32-bit FAST RAM vs. ZII FAST RAM

[polbit]

Well, I have another Amiga question, believe it or not!

A2000, with GVP '030/50. I currently have 12Mb 32-bit RAM on board, and 1Mb Chip RAM. I will be installing CyberVision 64/3D in it next week, which comes with 4MB RAM. My understanding is that this 4Mb gets mapped into the 24-bit memory address space, right? Now, how does that work exactly? Is this RAM usable by general programs?

Also, if I add another 4Mb fast ram to max out the 24-bit address space, will I make my Amiga slower because not all Fast RAM is 32-bit?

PolBit

[alexh]

The graphics card memory is not "normally" usable by general programs. It is for RTG frame buffers only.

It depends on the RAM card but adding an extra 4 meg of Z2 fast ram in a system like this is not recommended. It accomplishes nothing!

[polbit]

Ok, so the memory on the graphics card is protected memory, but it is mapped into the 8mb space, right? And the rest of Fast RAM is then used for image manipulation, etc. so I have to assume that some of my Fast RAM will be taken over by the system for the card, right?

Sorry, these might be basic questions to some of you, but I haven't used an Amiga in a very long time, and I am just trying to understand the limits and problems of 24-bit address space vs. 32-bit address space vs. chip ram.

I have seen a lot of set ups in both older magazines as well as on the internet where people have an accelerator board with 32-bit ram in their 2000, yet still have some old board with 8mb ram, or the 2091 with 2mb ram, etc.

[alewis]

Correct, the card is mapped into the Zorro II 8meg autoconfig space, and takes 4mb. Your expansion choices are then somewhat limited.

1. Add 4mb (max) of 16bit fastram. As Alexh says, why? Are you using anything that needs >12mb? The 16bit memory will always be slower. The only issue that *might* occur (that I can think of) is the C64 doesn't like using the Impact030 32bit fastmem. Given the C64 recommended min spec calls for a B2000 with 68020, this shouldnt be an issue.

2, Add 2mb, just in case, and hope that any future cards only require 64kb entries, or at a max, 2mb.

3. Do nowt, and leave the remaining autoconfig space free, just in case.

An no, the C64's video memory is not accessible by AmigaDOS, so cant be used.

Do remember that the display driver keeps a copy of the screen in fastmem, which uses memory. More so if you use certain options.

Regards fitting, stick it in the slot nearest the Impact030 / PSU. That way it is recognised first (just in case you have multiple cards already). And dont forget to set BOOTLOGO=YES in devs/monitors/cvision3d tooltype. Oh, and set the environment variable ALERTEMU=1 if you want guru's on the RTG screens. Also check that the cgxvideo.library and cgxsystem.library are the 030 optimised versions...

enjoy, its not a bad card.

[Zetr0]

I have the same card (CV64/3D Mk II), i have admit its an awsome card.

I even had fun with 3d texture mapping with it... (under warp3d)

an RTG workbench is most certainly the way to go ... very nice indeedy

[polbit]

Thanks for the explanation, that clears up a lot!

I can't wait to get the card, so glad Software Hut still had the C64/3D new, and pretty cheap... Initially I wanted to get the PIV I saw on eBay, but the price is a little crazy - 4 days to go and $236 already...

Now if I can find a 3000T and the scan doubler module for the C64/3D, I would be in Amiga heaven... But I should be quite happy with the 2000 for a while, just wish it had ZIII slots.

[polbit]

Ok, I need to vent a little... I hate Buy It Now on eBay, 'cause it makes me buy stuff that my wife can easly kill me for later! Just got an A4000/040 for too much money, now I gotta sell something... Time to offload my 8-bit collection I guess.... This past week I think I spent more than $500 on Amiga stuff

But, now I will have a nice ZIII Amiga

[alexh]

Originally Posted by polbit

Now if I can find a 3000T and the scan doubler module for the C64/3D, I would be in Amiga heaven...

The 3000(T) has a built-in scandoubler, all you need is a monitor with dual inputs or a monitor switcher.

[polbit]

I know, just figured for the ease of use and display quality, the C64 scan doubler/flicker fixer is better. Then again, there is always the PIV with its programmable scan doubler/flicker fixer...

[alexh]

I know, just figured for the ease of use and display quality, the C64 scan doubler/flicker fixer is better

Not true, the AMBER is as capable as the C64 SD/FF. But congratulations you are now officially an Amiga freak.

You have proved that without thought or reason you are willing to spend money on something that will add nothing to your Amiga experience what so ever.

You have that rare quality that I will never own and for that sir I salute you

[patrik]

Originally Posted by alexh

The 3000(T) has a built-in scandoubler, all you need is a monitor with dual inputs or a monitor switcher.

Or a CyberVision64 card - the original and better non-3d version, which does the switching for you automatically .

[polbit]

Hey now, I didn't actually buy the C64 sd/ff, I was just talking about it. I thought that the module outputs the signal through the same connector as the C64/3D board, but looking at it closer, the module has its own connector, so I'm taking it all back!

And getting a 3000T these days is pretty much impossible anyway, unless you are willing to spend crazy money.

[alewis]

The C64/3D sc-ff switches automatically. You fit the 15pin DSUB connector into the blanking plate on the C64/3Ds rear plate, unplug your monitor from the C64's DSUB, and plug it into the DSUB you just attached.

When the screen displays a non-RTG (standard Amiga) screen, the module displays that. When it displays an RTG screen, the module automtically switches the displayed output. Think of it as an auto monitor switcher built in to the SD-FF module as well. Extremely useful, and something that I hope a certain owner of an A4K tower system will appreciate in a months time :-)

[alewis]

The memory address is mapped into the autoconfig address space.

And yes, some fastmem is used to hold display images (so it can be passed to the card memory).

32bit memory on an accelerator does not - generally, it depends on the accelerator and/or the means in which it can be configured - map into the 8mb autoconfig space. It is wholly seperate. Which is how some accelerators can have >8mb. because it sits outside the 8mb autoconfig space, access to it is handled by the accelerator card. In addition, as it sits outside, the autoconfig address space is availabe to the system, for other expansion cards/memory.

Look at the Impact030 manual, there may be ajumper to configure the 23bit memory as "extended" memory (outside the autoconfig space), or as "slow"[er] memory within the autoconfig space, subject to a max of 8mb, either in total or in addition to any other expansion cards and memory. Certainly the G-Force030 could do this, and I wired a switch to the jumper so before powering on my B2000 I could switch the 030 on/off, and configure the 32bit RAM as fast "extended" memory, or "slower" fastram (slower in the sense it was considered 16bit RAM, and accessed at standard Amiga memory access timings).

Perhaps "slower" is a misnomer - it was still faster than chipram, or the so-called "slowmem", but I cant recall whether it was any slower than a.n.other fastmem present in the system.

Originally Posted by polbit

Ok, so the memory on the graphics card is protected memory, but it is mapped into the 8mb space, right? And the rest of Fast RAM is then used for image manipulation, etc. so I have to assume that some of my Fast RAM will be taken over by the system for the card, right?

I have seen a lot of set ups in both older magazines as well as on the internet where people have an accelerator board with 32-bit ram in their 2000, yet still have some old board with 8mb ram, or the 2091 with 2mb ram, etc.

[Secret Vampire]

My understanding was that any 32 bit RAM on an accelerator would be faster regardless of whether it was mapped into ZII autoconfig space or not.
The above should hold true unless the designer of the accelerator performed some tricks to make the RAM accessible to the 68000, that is, made it actual 16 bit RAM.
In most (all?) cases though, the RAM is 32 bit, and accessed at whatever timings the accelerator itself uses, regardless of which bit of address space it uses.

There are special cases too, with regards to having 32 bit and 16 bit in a system, Zorro II DMA cards tend to have trouble doing DMA to 32 bit RAM, what with being unable to see anything outside ZII address space and all, so if you need more than 8MB and want to use a DMA card such as the A2091, you can get better performance by having a bit of 16 bit fast
Whether this would be applicable to the CV643D or not is another matter, I'm not sure if it uses DMA to transfer data between the buffer in fast ram and its own framebuffer, but if it does it may well help to have 4MB of 16 bit fast and force the RTG software to use that memory.

Incidentally, if Cybergraphics and Picasso96 use fast RAM as a buffer, why is it so hard to produce drivers for the few cards that use a 64k window like the Retina ZII and TIGA board? and why couldn't Macrosystem make RTG software that used fast RAM instead of chip to draw the display?

[alewis]

You could well be right SV, its been [too] many years. I just have a nagging memory that switching the 32bit memory into the autoconfig space resulted in [slighty] degraded performance.

re the DMA to 32-bit memory, istr a *few* packages/adverts/write-ups that specifically highlighted the ability of *their* product to do this. Again, this could have been the blurb concerning the SCSI controller of *accelerator* cards (such as the G-Force)... been too many years!

[Zetr0]

@Secret Vampire,

I am not sure... its been a very long time.... but wasn`t there a fastmem hack on aminet for the Macrosystem... i think it was called something like ForceFast or somthing like that....

reason i remeber it is because a friend had a Retina i believe and he was always complaing that the ForceFast (or FastForce) had an issue with 060.. ( i only had an 040 at the time so i fear it was more of a drop to me than him caring about this program. mind you when i got the CV3D64Mk2 he got his deserves )

it may of been for somthing else.... its been a while.... pretty sure it was for a retina graphics card though....