MiniMig with AGA

[Hattig]

I'll definitely be interested in one of the Rev B products.

Are there pin headers on board for parallel and serial? How about pin headers for expansion (CPU bus, etc) - although I guess that the FPGA is better for expansion internally, if there is space. I like the idea of adding an AHI compatible 16-bit audio system inside, with an internal mixer for paula audio as well.

It would be nice to have a centralised website where all these projects could be hosted and have updates - it's very hard to keep track of them all, all the tweaks, the softcores, etc.

Good luck with the testing and production builds!

[Hattig]

I disagree with that because nobody is going to write drivers for that extensions. So zero software would actually use it. In contrast if a gfx chip supported by P96 or CGX was supported we suddenly would be able to run ALL amiga RTG software.

Why not expose the new chunky "AGA++"/AAA modes via a P96 driver?

The biggest difficulty with adding chunky modes (or a chunky P96 compatible video card alongside the AGA functionality) is that it isn't just the video output circuitry that needs to support chunky, the blitter and other features also need to support it - possibly including chunky-planar conversions.

The biggest advantage on a modern design is RAM quantity and speed. Hopefully this will be quite nice on this card, so 256-colour AGA will work nicely without slowing down the rest of the system... :p

[Hattig]

If youvare going to use a P96 driver, why not use a nice modern chip from ATI or Nvidia? It will be cheaper and more powerful than any FPGA solution.

I think that misses the point of the FPGAArcade by a clear mile, whilst extending AGA with chunky modes for fun and then looking for a way to make them accessible to serious applications doesn't.

[Hattig]

Plane mode is least efficient from a DRAM point of view. What I did with the ST is to burst read a chunk from each plane from the DRAM and hold it in a local RAM cache. I can then combine each plane how I like. The DRAM access is very efficient as it is reading 8 word burst access per plane. I wrote it to support 32 planes, and they all share one RAM block.

I didn't even know the ST had planar graphics modes. I thought it supported 16, 4 or 2 colours on screen, in a chunky format.

However planar graphics in conjunction with a fast (potentially full scanline, but actually only needs to be the optimal size of a DRAM burst access per plane) cache as you describe isn't so bad for outputting the graphics to a monitor.

[Hattig]

How is the software / firmware / softcores side of things progressing?

[Hattig]

We could see all sorts of things emerge, an extended AGA chipset (more planes/resolutions/sprites), AGA but with no DMA contention, a built-in p96/cgfx compatible RTG graphics card and vga switcher or anything else we can squeeze in.

I think the NatAmi is doing a lot of this.

Personally I think that it is better to do enhancements that enhance existing software, than to add new features that require developer support.

Therefore a contention-less AGA would benefit people, and more resolutions would provide more options for software, but more planes or sprites or chunky bitmap modes would have to wait for software to be enhanced to use it. I guess the chunky modes could have a P96 driver created for them.

[Hattig]

The WinUAE Amiga side P96 driver is closed source. Nevertheless the Minimig on the Replay board will have a true-colour RTG board and dedicated P96 driver (yes, I will write it). It won't happen before the 020 compatible CPU core is available (guess why).

The AGA chipset will be enhanced with chunky modes very soon. The chip memory bus access is already contention free but will be improved significantly in the near future.

Brilliant stuff, and good to know about the need to access lots of memory.

How's the performance currently?

[Hattig]

Nope it's not faster then a 030 50Mhz.

The softcore speeds are:

TG68K @ 7.09 MHz : 4.27 x A500 / 1.85 x A1200
TG68K.C @ 7.09 MHz : 5.81 x A500 / 2.52 x A1200

To be fair, yaqube did say he would try 14.18MHz with a memory buffer (cache?). That would be 5x A1200, which would be akin to a 70MHz '020 / 90MHz 68000.

I don't know if the A1200 speeds were with fast RAM or not, but looking at the numbers I think not.

Any more information? It's not really the most important thing, it's already ~a 45MHz 68000, which beats the old Supra28 A500 accelerator!

[Hattig]

Best of luck with this! It's getting to the exciting stage. Looking forward to the photos of the board running something AGA!

[Hattig]

I've looked at these before. The Coldfire is a really bad fit for Amiga... Ok we can now build AROS for the coldfire, so we can get AmigaOS functionality, but that doesn't magically make Amiga programs coldfire capable... There is a really good link on the Coldire wiki page that lists the differences between the coldfire and the 68k... It's pretty horrific... Coldfire software should run on a 68k, the reverse isn't true

Good lord, don't stop him porting AROS to this board - look at them running some form of Atari ST TOS operating system, we must stop their suffering.

[Hattig]

Hurrah! Good news everybody :-)

The AGA Minimig core with the TG68.C V4.0 (partial 020 compatibility) takes about 60% of the FPGA.

The aforementioned CPU core takes about 20% of the FPGA.

So plenty of room for expansion! :-) An entire AGA Amiga in around 1 million gates?

How is the TG68.C V4.0 in terms of performance on the Xilinx XC3S1600E?

[Hattig]

That's good news! I didn't even know that the Replay had a different audio mechanism. Does it mix with the standard Amiga audio output?

[Hattig]

Can the DAC also do sampling in? Of course that would require a Mic input...

Good work, btw, both yaqube and mikej. It's much appreciated.

[Hattig]

Results for 28 MHz clock and 256-byte instruction cache. Maybe the speed gain is not impressive but I have yet another option to try.

http://www.yaqube.neostrada.pl/images/SysInfo28-256.gif

I think this is very impressive for a core that was merely a 68000 core a short time ago. Is this with the 16-bit data bus still? Would there be any advantage to implementing a data cache?

5x faster than a stock A1200 and faster than an A3000 is not to be sniffed at.

[Hattig]

Hi,

@cammy,

Yes that is impressive, and with a few more instructions maybe we can have the new expensive minimig board roll over and play dead. Yes I think that the minimig board that duplicates the A500 was a good school project but I saw something the other day walking through a flea market, an Amiga 500 game console that came with 50 Amiga games all embedded in a joystick. It ran off of a couple of double A batts and you could carry it around take it over your friends house and play the 50 Amiga games on it, you could even order a flash card with another 50 games for $49.95.

Now that was impressive.

smerf

And you didn't buy it? You didn't take photos to post here?

Most likely some Chinese company took the Minimig VHDL and made an ASIC with it. Or it's emulated on a fast ARM SoC.