8 meg agnus mod

[source]

I was just wondering with all that is being emulated through fpga is ther a way to create a small piggy back board that will give a means to have a 2 meg agnus or even an 8 megapixel agnus through fpga. The vampire is supposed to be a stand alone all running through fpga. Winuae does it through software. I was on the individual computer site and they are saying that his boards are the last to be made due to lack of chips. I just got me to thinking that's all. I have no understanding of how the hardware works etc. Could the a500 address a 8 meg agnus? Is there hardware limitations in the first gen amigas that will only allow adresssing 2 meg. Just out of curiosity.

[Jope]

The only machine that can potentially access 8MB of Chip RAM without modifications is the A4000. No-one knows if the motherboard really supports it, as no such Alice has been created.

For the A500, all of the Chip RAM would have to be on the Agnus board and you'd need to add jump wires for the missing address lines and such. Quickly thinking, GARY will probably need an FPGA replacement also.

[1NOM155]

That is a noce idea, and some people already talk about for a while.
in mean time Apollo team already made some test on Vampire, have a check here http://www.apollo-core.com/knowledge...209&order=&x=4 .

[Jope]

The Vampire re-implements a lot more in its FPGA, so it is of course possible to do this with the Vampire.

It's not relevant to this discussion however, source asked about something that replaces the Agnus (presumably this would plug into the Agnus socket).

[zero]

Is there any advantage to have 8 meg chip? What would you use it for?

[Jope]

If you open up lots of colourful screens, it tends to run down pretty quickly.

[1NOM155]

Originally Posted by Jope

The Vampire re-implements a lot more in its FPGA, so it is of course possible to do this with the Vampire.

It's not relevant to this discussion however, source asked about something that replaces the Agnus (presumably this would plug into the Agnus socket).

Not total correct (about A4000) Alice for example never had support for the 11bit memory address so in this case can't address the 8mb chip ram, the jumper exist but don't anything.

Also Team Apollo explains how they can address 11mb of chip ram. Combining 2mb + 8mb (usually knowed by fast ram zorro 2 24bit ) + ranger ram aka slow ram 1.8mb, it's a matter of reading more about the thread were someone explain the Amiga memory range table.

If team Apollo want, they can release part of their core to support a custom bigger Agnus / Lisa chip but to be honest in mine opinion, and what I saw about the future of the project Vampire doesn't make sense releasing a custom chip (fpga) only to support it, thats why I linked the thread about a bigger chip ram implemented in a test core.

They have release Vampire 500 V2+ support A1000 / A500 / A2000, 600 V2+ for A600, the new V1200 for V1200 witch the later can be adapt to support the CD32 and later they will make one to support A3000/A4000. if the tests run smoth they will implement a bigger chip ram for all.

I understand some people dont like the project and want to have the original accelerators, RTG cards, Sound cards, I/O etc but the prices of thoses are increasing due to their rarity. I belive Vampire will provide us the future to expand our Amigas further, but everyone is free to choose what ever they want for theirs Amigas.

[Jope]

Originally Posted by 1NOM155

Not total correct (about A4000) Alice for example never had support for the 11bit memory address so in this case can't address the 8mb chip ram, the jumper exist but don't anything.

Yes. And I quote myself:

Originally Posted by Jope

No-one knows if the motherboard really supports it, as no such Alice has been created.

Originally Posted by 1NOM155

Also Team Apollo explains how they can address 11mb of chip ram. Combining 2mb + 8mb (usually knowed by fast ram zorro 2 24bit ) + ranger ram aka slow ram 1.8mb, it's a matter of reading more about the thread were someone explain the Amiga memory range table.

If team Apollo want, they can release part of their core to support a custom bigger Agnus / Lisa chip but to be honest in mine opinion, and what I saw about the future of the project Vampire doesn't make sense releasing a custom chip (fpga) only to support it, thats why I linked the thread about a bigger chip ram implemented in a test core.

They have release Vampire 500 V2+ support A1000 / A500 / A2000, 600 V2+ for A600, the new V1200 for V1200 witch the later can be adapt to support the CD32 and later they will make one to support A3000/A4000. if the tests run smoth they will implement a bigger chip ram for all.

I understand some people dont like the project and want to have the original accelerators, RTG cards, Sound cards, I/O etc but the prices of thoses are increasing due to their rarity. I belive Vampire will provide us the future to expand our Amigas further, but everyone is free to choose what ever they want for theirs Amigas.

Seriously? How is this relevant in a thread about a plug-in Agnus replacement? Perhaps it's best if you leave any information about spinoff projects to be communicated by Team Apollo?

Of course they are free to do whatever in the Vampire - it re-implements many of the Amiga custom chips in its FPGA!

This is a totally different playing field than recreating individual custom chips in FPGA as plugin modules. The Vampire plugs into the CPU slot and much of the action happens on the Vampire board, not in the Amiga's motherboard any more.

Please stop talking about the Vampire in this thread now, it is not relevant to this discussion.

[AmigaHope]

Yeah, general consensus 8M CHIP is basically not possible unless you rolled yourself a whole new Amiga platform including new Lisa and Alice, and a new motherboard. The address lines just aren't there. I too have looked longingly at that 8M CHIP jumper on my A4000, but it's a phantom dream.

[zipper]

It's meant for free sex and beer...

[Jope]

Originally Posted by AmigaHope

Yeah, general consensus 8M CHIP is basically not possible unless you rolled yourself a whole new Amiga platform including new Lisa and Alice, and a new motherboard. The address lines just aren't there. I too have looked longingly at that 8M CHIP jumper on my A4000, but it's a phantom dream.

It depends on the SIMM geometry that you plug in.

10 address lines are there (1024 bits). Two rows, four columns are controllable by U212.

Each RAS signal can hide behind it a geometry of 1024x4096 bits = 4MB. But you have two RAS signals.

So the SIMM you use must be at least 2048x4096 total and it must respond to two RAS signals. Not impossible, but "not just any" stick will work, as some have a very shallow geometry thanks to the amount of address pins available in the 72 pin SIMM specification.

[AJCopland]

How does a, for example, A1200 know how much chipram it has?
Is it autoconfig'd?

It's just that there's been a couple of remakes of the A1200 motherboard recently, so if the address lines are there would it be possible to wire up 4MB of chipram?

[Daedalus]

Chip RAM isn't detected through the normal Autoconfig route. It's probed very early in the Kickstart hardware initialisation process. Adding address lines to the board isn't going to help things since the custom chips themselves also lack the address lines necessary, so you would need a completely new chipset to go with the new boards.

[Toni Wilen]

"Only" Agnus/Alice needs address line upgrade, Agnus/Alice is the only chip that addresses Chip RAM. Other chips talk via data bus and separate 8-bit RGA address bus that Agnus uses to DMA access Denise/Lisa and Paula registers, CPU also addresses Chip RAM indirectly via Agnus/Alice.

>2M Chip RAM will conflict with Z2 RAM boards (E90000+ space IO boards work fine). Enabling (ROM patch) >2M Chip detection is the easy part..

[AJCopland]

Ok, thanks for the info.
I figured that it wouldn't be easy even if possible because of the memory map.
However if it requires replacing Alice entirely, rather than using existing address lines on the existing Alice chip, then I guess it's a no go.

[Toni Wilen]

Perhaps it would work after all.. Megachip like extra board with Agnus socket + 2M or 6M of RAM (or perhaps it is easier to have all chipram on board) + FPGA + 1 or 2 address lines that needs connecting to CPU pins.

FPGA could snoop RGA bus addresses and store extra 2 address bits for each xPTH (High word DMA pointer) register when written. (Even internal Agnus register accesses are visible in RGA bus which enables storing copper writes)

When Agnus does DMA, FPGA would match RGA bus value with stored DMA pointer value and set chip ram extra 2 address bits to select correct 2M chip ram "bank".

(DMA crossing banks needs extra care but it should be simple enough to handle by comparing if previous same DMA channel's memory address value was very large and new very small or vice versa)

I think this should work, assuming there is time to snoop RGA bus and select correct extra chip ram "bank" before Agnus does the DMA transfer.

[xyzzy1]

If you have to resort to using an FPGA then I think a better strategy would be to make a small board with FPGA and high-speed SRAM on board and add a port to allow an accelerator card or CPU wedge to access our new 8MB chip ram at full speed. The FPGA would arbitrate between chipset and cpu accesses.

[AJCopland]

Well since there's the ReAmiga 1200 (http://eab.abime.net/showthread.php?t=93836) and Amiga 1200+ (http://www.amigaclub.be/projects/amiga1200plus) replacement motherboards how about recreating the ChipRam read performance feat implemented by Index Information on the Access (http://amiga.resource.cx/mod/access.html) which: "Access incorporates an improved Chip RAM interface which couples it directly to the processor, giving twice as fast memory read performance as of a standard A1200."

Maybe getting 4MB of ChipRam isn't easy. Although Toni I do like your idea of using a FPGA/CPLD to duplicate the functionality.

[imi2003]

Just a quick question regarding the 8372a Agnus, I'm correct in assuming that this Agnus has 20bit pointers internally as it can access 1mb ram. Since the 512kb versions have 18bit pointers and there are no references to a 19bit version.

If so would the 8372a be "easier" to add the missing signal lines so it could address the full 2mb.
Maybe this would be a good starting point in implementing a full replacement.