TG68 - The Open Source Minimig CPU into the FPGA

[tonyyeb]

alexh wrote:
If you have a MiniMig v1.x board you dont care.

If are a potential MiniMig v1.x owner, but cannot find one cos no one makes them, and were considering getting buying one of the slightly expensive off the shelf Altera DE1 or DE2 boards then you are wetting your pants as a more compatible DEx-MiniMig is on it's way.

So what will be the future? I buy a dev board and then i can download new code and use Amiga software on it?

[downix]

tonyyeb wrote:

alexh wrote:
If you have a MiniMig v1.x board you dont care.

If are a potential MiniMig v1.x owner, but cannot find one cos no one makes them, and were considering getting buying one of the slightly expensive off the shelf Altera DE1 or DE2 boards then you are wetting your pants as a more compatible DEx-MiniMig is on it's way.

So what will be the future? I buy a dev board and then i can download new code and use Amiga software on it?

Right

[amigadave]

@Belial6,

Thanks for the reply, but from what I have read so far, it did not appear to me that Dennis and Tobiflex are arguing.

My understanding is that the DE1 board just barely is capable of running Minimig v1.1 and has no capacity for improvements which is where many others, including Dennis want to go (v2.0 and beyond).  Maybe I am wrong and have misunderstood some of the previous messages that I read too quickly, but I am pretty sure that Dennis is interested in v2.0 and less interested in just porting v1.1 to other boards.

In any case, I still like freqmax's suggested board, but have not checked on what it cost, I'll go back and look it up and try checking for pricing?  Perhaps someone besides Tobiflex or Dennis will take enough interest in it to try to port the Minimig code, open core 68000 code and some other enhancements to it in the near future.

@Tonyeb, that is what is so exciting about the work being done on the Minimig project.  It has "legs" and anyone with the skill and desire can take what has already been done by Dennis and Tobias and "RUN" with it to create a bigger and better project.

[Belial6]

I did not mean to imply that Dennis and Tobiflex are arguing.  Tobiflex suggested the DE-1 board to Dennis.  As far as I know, Dennis has not stated his opinion on the DE-1 board one way or the other.  I do know that he has not responded one way or the other to my PM with an offer to just buy one for him.

It is other people that have argued that the DE-1 was unsuitable for the MiniMig.

Unfortunately, it looks like the DE-1 code is just that.  Code.  I am eagerly waiting for someone to compile everything so that I can download a zip file that can be unzipped to an SD card that I have put my own kickstart images on, and boot on the DE-1.  As soon that is available, I will order a DE-1 board for myself.

[amigadave]

freqmax wrote:(edited)
Or you could use: http://www.xess.com/prod035.php3

1 million gate FPGA plus all the features below for $199

Features:

    * XC3S1000 FPGA
    * XC9572XL CPLD
    * 32 MByte SDRAM
    * 2 MByte Flash
    * 100 MHz oscillator
    * Parallel port
    * Keyboard/mouse PS/2 port
    * 512-color VGA port
    * 7-segment LED
    * 2 pushbuttons
    * 4 DIP switches
    * 84-pin prototyping interface (65 free I/O pins)
    * 5V DC power jack
    * 5V / 3.3V / 2.5V / 1.2V regulators
    * Downloading cable
    * XSTOOLs CD
    * Works with XILINX ISE, WebPACK, iMPACT and ChipScope software
    * Requires Win2K or WinXP. Not supported by Win98 or WinME!

Looks very tempting to me at that price.  I hope someone attempts to get the Minimig running on it and an open core soft 68020+ is created to put in it.  (I wonder how long it would take me to learn Verilog and VDHL?) :lol:  (probably pretty long since it is VHDL, not VDHL)

Oh, and don't forget this is included too!

Documentation:

    * XSA-3S1000 Board Manual (PDF file)
    * XSA-3S1000 Board Pin List (Excel file)

And these are available:

Accessories:

    * XSUSB USB interface
    * XStend Board 3.0
    * 5 VDC power supply (North America)
    * 5 VDC power supply (Europe)

[freqmax]

There's also this alternative:
http://www.enterpoint.co.uk/moelbryn/darnaw1.html
(140 USD, 110 EUR)

Then add a through hole accessory board with:
 * Video out (VGA/SCART)
 * Audio  
 * Joystick
 * Flashmemory card
 * Keyboard (PS/2)
 * RS232
 * +5V and +3,3V regulators

And it will do the same as a Minimig1 provided the softcpu and sdram interface is up to the task.

[amigadave]

@freqmax,

I liked your first suggestion much better, and for an A1200 like Minimig, the 32mb RAM should be plenty.  I am sure it is going to take a huge amount of developing effort to get from v1.1 OCS to v3.x AGA w/68020+ soft CPU.  Perhaps as much, or more than Dennis had to put into the whole of his design?  It would be great if a small team of programmers would take it on and speed its completion.  A Minimig that is equivalent to an 030/50mHz A1200 in a small form factor and at a cost of $300 to $450 would be quite an accomplishment.

[downix]

amigadave wrote:
@freqmax,

I liked your first suggestion much better, and for an A1200 like Minimig, the 32mb RAM should be plenty.  I am sure it is going to take a huge amount of developing effort to get from v1.1 OCS to v3.x AGA w/68020+ soft CPU.  Perhaps as much, or more than Dennis had to put into the whole of his design?  It would be great if a small team of programmers would take it on and speed its completion.  A Minimig that is equivalent to an 030/50mHz A1200 in a small form factor and at a cost of $300 to $450 would be quite an accomplishment.

The FPGA you'd need to fit the 68020 + AGA would run you approx $300 on it's own.  Add in the other parts, you'd likely be nearing $1000.

Give it a few years.  We need to hit ECS before we go AGA, and by that time, prices/density should have come down.  Heck, the chip density used in MiniMig 1.0 would have cost almost $1000 not even 3 years ago IIRC.

[FrenchShark]

downix wrote:

amigadave wrote:
@freqmax,

I liked your first suggestion much better, and for an A1200 like Minimig, the 32mb RAM should be plenty.  I am sure it is going to take a huge amount of developing effort to get from v1.1 OCS to v3.x AGA w/68020+ soft CPU.  Perhaps as much, or more than Dennis had to put into the whole of his design?  It would be great if a small team of programmers would take it on and speed its completion.  A Minimig that is equivalent to an 030/50mHz A1200 in a small form factor and at a cost of $300 to $450 would be quite an accomplishment.

The FPGA you'd need to fit the 68020 + AGA would run you approx $300 on it's own.  Add in the other parts, you'd likely be nearing $1000.

Give it a few years.  We need to hit ECS before we go AGA, and by that time, prices/density should have come down.  Heck, the chip density used in MiniMig 1.0 would have cost almost $1000 not even 3 years ago IIRC.

I am actually starting this kind of development.
My ultimate goal is to use a Stratix II dev board to create an accelerated AGA Amiga.
Here are the characteristics:
- 32 MB of NOR flash (have already replaced the 16 MB chip by a 32 MB one) -> it will contain the kickstarts and the workbench flash disk.
- 32 MB of DDR SDRAM (64 MB planned) -> it will be the fast RAM
- 2 MB of SSRAM (8MB planned) -> it will be the chip RAM
- Planned (chip and fast) bus speed : 114.5 MHz.
- 100Mb Ethernet
- IDE port
- Stratix 2S60 (3 million gate chip)
- VGA, S-Video, Joy ports, parallel, floppy and keyboard will be on a Santa Cruz daughter board
I want to mount the dev board into an A1200 case with a slim DVD-ROM drive and a compact flash drive.
That will be my dream Amiga :-D

[jkonstan]

The Stratix II development board sounds like a nice idea; however, it is really expensive right now ($2K to $3K) unless you have access to one via work or school. Affordable FPGA evaluation boards today, would be Altera Cyclone II/III or Xilinx Spartan 3e.

 :-)

[amigadave]

@downix,

Not being as tech savvy as you, I have no idea how many gates the FPGA would need to do the AGA Minimig code plus a softcore 68020+ and thought the 1 million gates could maybe do it.

The price of $199 for the 1 million gates plus all the other features seems like a very good price and it has expansion capabilities which could perhaps contain either the 68020+ in another FPGA, or the AGA, or better display device code.  I just thought that for the price, there was a lot going for the XSA-3S1000 Spartan 3 Prototyping Board and since price is a big consideration for many wanting a Minimig, the XSA-3S1000 was worth looking at a little closer by someone with the capability to produce something.

So, I brought it up a few times just in case it would help point it out to the right people.

Oh, and I totally agree that ECS needs to be next and you are right that we must walk before we can run.  I know that AGA and/or 68020+ will likely be a year or more into the future.

@FrenchShark,

Sounds like a great plan, are you going to be working alone, or with others to make it a reality?  The reason I ask is that I hope that the huge effort to create AGA and/or 68020+ softcore code is at least done in cooperation and not several individuals working on the same thing at the same time and having it take many times as long to complete as it would if they work together.

@Dennis,

See what you have started!  My congratulations to you again sir.  You have done more for the Amiga community than any single person since Jay Miner.    :bow:

[amigadave]

jkonstan wrote:
The Stratix II development board sounds like a nice idea; however, it is really expensive right now ($2K to $3K) unless you have access to one via work or school. Affordable FPGA evaluation boards today, would be Altera Cyclone II/III or Xilinx Spartan 3e.

 :-)

My point about cost and the XSA-3S1000.

[jkonstan]

I would recommend a similar approach on starting to get on route to an A1200 level of a MiniMig 2.0 by use of a larger FPGA, SDRAM, and an external 68020 CPU (need CBT level shifters) since it is going to take some work to get that really nice soft VHDL 68K core upgraded to 68020 level (Icache, pipline etc ...) and to get AGA graphics.

  :-)

[freqmax]

For development purposes a real MC68020 plus 5V<->3.3V interface would be useful. Since you can then run the softcore  "68020" and the real thing in parallel executing the same instructions.
Then trigger points within the code can catch any anomalities. And the transactions causeing it can be logged.

Also what benefits does 68060 have over 68020 that really makes a practical difference from a functional view?, as the ballgame becomes different when you can do things in HDL.

Is there anyone that could make a realistic estimate of the FPGA capacity needed for AGA ..? (diesize x nm process), to nail weather it's feasable or have to be postponed for years until logic capacity goes down in price.

I did an rough estimate in Oct-2005 on the number of gates in the Amiga using the pictures of Lorraine#1 #2. By using the average number of gates in 74xx logic chips say 60 gates x 30 rows x 5 columns x 15 boards = 135000 gates.

[ChaosLord]

freqmax wrote:

Also what benefits does 68060 have over 68020 that really makes a practical difference from a functional view?, as the ballgame becomes different when you can do things in HDL.

68060 has 8k data cache
68020 has 0k data cache

68060 has 8k instruction cache
68020 has 0.25k instruction cache

68060 can execute 2 instructions simultaneously (3 if you count predicted branches)
68020 can execute only 1 instruction at a time

68060 has many floating point instructions
68020 has no floating point instructions

[Belial6]

So, from the software running on the machines point of view, the only difference is that the 68060 has floating point?