4MB Upgrade & 16MHz Booster progress

[qq1975b]

Here you can see a Mega ST motherboard: http://www.atari-wiki.com/images/9/9b/Mega_4.jpg

All revisions are the same (only TOS configuration may change between 2 and 6 chips). Maybe your design is already compatible with it? Just if it doesn't affect the Mega bus it would be Ok. The Mega bus cards face always back to the Mega ST.

Probably the blitter patch have to be taken out...

[joska]

IDE ports is interesting, but there are already kits for that, this project would need new software and hardware to work with fast-ram.

Not sure what you mean by that. I would think that when you already have a CPLD you'd only need an IDE connector and a buffer/driver to add IDE? Or maybe you need a bigger CPLD for that?

Flash ROM is possible, but also have to remember each time something is "added" cost will be going up. The ROM is in a PLCC, so anyone can program a new chip with what ever they want anyway.

Again, not sure if I understand. You're already adding a "big" EPROM and logic to switch "banks" and support 2.06. You already have a CPLD for the necessary logic. What else do you need to use flash EEPROM? Remember that most users don't have EPROM programmers. I think it would be very useful to have this feature, e.g. to put MagiC in flash.

[exxos]

Here you can see a Mega ST motherboard: http://www.atari-wiki.com/images/9/9b/Mega_4.jpg

All revisions are the same (only TOS configuration may change between 2 and 6 chips). Maybe your design is already compatible with it? Just if it doesn't affect the Mega bus it would be Ok. The Mega bus cards face always back to the Mega ST.

Probably the blitter patch have to be taken out...

Looks like there is more space on that than the STFM which is good. The blitter patch won't be a problem as the blitter clock as to be cut to install the mod anyway. Though it won't effect the code as machines without blitter TOS will not access blitter registers anyway, so the code won't do anything.

[exxos]

IDE ports is interesting, but there are already kits for that, this project would need new software and hardware to work with fast-ram.

Not sure what you mean by that. I would think that when you already have a CPLD you'd only need an IDE connector and a buffer/driver to add IDE? Or maybe you need a bigger CPLD for that?

Flash ROM is possible, but also have to remember each time something is "added" cost will be going up. The ROM is in a PLCC, so anyone can program a new chip with what ever they want anyway.

Again, not sure if I understand. You're already adding a "big" EPROM and logic to switch "banks" and support 2.06. You already have a CPLD for the necessary logic. What else do you need to use flash EEPROM? Remember that most users don't have EPROM programmers. I think it would be very useful to have this feature, e.g. to put MagiC in flash.

There isn't a problem of adding stuff to ports, we have around 80 ports to play with, around 50 are spoken for with current stuff. What I wanted was the IDE to have the option to load and run in fast ram. IDE is normally directed to ST RAM which is slower. We can't just copy paste code into this project to add IDE, and IDE will have to support fast ram and higher speeds also. This will probably need special software and drivers to work in fast ram. We could add normal IDE on there, but again, where will the software come from. With so many hard drive which work on the DMA anyway, IDE isn't as fast. To make IDE worthwhile it needs to operate at faster speeds with options to use fast-ram. Which I can see will be a lot of work. I agree it could be useful to some people to have IDE on there. Myself and rodolphe talked briefly about adding a flash drive on there a couple weeks ago. Though its also adding to final production cost. Same for alt-ram with costs, the best deal we could find was to add 4MB fast-ram. chip cost for example £20. 8MB isn't as good value, £50-£60. We could use 2 chips ( instead of a single chip), but this takes up more space, which we do not have. If you think £50-£60 for just the 8MB fast ram IC, then add all other costs, we are already at a kit which would cost over £100 to produce.

[joska]

There isn't a problem of adding stuff to ports, we have around 80 ports to play with, around 50 are spoken for with current stuff. What I wanted was the IDE to have the option to load and run in fast ram. IDE is normally directed to ST RAM which is slower. We can't just copy paste code into this project to add IDE, and IDE will have to support fast ram and higher speeds also.

I did not realize that you're going to add fast-RAM to this too. That's very good news Even 4Mb would help a lot. So we're looking at 16MHz CPU, with 4Mb 16MHz fast-RAM and a 512 (?) Kb ROM?

I agree with you that it would be nice to have faster IDE, but that's not necessary. Even the standard Falcon/STBook IDE interface is as fast as ACSI on a plain ST. Having a plain IDE port on this card would be more cost efficient and simpler than buying an additional SD/CF storage adapter/device. E.g. I was considering buying a £50 SCSI2SD card for my Mega (it already has an internal SCSI adapter) but would rather spend a little bit more on your accelerator instead.

Not so sure if you need a special driver to take advantage of fast-RAM either. AFAICS the standard drivers would work without changes even if the IDE registers are accessed at 16MHz. Maybe the only required change would be to put the driver and read/write-buffers in fast-RAM. Uwe Seimet seems to be willing to support new hardware - look at HD Driver 9 with support for 8 devices with Alan's IDE interface.

[exxos]

The goal for the 68000 card is

Switchable TOS104, TOS206 ROM,
Switchable 8/16mhz CPU/TOS/BLITTER
4/8MB Fast 16mhz Alt-ram.
Optional blitter socket.

The alt-ram comes down to cost, 4MB is more cost effective, but 8MB can be used if someone wants to fork out the higher costs. Of course there isn't a "must" to have any alt-ram. The card will work fine without it.

I have talked to Rodolphe about other things, but we have decided to keep focused on making the machine run faster than adding generic "features". We both agree it stuff can be added at some point, but each thing we add is going to ramp up cost and delay the release of the project. If we keep adding things, then it probably will never get finished.

What I am still working on is breaking the 16mhz barrier. Though this is mainly down to the CPU not being able to run any faster. Though we are thinking of adding a PLCC socket for the PLCC 68000 as those are listed as 20mhz. I'm not saying any CPU will run faster than 16mhz, but we can try.

Also we have decided to add the blitter on the board also. We cannot supply the blitter, but anyone who has a spare blitter has the option to add it to the card. This I think is good idea, as the blitter clock would have to be cut and re-wired. With so many motherboard revisions it would be a total nightmare to hunt down and document where the pcb tracks are to be cut and re-wired. so overall its a lot easier just to put the socket on the booster card.

Long term, we hope to speed up other areas of the ST, like said before, I want to try and break apart the ST-ram circuit and speed that up ultimately. Though its a lot of work and probably may never happen. Once this current 68000 booster is completed, we are going to mainly focus on creating a 020 booster card. We think 32mhz is possible with 32bit ROM& RAM access. We really want to get started with the 020 project in the new year.

As a side note, my simple 16mhz booster pcbs , I am going to try and start building them this week and test them out. They will be put on sale in my store once tested. They will be reasonably cheap. I hope to generate some capital from those sales to help fund the current booster projects.

[GadgetUK164]

Totally amazing! I really can't wait for the finished board =) Just the 16Mhz with switchable TOS, 4 or 8Mb of Alt-RAM and optional Blitter socket is like a dream come true lol! The 20Mhz chip sounds like a great idea but I do worry about the compatibility with that mod (not the clock speed per se, but the difference between the processors - doesnt the 020 have some major differences?). It's easy for people to say that the 16Mhz will break stuff, but to be honest I've found very little negatively affected by 16Mhz clock speed. So far there are like 2 games out of around 300 or so which have issues. Yes the raster stuff is affected on PP's HDD adapation title screen images etc, but at least 99% of stuff I've tested and played works fine besides that minor title glitch. Compatibility might drop a bit with fast TOS access, but if its possible to disable fast TOS access that brings the best of both fast GEM apps and smoother gaming experience. I think the big problem with an 020 processor is that a lot of the community is in it for the gaming angle, and that could limit the popularity of the 020 mod. I personally would consider buying another STFM / STe just for the purpose of that mod, but stick with the 16Mhz for my main gaming machine.

Great job so far - Really appreciate the huge effort you both (Exxos & Rodolphe) put into this to get this far!!! I am loving my ST since I first applied Exxos's 16Mhz mod - It just makes the machine feel much snappier and gives a better all round experience, and with the fast TOS and Alt-RAM this can only get better imo. The way you plan to get this initial board working with all the aforementioned features will mean its pretty much 'the mod' to apply to an ST in order to get the best of everything.

[exxos]

If you remember the boost in speed from 8mhz to 16mhz, then imagine that boost again for fast TOS and you get a rough idea the 020 is 99% ish compatible. I think there was some instructions missing from it, though be what I have read, not much software used it anyway (I guess its why motorola took it out). Though don't know much about actual instructions personally. The 020 has a instruction cache so some things get a boost in speed there. The 020 is more cycle efficient than the 68000. It takes less clock cycles to run the same instructions. So that in itself is another speed boost. Also its 32bit, so TOS and alt-ram actually run at twice the speed. Also , if that wasn't enough, there are 25mhz versions of that CPU. I am pretty sure we can clock that to 32mhz. So the the CPU/TOS/RAM will run at double speed again. Running apps in fast-ram is even faster! Its basically very similar to the Veloce 020 booster. Though we are forced to use TOS206 for that. TOS206 in itself will break lots of stuff, so the stuff the CPU breaks i suspect isn't as much in comparison.

Rodolphe really wants to get started on the 020 board, so hes busy finishing off the 68000 board so we can get some of those made hopefully before xmas. Once the 68000 board is finished I want to try and boost other things on the ST. So its not going to be the end of the 68000 progress once these boards are done. Though the 020 card will take priority overall

[rpineau]

Regarding the 68020 there is actually more instruction that on the 68000 one of the big change from the 68000 to the 68020 instruction is move from SR, it use to be unprivileged on the 68000 but it now requires supervisor access (which is part of what TOS 2.06 fixes). That is the only non compatible instruction of the 68000 instruction set on the 68020. This of course can break a lot of games.

And as Chris said .. all 32bit for TOS and FastRam with possibility to run at 32MHz (we'll probably try at 16 first ).
As for the game compatibility ... probably going to break a lot of them like the TT and Falcon did. I think the 68020 card is more for user wanting to play with MiNT and Gem app than for people playing standard ST games.

Rodolphe

[Cyprian]

why not 68030? it's much better than 020. e.g. it has data cache which would compensate a slow access to ST-Ram.

[rpineau]

why not 68030? it's much better than 020. e.g. it has data cache which would compensate a slow access to ST-Ram.

Because it's a lot harder to adapt a 68030 to a 68000 bus than a 68020 , and because I'm not good enough to do it. Also you'll need a TOS 3.x with a lot of patches so that it doesn't try to access all the TT stuff that aren't present in a ST.

Rodolphe

[exxos]

why not 68030? it's much better than 020. e.g. it has data cache which would compensate a slow access to ST-Ram.

"Shockwave" on here has project "re-pak" So the 030 and up CPUs have already been done. Though I sent him a message weeks ago, but he seems to have vanished. So as the 030 has already been worked on, there is no use in use starting the same project.

Originally I did the simple 16mhz booster, simply due to it being low cost, and did not break much software. Though I always wanted to speed up TOS as well. As This will speed up GEM which it really needs. Speeding up GEM shouldn't break anything either. So when Rodolphe offered to help, we got the ROM speed up done a lot faster and recently adding the blitter and alt-ram. The goal has always been not to break software. Moving to the 020 will break a lot of things, but that is more a TOS issue.

My STE has the veloce 020, and not many games will work on it. Though its really madness to run a 020 with some games. For example, I got Vroom working once, though as soon as the game starts, you are in a tree! The game is unplayable since in under 1 second your are crashed into a tree. So while it breaks a lot of games, its probably a lot of games simply wouldn't be playable at those speeds anyway. The 020 is a fantastic upgrade but each change in the spec breaks things. This is why I wanted to keep the 68000 CPU and just speed the ST up as much as possible without breaking much software. There is as always cost, as a factor as well. Nobody really is going to spend £200-£300 on a addon card for the ST. More likely cheaper solutions will be more popular.


Talking of cheap, the simple 16mhz mod page should be up later today. It is probable I will supply the CPU with the kit fitted on top ready to solder in. I could sell the thing as a kit, but most do not like SMT soldering, and as there is only 3 ICs I think there is little point in selling the parts for this. I will be taking pre-orders later also for this kit. Probable price will be £25 which includes the kit assembled and the 16mhz CPU. This kit only fits DIP type CPU (mars bar type). Hopefully a few people will order this kit and it will help put some money into the pot to keep working on these projects.

pre.jpg

[joska]

the 020 is 99% ish compatible. I think there was some instructions missing from it, though be what I have read, not much software used it anyway (I guess its why motorola took it out).

IIRC there's some differences to the stack frames too.

[joska]

I think this is a great project. I will probably buy the 16MHz + fast-RAM card for my Mega, and upgrade to the 020-version when it's ready.

[exxos]

The simple 16mhz mods is up for pre-ordering. I did a separate post, but in case anyone missed it..

http://exxos.www.idnet.com/IMPULSE/atari/last/v1booster/index.htm
http://exxos.www.idnet.com/IMPULSE/atari/last/store.htm

Hopefully a few of those kits will sell quick as we need funds to get the next 16mhz mode pcb manufactured

[rpineau]

the 020 is 99% ish compatible. I think there was some instructions missing from it, though be what I have read, not much software used it anyway (I guess its why motorola took it out).

IIRC there's some differences to the stack frames too.

Yes you're right. I forgot about that.

Rodolphe

[exxos]

booster2.png

Rodolphe has been busy Though unfortunately we cannot route the SRAM chips on this board. It is already 4 layers and we still cannot route it. So rather than going to 6 layers we have decided to add a expansion bus to the card instead. This way, the alt-ram will just be a optional plug in board. I think having a expansion bus is a better idea as it allows future expansions to be done. Of course if anyone wants to develop their own expansion kit to work with 16mhz CPU then they are free to do so

[exxos]

I am building up the kits today. So should be able to start shipping in a few days

I have 18 kits built but only about 12 16mhz CPUs left. So once my stock of CPU's has gone, people will have to take a gamble and source them elsewhere. I did get about 50 pcbs made so I can make up more kits in the future if there is demand. Though again, any future batches will be supplied without the CPU, and of course the pcb only price will be cheaper.

[GadgetUK164]

I know I've mentioned it before but have you looked at the MACH 16 stuff on this site?

http://atari4ever.free.fr/

It basically contains all the GAL code for a 16Mhz mod plus the cache. It would be really nice to have a cache feature as well as fast TOS. If we could get a board designed vs the schematics it should be easy to build one of these ourselves.

It also shows performance there, and there's a math coprocessor socket oO

[GadgetUK164]

booster2.png

Rodolphe has been busy Though unfortunately we cannot route the SRAM chips on this board. It is already 4 layers and we still cannot route it. So rather than going to 6 layers we have decided to add a expansion bus to the card instead. This way, the alt-ram will just be a optional plug in board. I think having a expansion bus is a better idea as it allows future expansions to be done. Of course if anyone wants to develop their own expansion kit to work with 16mhz CPU then they are free to do so

That's awesome! I love the idea of expansion RAM - take my money lol! I don't suppose its possible to squeeze a math co processor onto the RAM expansion board as well? I want my cake and I want to eat it, and it has to have extra chocolate pieces and whipped cream lol

[exxos]

I know I've mentioned it before but have you looked at the MACH 16 stuff on this site?

Thanks for the reminder, I had totally forgot about that.

[exxos]

That's awesome! I love the idea of expansion RAM - take my money lol! I don't suppose its possible to squeeze a math co processor onto the RAM expansion board as well? I want my cake and I want to eat it, and it has to have extra chocolate pieces and whipped cream lol

It would be 8MB alt-ram, but I am calling it fast-ram as it will be running at 16mhz.

I think there was some discussion about the FPU before somewhere. Its not a problem to add it, but really, what program use it ? Simply putting on on the card will do nothing whatsoever. You could treat it like a intel cpu command set, where/when "MMX" came out, it made some new instructions faster for doing some things, but unless a program was written to take advantage of MMX instructions it was pretty much useless. Similar with the FPU on the Atari. I think some CAD related software would use it, but I doubt any software out there would likely even be used on a Atari anymore. I don't think GEM uses it either, but not 100% sure. Some compilers like devpac might use it for compiling code. Can't really think of any other software which would possibly use it.

So its not really a matter of "can we add it" its more a matter of "is there any point".

[joska]

I had a 68881 in my Mega STE about 20 years ago and it was of very little use. POV-ray would render a picture in hours instead of days, but that's it.

[exxos]

I had a 68881 in my Mega STE about 20 years ago and it was of very little use. POV-ray would render a picture in hours instead of days, but that's it.

I forgot about POV-ray, Its for crunching data fast that's for sure. I wonder if anything like STzip would use the FPU ? I don't recall it doing so, but its been a while since I looked at it...

[DarkLord]

If I remember correctly, at least one of the more well known picture viewers took
advantage of an FPU?

Can't remember the name right off hand...