Where do you draw the line ?

[exxos]

I guess if you ever intended to support the 64bit Apollo 68080 core (or perhaps some other hacked core with LL op sizes), the redesigned ancillary chips would need to be FPGA anyway

It is unlikely that will be used because it is a commercial product and when I asked them all years ago they basically laughed at me and said they would never support Atari machines. It was via email and on their forum. But they deleted the forum thread. I think some developers are working with them but to be honest they don't seem to have a good reputation with people.

Tuxie (lightning team) when I was talking to him years ago, I believe he was going to obtain to board so we could adapt it to Atari. I never saw a board, I know Tuxie had one, as far as I know, he gave up trying to get it to work on the Atari in the end.

I would much rather have things like what @dad664npc is working on as a alternative CPU. I do like the appeal of the regular 030 CPU's etc, but they wouldn't likely be attractive to newcomers to the scene.

I would like to see a proper 32bit bus on the FPGA system eventually. I don't know how difficult that would be. But anything is possible if we can create our own silicon gates.

If the ST line was still going, Atari (or whoever) would have long ago moved towards using more custom chips. They'd have followed the trend and had a 'bigger' bus for more cards - graphics/sound etc. Like you're doing with the H5 anyway

Atari of course, integrated the BLITTER,GLUE,MMU into the STE GST chip. I think natural progression would have been the shifter then the DMA chip. It sort of went this way with the Falcon anyway.

While graphics cards can improve things they are basically unobtainable now. But the thing is, if you have a faster bus, you can push data to a shifter a lot faster anyway. You have already opened doors very easily for higher resolutions. But of course the downside of it all, is that without developers making use of the extra functionality then it somewhat becomes pointless doing too much.

I've always gone with the philosophy that the biggest software library is with the ST. So any Atari machine compatible to run the software. Of course demos are a big part of the ST culture. So I always want something which is going to be backwards compatible. This means it's going to have a 8MHz mode and a switchable 32MHz mode. A lot of games and software runs pretty rubbish at 8Mhz. So 32MHz can benefit the majority of software.. As long as they are not using time independent code that is. But thats why I want to keep a 8MHz mode for legacy software.

[exxos]

Looking later into the 90s they would probably have seen what's coming and offered an expansion bus, probably PCI, as many other non-Intel platforms started to do, and put graphics onto that rather than having them memory mapped, maybe with a co-processor talking to the graphics card to maintain backward games compatibility. They would have moved to a 32 bit data bus too.

I think what I'm saying is the line to be drawn is really where everything is simulated in one chip as a software defined architecture.

Yeah. It poses somewhat of a conundrum because if everything is emulated in FPGA like the MiST, the alternative is to not even bother and just grab some inexpensive PC mini motherboard and run Steem on it. It goes back to my point of when is a ST not an ST anymore.

I think part of the ST heritage is the ability to add upgrades and hack stuff in. People can get their hands dirty with real hardware.

[stephen_usher]

Looking later into the 90s they would probably have seen what's coming and offered an expansion bus, probably PCI, as many other non-Intel platforms started to do, and put graphics onto that rather than having them memory mapped, maybe with a co-processor talking to the graphics card to maintain backward games compatibility. They would have moved to a 32 bit data bus too.

I think what I'm saying is the line to be drawn is really where everything is simulated in one chip as a software defined architecture.

Yeah. It poses somewhat of a conundrum because if everything is emulated in FPGA like the MiST, the alternative is to not even bother and just grab some inexpensive PC mini motherboard and run Steem on it. It goes back to my point of when is a ST not an ST anymore.

I think part of the ST heritage is the ability to add upgrades and hack stuff in. People can get their hands dirty with real hardware.

Indeed. An ST derived system with access to a PCI bus would be highly useful. There are still a lot of PCI peripherals which could be made to work with the correct drivers. One solution for this would be maybe an embedded Pi Pico SoC on the board to handle the arbitration etc. and address decode (so that the PCI bus would be a relatively small memory aperture in the ST memory map) and have the 32 bit data bus link into the PCI bus. That should simplify things and not take the system too far away from the original architecture.

As for the main bus speed, that would be the same as the CPU but the memory could be clocked at double that and have the graphics sub-system access the memory on the odd clock, so as not to slow the CPU.

Would a PiStorm be too far as a CPU? Probably not if it uses the m68k socket, but that restricts the system to a 16 bit data bus.

[exxos]

Indeed. An ST derived system with access to a PCI bus would be highly useful. There are still a lot of PCI peripherals which could be made to work with the correct drivers. One solution for this would be maybe an embedded Pi Pico SoC on the board to handle the arbitration etc. and address decode (so that the PCI bus would be a relatively small memory aperture in the ST memory map) and have the 32 bit data bus link into the PCI bus. That should simplify things and not take the system too far away from the original architecture.

There is a lot of PCI still for all sorts yes. The problem is nobody is ever going to write drivers for it. I would think someone would have done it by now if I was going to. It is a battle to even get a LAN solution going on the 68K bus. It needs hardware and software developers working on it. But with PCI, it probably needs to be more like a Milan based machine. I did inquire about the possibility of recreating it years ago but while he replied to some questions, he didn't really respond to the idea of making them again. I think mostly he lost motivation because nobody was interested in the project.

As for the main bus speed, that would be the same as the CPU but the memory could be clocked at double that and have the graphics sub-system access the memory on the odd clock, so as not to slow the CPU.

Its possible it could be adaptable split. I mean if you are only running low resolution then you could allocate something like 75% bandwidth to RAM from the CPU.

32MHz is really just an arbitrary figure. Mostly based on the fact that the 68K maxes out at that speed. It is possible to increase to 64MHz with the SEC booster CPU. It will of course need SRAM (expensive) to keep up.

My sub- goal was always to max out the 68000 as much as possible. Then start to think about 32bit. But even getting a 32MHz system working in FPGA going to be a lot of work in the monumental achievement I think. Really the platform needs a lot more interest before I seriously consider the next step afterwards.

Would a PiStorm be too far as a CPU? Probably not if it uses the m68k socket, but that restricts the system to a 16 bit data bus.

My philosophy has always been to use a real 68K chip. Though things are not that simple because if we want faster or better CPU like the PiStorm, then it's going to have to move away from real silicon. I think giving people the choice really covers everything anyway. Of course the CPU's are not going to be around forever. So ultimately they will have to be some alternative.

From a pure cost standpoint. We were sort of end up back to the MiST again. Integrate everything into a single FPGA. With or without legacy ports is another issue. Because if newcomers just want to play the games they are probably not going to be bothered about legacy hardware anyway. But I don't think there is any point building such a system because again the MiST exists.

[elterwater]

I'm going to come at this from a different direction.

As someone reliving my youth (all those hazy decades ago), I'm enjoying being able to explore bits of hardware I could never hope to have owned as a child (ACSI2STM) as well as fixing and upgrading bits and bobs on my STEs (gotek, VGA output etc). I've ripped out the PSU in one of them and replaced it with a picopsu because... well I wanted to see if I could make it work.

I can satisfy my curiosity and explore TTs and Falcon-like experiences with Hatari to my heart's content, but I quickly drop back to STE-level configuration. Why? Well, after booting into those desktops I very quickly ask myself "So that's what it would have looked like! Ok, what do I do with this now?" and almost as quickly realise that there's not much I really want to do with those comparatively more powerful configurations.

So let's say that a lot of time, expense and development is expended and we get to the point of new motherboard infrastructure that's on a par with a MiST or a Pi solution. Great! But what are we, as enthusiasts going to do with it and more importantly, are there enough of us left who are going to pay for it over something like a MiST?

For me, the line would be where if we had some sort of daughterboard solution which could be placed within the case of an ST that could slowly replicate the functions of some of these hard-to-find components but still retain as much as possible of the original machine. The other side of the line is the daughter board is just a junction box for the different legacy ports soldered onto the motherboard.

Once we cross that line, we may as well forgo goteks, ACSI2STM, real TOS etc and just be running on a Pi with emulated hdd :)

[exxos]

I can satisfy my curiosity and explore TTs and Falcon-like experiences with Hatari to my heart's content, but I quickly drop back to STE-level configuration. Why? Well, after booting into those desktops I very quickly ask myself "So that's what it would have looked like! Ok, what do I do with this now?" and almost as quickly realise that there's not much I really want to do with those comparatively more powerful configurations.

So let's say that a lot of time, expense and development is expended and we get to the point of new motherboard infrastructure that's on a par with a MiST or a Pi solution. Great! But what are we, as enthusiasts going to do with it and more importantly, are there enough of us left who are going to pay for it over something like a MiST?

It is indeed a very good point. I mean sure, everyone wants better resolutions and more colours to make it more Falcon like. But then what ? If there is not software to make use of those features then they somewhat become irrelevant to a point. Indeed spending massive amounts of time and resources on developing things is something I don't really have time to do personally.

Though I think as things get upgraded one by one, are we then not basically getting towards building a PC anyway with higher resolutions and colours etc ? I sometimes feel it is going a little bit against the charm of the humble ST at times.

Mostly what I have been keeping in mind is to keep compatibility with legacy software and games. A lot of games can make use of more CPU power in particular 3D games. Developers using real hardware can make use of a faster CPU ( and possibly higher resolutions) for those people are basically the minority these days.

What I would really like to see is demo coders doing some demos to make use of the higher CPU speeds etc the problem being, it is unlikely many of them will be able to afford a complete custom-built for them. But then of course, if there are not many machines running 32MHz, they not going to spend the time programming stuff anyway IMO.

For me, the line would be where if we had some sort of daughterboard solution which could be placed within the case of an ST that could slowly replicate the functions of some of these hard-to-find components but still retain as much as possible of the original machine. The other side of the line is the daughter board is just a junction box for the different legacy ports soldered onto the motherboard.

The FPGA blitter is cycle accurate and tested on real machines. It is my intention that they could be used because blitters are impossible to find. FPGA prices are very expensive currently, it sort of isn't really practical at the moment to do individual chips. But indeed these are actually being worked on. It is becomes more cost-effective to integrate a few chips into a single FPGA.

I think there is going to be a very fine line between cost vs features here. We need to encourage more people are more coders onto the platform. Though this is only going to stand any chance if the platform is affordable to the masses. Which is very problematic when you're dealing with low quantities, parts shortages, large PCBs etc .

[mrbombermillzy]

I think the way to go is to follow the biggest available software base. (The 32mhz bus 32/64mhz CPU speed increase will encourage enhanced SW writing, but an existing dev/tool/utility/games SW base that can be fallen back to - or can be used with the system boosted - is IMHO needed initially.)

To explain:

If we are going to start looking at FPGA use to some extent, I would suggest instead of creating new GFX/Sound devices, the FPGA has something like Videl/ET4000/Mach32 for graphics (instant GEM/TOS software compatibility to lots of already written software from the get go, as there are drivers or software already written for these).

Same with sound; ST/TT compatible DMA sound system. At most, maybe a STFM/Ste software version 'splice' in the sound routines department is needed.

A bit more exotic, would be the addition of the 560001DSP, as long as it can resemble the memory mapping layout of the F030, or maybe a version mapped like the Startrak MegaBus DSP card, so we can use the appropriate MegaST sound software. Not sure if this can really be done, but considering the Sparrow/FX were boards put on top of an STe, it may be possible, with the right TOS ROMS obviously.

Now to add to this, the H5 board can be 'cost reduced'; i.e. shrunk to resemble a TF/Vampire v2 card which is wholly 68000 socket based.

The core can use (by easy means of a re-flash when needed) as much of or as little of the original STFM components as necessary, providing the output ports are actually on the daughterboard and enabled.

In that way, if the users blitter chip fails, just flash the core to one that has blitter onboard the FPGA. Real shifter broken? Plug in to the (e.g.) HDMI socket on the daughterboard and use the FPGA cores shifter. Same goes with other ICs that have been accurately emulated, providing the faulty part doesnt cause bus errors/system lock ups. The user then has the choice of how much or little of their original ST is used, if there are enough core updates/versions to cater for this. It would be very handy for failing components.

[mikro]

In the end it is always what makes it fun for you. Personally, be it your H5 or Pavel's (hopefully one day finished) Falcon030NG board, my motivation for purchase is / will be always the same - the security of having a modern motherboard without all the faults and cheap workarounds so I can enjoy my hardware as long as possible. That's it, nothing more. If you add 32 MHz GST, easy bus for upgrades or in Falcon030NG's case an overlocked bus... that's fine but I could live without those.

If your thinking of H5 is something like a platform for future (hardware) development then of course one day you'll have to move away from the original chips and connectors.

If I were to judge what to do next with the H5: STE, STE, STE! :D That's the only reason why I'm still hesitating with purchasing (and all the follow-up work with...) the H5 because I don't even have my STFM powered on for years. Everything ST-related I do on the STE because these days there's so many cool productions to run on it...

[exxos]

It was always my intention to use the suska cores as they are STE. But I am looking to use @ijor's cores as they are more mature than suska and @ijor actively maintains them. As to if he can add STE features.. maybe. I want to add DMA playback from the STE, but not sure what else the STE really has to offer which is of some actual use ?

[chronicthehedgehog]

It was always my intention to use the suska cores as they are STE. But I am looking to use @ijor's cores as they are more mature than suska and @ijor actively maintains them. As to if he can add STE features.. maybe. I want to add DMA playback from the STE, but not sure what else the STE really has to offer which is of some actual use ?

Well I think ultimately it's games that draw people in so the hardware scrolling and expanded palette of the STe would be nice.

Pie in the sky thoughts... don't take me too seriously :D
If we are going to draw in new blood then it's got to level up with that other machine and perhaps run DREAD twice as fast :D
And then some. Basically I'd like a Falcon on the cheap perhaps also with Jag compatibility
Be nice if it could handle X68000 games with a little porting.
Obviously this could all be run on emulators but what I'm trying to imagine is some kind of refreshed 16/32 bit platform much like
what the Commander X16 and Spectrum Next are doing for 8 bit.

More seriously, I've been reading about people putting STs back in their music studios so that tight midi timing (afaik as I've never done any) that the ST was famous for is a must.