New Universal Agnus 8372, 75 etc

[nonarkitten]

Originally Posted by alexh

Sounds interesting. Not only the Agnus/Alice but the RAM as well? Similar to a MultiMega?

You mean the DKB MegaChip? Yeah. I googled "MultiMega" and only got results about the Sega Multi-Mega (the CDX here).

[kipper2k]

Originally Posted by jbenam

One step closer to being able to build entirely new Amigas (and also save some old ones )

We are almost there, all the custom chips are on the radar, we already have a few a500+ clones out there so it is becoming a reality

[nonarkitten]

Originally Posted by AmigaHope

How does this work? Does it remove nasty mode DMA contention for the CPU side somehow? Only through max-speed blitter, or something else?

Simultaneous service of both data buses?

1. It's complicated
2. Yesish
3. This is not required but helps
4. This is not required but helps

You're absolutely right that running things async can have big benefits. For example, the CPU could still read from the prior burst fetch while Agnus is busy with the Copper or Blitter. Our pSRAM is more than fast enough and for chip-to-Denise, we just double/quad pump the bus. But CPU-to-chip is still a problem and the best we can do is have Gary sometimes give the CPU both odd and even cycles. This approaches but does not match AGA performance and C2P-heavy games and demos could suffer.

There are potentially some ways around this though.

[nonarkitten]

Originally Posted by Promilus

Oh, I see. ICE65 with 2x external 4bit PSRAM if I'm reading schematics correctly. And that's because when all banks are powered from 3.3V it makes I/O 5V tolerant so it doesn't require additional level shifters?

Correct.

[alexh]

Originally Posted by nonarkitten

You mean the DKB MegaChip? Yeah. I googled "MultiMega" and only got results about the Sega Multi-Mega (the CDX here).

Yeah, MegAChip or MiniMegi must have had a brain fart while typing.

[Turrican_3]

Very interesting stuff!

[nonarkitten]

Originally Posted by alexh

Yeah, MegAChip or MiniMegi must have had a brain fart while typing.

It happens to me all the time. No worries.

[btp]

A newly produced drop-in replacement is simply awesome!
Big congratulations to Renee and the rest of the developers & contributors for all the hard work!

[Dynamic_Computi]

Wow... This is a nice bit of news. I hope everything works out with it! Great job getting it to work!
If you want me (10 Minute Amiga Retro Cast) to do a review or preview of it, let me know.

[dirkies]

typo in nonarkitten's github page in configuration table:

8327b (twice)

[Photon]

Seems an exciting prospect! It would have been nice to pan the audio channels on the original chip. It would have been nice also if the Blitter could have gone faster on AGA.

It's important that blitting vs DMA saturation matches or exceeds the Blitter cycle diagram in HRM, so that the finish time of large and small blits are equal or faster.

To display a correct picture, timing all the elements to the same horizontal positions as original Agnus chips is important. Copper, Blitter, or CPU changes to COLORxx, sprite pixels and positions, and bitmap pixels should match up as on original Agnus.

Other aspects may be more forgiving.

[nonarkitten]

Originally Posted by Photon

Seems an exciting prospect! It would have been nice to pan the audio channels on the original chip. It would have been nice also if the Blitter could have gone faster on AGA.

It's important that blitting vs DMA saturation matches or exceeds the Blitter cycle diagram in HRM, so that the finish time of large and small blits are equal or faster.

To display a correct picture, timing all the elements to the same horizontal positions as original Agnus chips is important. Copper, Blitter, or CPU changes to COLORxx, sprite pixels and positions, and bitmap pixels should match up as on original Agnus.

Other aspects may be more forgiving.

Well, panning on *PAULA* might be possible, but rather tricky on Agnus.

But all the firmware will be open source, so if people want to do this, then power to them. In fact, I have no problem adding stuff if people are willing to write AHI and P96 drivers and that it doesn't do anything those libraries can't already do.

[Promilus]

@Photon - well Apollo Team already have proven getting "updated" chipset is viable. It just happens some ppl (like me) don't want to abandon on-board chipset. New IC replacement with new options and 100% compatibility with old ones is pretty neat idea.

[kipper2k]

Originally Posted by Promilus

@Photon - well Apollo Team already have proven getting "updated" chipset is viable. It just happens some ppl (like me) don't want to abandon on-board chipset. New IC replacement with new options and 100% compatibility with old ones is pretty neat idea.

i like compatibility... i like the idea to try to replace unavailable components... i like trying to keep Amiga as close to original as possible. I like the idea of using an Amiga in an Amiga case.

For this reason this is why we are trying to replace the custom components. AND making them open source so other people can help us if they can.

[nonarkitten]

Originally Posted by Promilus

@Photon - well Apollo Team already have proven getting "updated" chipset is viable. It just happens some ppl (like me) don't want to abandon on-board chipset. New IC replacement with new options and 100% compatibility with old ones is pretty neat idea.

We have it right on the github page that adopting some of the Vampire's AGA additions might be possible. While we don't quite have SAGA's bandwidth, somethings I think are doable, at least in theory:
- sprite multicolour mode
- sprite horizontal flip
- sprite parallax mode
- 8-bit chunky modes up to super hires (would need P96)
- 16-bit chunky modes up to hires (ditto)
- 24/32-bit chunky modes in lores (ditto)
- scan doubling for everyone (without mode-promotion)
- paula pan control (would need AHI drivers)
- paula second channel set (ditto)
- paula 16-bit audio (ditto)

Sure, all are doable (some even trivially). Not until everything else is working though.

[btp]

Originally Posted by nonarkitten

Sure, all are doable (some even trivially). Not until everything else is working though.

I applaud that you have set 100% compatibility as the base line here

[alexh]

One chip that is impossible to find is the A3000 Super DMAC-04 if there was a possibility to create new Super DMACs that solve the issues lots of A3000 users have it would be a big help to a small subset of Amiga users.

[grelbfarlk]

Originally Posted by alexh

One chip that is impossible to find is the A3000 Super DMAC-04 if there was a possibility to create new Super DMACs that solve the issues lots of A3000 users have it would be a big help to a small subset of Amiga users.

Not to sound like an asshole, but why do we want a replacement Super DMAC? I know because for A3000 users to support the DMA SCSI on the A3000. But with limited resources why isn't it a better idea to focus on replacing Lisa, Alice, Ramsey and Buster chips?

[alexh]

Originally Posted by grelbfarlk

Not to sound like an asshole, but why do we want a replacement Super DMAC? I know because for A3000 users to support the DMA SCSI on the A3000. But with limited resources why isn't it a better idea to focus on replacing Lisa, Alice, Ramsey and Buster chips?

Because iComp.de is already making a replacement Buster. Ramsey chips are readily available (at least at the moment). THIS IS an Alice replacement (didn't you read?). The Lisa chips are also available, the Lisa functionality has already been converted into as HDL (MiniMig AGA) so could with some re-work become a standalone chip (i.e. not as much re-engineering required).

There are simply no sources of Super DMAC.

It's obvious that you don't have an A3000, but I imagine A3000 owners, or those trying to make Chucky's ReAmiga 3000 would be glad of a source.

[pandy71]

Originally Posted by alexh

Because iComp.de is already making a replacement Buster. Ramsey chips are readily available (at least at the moment). THIS IS an Alice replacement (didn't you read?). The Lisa chips are also available, the Lisa functionality has already been converted into as HDL (MiniMig AGA) so could with some re-work become a standalone chip (i.e. not as much re-engineering required).

There are simply no sources of Super DMAC.

It's obvious that you don't have an A3000, but I imagine A3000 owners, or those trying to make Chucky's ReAmiga 3000 would be glad of a source.

All above is true but... but SDMAC doesn't define Amiga and it can be substituted by other DMA controllers (of course not P&P).
Also there is limited demand for this particular IC.
And yes, i own A3000 and i would be happy to have opportunity to get new SDMAC but let's focus on most important functionality especially with limited resources.

To be honest i'm not aware is there any detailed specification for Super DMAC.
SDMAC must be re-implemented based on software reverse engineering - how accurate this can be...?