MiniMig with AGA

[Louis Dias]

Can we please keep to something that is going to happen during this decade (or at all)?

Yeah cuz FPGA tech will never be able to run a Amiga software...

Oh wait?!  O_o

[Louis Dias]

Excuse me? Who claimed anything like that?

Everytime *Natami* comes up, you have something negative to say.
The people on the team have CURRENT (not ancient) industry experience.

Now that we have Minimig and FPGArcade, is it such a stretch to see a similar device with even more oomph?

Here you'll see Gunnar has already developed an Amiga 3D card years ago:
http://www.natami.net/knowledge.php?b=2¬e=16454&x=1

As you can see, yacube has an AGA core...is it so hard to expand upon or add to that?

There goal is an Amiga 5000 if Commodore had still been around in 1997ish or better if possible.  Seems attainable since *the community* has already replicated (with minor enhancements) what Commodore release to the public up to its demise in 1994.

The Minimig core is cycle-accurate with the 68000 and because it has a faster memory interface on the FPGArcade board, it performs 1.85x better (atleast that is my understanding).  The latter Amigas did not use the 68000 and still maintained high compatibility.

Here: http://www.natami.net/knowledge.php?b=2¬e=17027
You can see how much improved the N050 will be in processing instructions over legacy hardware because cycle accuracy is not necessary when moving forward in progress.  Alot of stuff will still run, some stuff will break just like when the 3000 was released and then the 1200/4000 from there...  Big deal.  The goal in an A5000, not A500/3000/1200/4000 emulation.

Recall the flak Dennis took on this very forum when he first announced his project.

As you are (iirc) a MorphOS developer, my feelings are that such a project succeeding may cause a conflict of interest for you here.  However, you could put your talent to use and join the team and write new libraries to take advantage of the platform.

[Louis Dias]

Forgot about that. For some reason I built an analogy in my mind that this would be similar to running software emulation on top of a host system, but indeed that is not the case. Here we have to deal with the memory interface directly, which indeed is quite slow. Just shows how much of a software guy I really am.


Nothing.

So I assume the 68000 softcore has some kind of built-in cache to compensate for the slow memory interface, too?

The memory interface on the NATAMI is much faster and uses DDR2(iirc).  It's also going to have 32k or 64k cache.
Also, I believe their bus will be 32bits.

[Louis Dias]

Hi,

When I initially designed the board I did a lot of analysis of the memory timings. DDR2 is not better than DDR1 in this application. DDR2 is basically optimised for faster clock speeds, the memory access time is about the same. For the speeds we can obtain with the FPGA (133-166MHz) then DDR1 is the better choice.

The memory bandwidth is not the problem, for example the 16 bit memory at 133Mhz will deliver over 4GBytes per second in burst sequential access. We are using the DRAM at the moment to emulate large amounts of static RAM, and we can get cycle accurate timing for the original hardware.

The softcores are not at the level of compatibility that I would like at the moment. The reason I have a 68000, 68030 and 68060 on daughter boards is to develop and test new softcores. As has been mentioned before, using FPGA internal SRAM as a cache can significantly increase performance.

People always complain about the cost, so for where we are now the memory system is about optimal cost/performance level.


The expansion board idea is a way to increase the performance above the basic card. We can develop boards with either real processors, or faster FPGAs with dedicated memory.

The most important thing for me is to get a stable board which can be mass produced which meets 90% of our current desires and has potential for expansion for the people that want that.

I imagine a future version without the expansion connectors and with several banks of memory instead.

4 connections to make on the layout, still on track for manufacture on Monday.

/Mike
in a bar in China....

Hi Mikej,

Kudos to you.  The board you designed is to emulate older hardware and I understand why it was designed the way it was and how you can get some enhancements as it exists now.  What I said earlier about the NATAMI was not intended to take anything away from your design.

I was merely pointing out, as people seem to forget, that the target for the NATAMI is a new evolutionary 68K Amiga platform.  An up and down evolution over the A4000, not an emulation of it or older hardware.

What you have done is excellent.  It's basically a hardware MAME and it does that job perfectly.

[Louis Dias]

Put a PCI bus on this bad boy and a couple of slots attached and you've got the Natami beat by a stretch!

How so?  Also, don't you think that will increase the cost?
Regardless, they are aimed at different markets with different needs.

I could see the NATAMI team purchasing one of these boards to test their cpu core.

[Louis Dias]

While a custom p96 driver would be great (eg UAEGFX), to just do the WINUAE trick and boost chipram to 8mb and make it a faster would allow Ratte to develop some nice new drivers/modes.   Even without any new modes you'd be able to run in 1024x768 and 256 colours....just try it in WINUAE if you want to see how it performs

I guess no one noticed in the NATAMI demo that they were running 16MB of chip ram...

[Louis Dias]

I guess you didn't notice in the MiniMig AGA demo that it was running 50MB of chip ram...

I guess you didn't notice the NatAmi is eventually supposed to have 256 Megs Chip and 256 Megs fast...

[Louis Dias]

Ohwell.. I took it as another pin mismatch so it won't work(tm).

Anyway  I found a page about connecting the Sega joypad to another console. He mentions "However, the newer Genesis controllers have a custom IC that adds the extra signals onto the same 9-wire cable. ". So the plain version should work out of the box. For the other one with an special IC one can use the following trick:

 (+5V input) --(Resistor 1-10k) -- (+3,3V output) -- (<3,3V Zenerdiode) -- (GND)

Then one just feed the +5V outputs of the joypad etc.. to this arrangement and connect the max 3,3V level to the FPGA. And ofcourse the joypad has to be feed with +5V supply.

That must be for the 6-button pads(+mode button)...

[Louis Dias]

Good luck finding USB keyboards and mice that still support the PS/2 protocol.

[Louis Dias]

IIRC Jens stated that the Keyrah will not work with a USB-PS2 adapter.

I just tested it with my Minimig V1.1 using the adapter and the keyboard did not function.  I then removed the adapter and plugged it into a PC PS2 port and (after loading the drivers) it worked.  I'll take a guess and suggest that perhaps whatever the Keyrah needs for power to work isn't supplied over the PS2 port (remember that the Keyrah is more than just a keyboard adapter - it also supplies two Amiga joystick ports).

Either way we need something else (and should something else come along we'll also need an alternative for the now non-existant PC "F12" key).

Well, supplying power shouldn't be much of an issue...

Did I miss a link to a keyrah?

[Louis Dias]

Alright, I'll rephrase my statement... TWO guys in their bedrooms built a better CPU than Motorola's entire CPU dev team in the whole 80's...

Well, cache greatly improves performance on any cpu but back then perhaps the bottlenecks were elsewhere...

[Louis Dias]

2 more weeks!

:-)

But seriously, even without "feature creep" these things take time.  Considering Commodore bailed in '93/94, you'd think patience would be the order of the day in this community...  

FYI: In a few weeks this thread will be 2 years old!