

Unlike those poor Atari savages, we've had companies like Phase 5, GVP, DCE, Elbox developing our hardware. But then, when most of those companies disappear cos there's no profit left in this market we have no homebrew hardware scene to take up the slack, and our maxed out A1200Ts are outperformed by Falcon 060s...

God knows making an accelerator card isn't remotely easy, but it's a lot easier now than it used to be, and we seem to have a lot more hardcore solder junkies around than we used to.
I'm not gonna start a campaign to get a new card out on the market or anything, but maybe the cynics could back off a little bit so the people with the know-how aren't so scared to come forward and make their contribution.
I know it's hard to stay calm when someone posts dodgy info or uninformed opinion, but this is just a discussion on a web forum. It's not life or death.

I wonder what that Rodolphe Czuba guy has been doing for the last year.
After all that hard work on the Phenix, it must've been heartbreaking to be let down by others so late in the development.
Take his 060 card, SDRAM interface, PCI bridge, and marry it to the basic 'magic-happy-chips' that every A1200 accelerator has (I don't know much about that stuff...) and you'd have an ultimate 1200 upgrade that I'd be happy with, even if it cost £200 for a board without CPU or any other socketed components.

[DamageX]

1) An IO mapped eprom is hardly the same as a multi-master tristate bus but hey...

Nobody said a ROM is the same as a bus, or that it is IO mapped, the point is that it can co-exist.

2) The x86 Northbridge (or is that southbridge?) has like 2,000,000 gates or more to convert between buses and standards. Obviously I am not saying you would need even a fraction of those gates, but it is not going to be a sunday afternoons work.

I agree that it's not likely to happen, in an afternoon or otherwise. So maybe it's a pointless argument, but I disagree with your characterization that using some other CPU is substantially more complicated than existing accelerators. Dunno about the Duron, but I've looked at the socket 7 pinout before and the majority of pins are for power, debugging, and multiprocessor systems. The stuff that is relevant to interfacing the CPU to an Amiga bus is not really exotic.

[Protek]

I consider myself a total beginner at electronics. I've, however managed to get myself acquainted with the subject, so that I have a glimpse what electronics is all about. Having followed the development of great homebrew Amiga projects on this forum and others (i.e. Dennis' MiniMig), I've got an itch to get my hands dirty in a project. Although it's starting to  seem that I'm way over my head with this project. But hey, you can't blaim a guy for trying.  :-D

Obviously, I would get easier and probably cheaper off if just bought an accelerator second hand. Of course, when I came up with this idea, the first thing was to try to find out whether similar homebrew projects exist already. The lack of such projects made me to bring this subject up on this forum. Having followed this thread, I've let myself understand that there isn't a modern chip available that would be downwards compatible with the MC680x0 family of chips. PPC is one thing but I suppose that is more viable if you want to use OS4.