Here is something else to throw out there. If the converter supports Freesync, then it would work with every v-sync value up to the max refresh rate of the monitor... you would have to have a Freesync monitor as well but that is easy enough to do.

The converter would also need to either scale to a different resolution or do letterbox type output as HDMI and Displayport only support certain resolutions. No way around that.

Buy an Mister FPGA and call the day

A real hardware solution that meets these requirements is absurdly difficult and impractical. There are so, so many reasons why this wouldn't and couldn't work.

I doubt you are serious, but if you want to actually get somewhere emulation is the only real way to do it. You could build a cute little housing around an x86 or ARM based SOC, theoretically license a bunch of games, and then configure in software things like optional controller support and a nice UI. That would already be super expensive to implement and do properly, not to mention licensing games is a nightmare, but at least it could conceivably be done.

mothergoose729 wrote on 2020-12-10, 09:02:

an x86 ARM based SOC

A what now?

jmarsh wrote on 2020-12-10, 09:06:

mothergoose729 wrote on 2020-12-10, 09:02:

an x86 ARM based SOC

A what now?

😜

There is Raspberry PI and bare metal Dosbox. So this is more or less covered. Mister FPGA is ridiculously expensive.

The performance of DOSBox running on a Raspberry Pi is abysmal, it should not be considered a serious option.

Ozzuneoj wrote on 2020-12-10, 05:02:

*no commercial product can be sold running MS-DOS or Windows without support from Microsoft (or breaking the law) because the games rely on MS software to work... unless you use an MS-DOS alternative, and forego Windows... but you've just limited the system's usefulness to pre-1997 games, and there are likely to be more setup headaches for each game.

I am initially targeting DOS games only, which admittedly is 1000s of titles. You don't need Microsoft's permission to do this; aside from FreeDOS, there are other non-Microsoft versions of DOS. For really old titles, Microsoft made MS-DOS 2.0 open-source, but I don't see that as a realistic option on bare metal.

What about Windows 9x games? Getting Windows 98 SE licenses would cover the legality, although it would increase the price of the console. I think it's actually harder to choose hardware for a compelling Windows 9x gaming console, since some folks will want real 3dfx chips while others will want the best Direct 3D experience from Nvidia, some will demand Aureal Vortex 2 sound, while others swear by Sound Blaster Live! / Audigy...

Basically, if I do actually make this, and it does actually succeed, then there are ways to expand the game library to Windows 3.1 and Windows 9x and do so legally, but the engineering effort is non-trivial and the coverage will be incomplete at best.

cyclone3d wrote on 2020-12-10, 07:40:

Here is something else to throw out there. If the converter supports Freesync, then it would work with every v-sync value up to the max refresh rate of the monitor... you would have to have a Freesync monitor as well but that is easy enough to do.

The converter would also need to either scale to a different resolution or do letterbox type output as HDMI and Displayport only support certain resolutions. No way around that.

My initial plan is to use the Variable Refresh Rate support built into HDMI 2.1, although some HDMI 2.0 TVs can apparently get VRR support from a software update. As for scaling, cropping, letterboxing / pillar-boxing to make a 320x200, 320x240, or 640x480 game scale nicely to 720p or 1080p, you're right there is no way around that. I would really prefer to use an off-the-shelf HDMI transmitter IC or video processor IC that supports pixel bus input, VRR, and scaling, but with HDMI output being integrated into modern SoCs, it looks like the old vendors (Analog, Silicon Image, NXP) are getting out of that business. Maybe I'll find the right chip for the job, but I could eventually be forced to buy an IP core and put it into an FPGA, at which point the costs start really rising for this project and also going beyond my areas of expertise.

Almost anything is possible with enough effort (money+time), but you have to consider your target market and whether this could actually come close to paying for all that effort, let alone being profitable.

The target market would be people who:
1. Like DOS games enough to want to play them and be willing to either play a specific set of specially made DOS games that don't need tweaking, or are willing tweak things to make them work... and do it on a proprietary system that doesn't work like the MS-DOS they may have known... and again, where do they get the games, and who will help them sort out problems?

2. Aren't nostalgic\picky\old enough to care about anything other than the games themselves (they don't mind a completely made up modern interface for launching games, and they don't want the system itself to look, sound or feel like what they remember... because the system can at best resemble ONE computer, not all the varieties people had)

3. ARE nostalgic\picky enough to not want to run GoG versions or DOSbox (which does an amazingly good job and solves almost all of the issues you have posed, aside from running real hardware)

5. Will find the experience enough of an improvement over DOSbox to keep the product and review it well enough to make others want it (they have to be able to discern high-quality emulated OPL3 or MT32 music vs the real thing... and then say it's worth several hundred dollars)

So really, who is this for? There may be people who fit most of these, but that number is extremely low. The number that fit all of them will certainly not be high enough to ever pay for the work needed.

Again, the best chance of making this work is getting with GoG and making a system that is built to run their updated\emulated versions of games. Sadly, most people already have access to computers that can do that, so there is very little incentive to buy an expensive console for the same purpose. For a more authentic experience... buying old hardware and then reading this forum is the best chance people have at playing old PC games. And there's no way to turn that into a product.

If you want to be able to mass produce a DOS computer from new hardware you are going to have to completely engineer everything from top to bottom. There are no chips still being produced today that will give you a complete sound card or graphics cards. I don't believe it is possible to source an x86 platform that will have legacy support for things like IRQ assignment and DMA.

Even on period correct hardware you can expect to run into a number of compatibility issues.

Speed sensitive DOS games
https://www.vogonswiki.com/index.php/List_of_ … sensitive_games

VGA compatibility matrix:
https://gona.mactar.hu/DOS_TESTS/

And of course, DOS sound, which is too complicated to be condensed into a single list or matrix

You would likely need to develop your own sound card, perhaps your own graphics chip, your own drivers and TSRs and game patches. It would take years to develop, in all likelihood, and be ungodly expensive.

Even as a passion project just for yourself, this is a daunting amount of work with many, many technical and logistical challenges. And for what purpose? Other than "real hardware is cool" there are no benefits to the end user over emulation.

You should take a look at Vortex86 stuff. You basically get everything in an SOC design except a DOS compatible sound card and a 3d accelerator.

The CPU is also speed configurable. Not sure exactly how low they can go though. You would probably need to contact them.

http://www.vortex86.com/

In their current models, the Vortex86DX (800Mhz) and the Vortex86SX (300Mhz) are probably the models you would want to look at.

Both of those have PCI and ISA. The DX has FPU, the SX does not.

Both of those would need the VortexVGA chip as well as they don't have onboard video.

mothergoose729 wrote on 2020-12-11, 01:45:

You would likely need to develop your own sound card, perhaps your own graphics chip, your own drivers and TSRs and game patches. It would take years to develop, in all likelihood, and be ungodly expensive.

Even as a passion project just for yourself, this is a daunting amount of work with many, many technical and logistical challenges. And for what purpose? Other than "real hardware is cool" there are no benefits to the end user over emulation.

I have no intention of developing my own sound or graphics card, even if I were to just borrow from ao486. That would be insane. I'm going to pick just one sound chip and one video chip, with points given to solutions that don't require support chips or lots of discretes to function. I don't mind a solution that has a few incompatible games or games with compromises so long as I can cover most games for, let's say, PC/AT up to slow Pentium. (The CPU has to have configurable multiplier and cache settings for that to work, I know.)

Just getting the HDMI output at 70 Hz using variable refresh rate is an engineering challenge enough. If I can't find an off-the-shelf video IC for that, then this project becomes unfeasible for one person.

As you said, the point is, "real hardware is cool." If there's enough folks who think such a product would be compelling because it's real hardware, then this has value beyond idle daydreaming. If not, then okay, it was fun to think about.

cyclone3d wrote on 2020-12-11, 02:05:

You should take a look at Vortex86 stuff... In their current models, the Vortex86DX (800Mhz) and the Vortex86SX (300Mhz) are probably the models you would want to look at.

Indeed, I'm familiar with them. I own an EBox 3350-MX, which has a Vortex86 MX in it. My original post mentions the Vortex86 EX2. If you haven't looked at it, it's fascinating. It's not dual-core, it's dual-computer with shared peripherals and RAM. My "option 2" in the OP could use it; the secondary CPU would run Linux while the primary CPU runs DOS (because only the primary CPU has access to the ISA bus.)

The Vortex86 VGA is interesting because it's got embedded VRAM. That makes it much simpler to integrate and reduces risk. It's also a new part as opposed to recycled or NOS. Its DOS compatibility is a complete unknown. I would need to get the mini-PCIe board from x86duino, put that in a mini-PCIE to PCI-E adapter, put that in a PC that can still boot DOS, and then run a bunch of compatibility tests on it.

Completely missed that... very cool.

Does the EX2 support ISA DMA? It says no LPC bus so I'm not so sure it does. And the MX says no ISA.

I have an old SOC development board that has PCI and ISA slots. I forget what the SOC is. I'll have to take a look since I haven't messed with it in quite a while.

It is like a fast 486 with SDRAM.

I also have a a System On Module board that has a Celeron 800Mhz? CPU which is PCI and ISA and pretty sure the module has the VIA onboard audio which supports DOS but the backplane doesn't have the traces / output run for the audio chip so I would have to mod it in order to use the onboard audio.

I wonder if you could contact Vortex86 and see if you could get them to send you some evaluation boards.... maybe for low cost or even free.

LightStruk wrote on 2020-12-11, 03:10:

As you said, the point is, "real hardware is cool." If there's enough folks who think such a product would be compelling because it's real hardware, then this has value beyond idle daydreaming. If not, then okay, it was fun to think about.

That's why I like this thread. I'll leave it to others to discuss commercial viability. But I find the idea compelling, and even as a thought experiment it makes for a really interesting conversation.

Pierre32 wrote on 2020-12-11, 04:47:

I'll leave it to others to discuss commercial viability. But I find the idea compelling, and even as a thought experiment it makes for a really interesting conversation.

"Coolness" is all so subjective, though. I can't imagine why an x86 instruction executed in silicon should be objectively more cool then the exact same instruction executed some other way.

It makes me think of how people seem to like Gentoo Linux for recompiling everything from source when it produces binaries indistinguishable from what you could get from any other package. Or how other people try to extract game ROMs from commercial distributions when the resulting file is indistinguishable from what you could easily download from slightly dubious sites.

But de gustibus non est disputandum.

The Gentoo folk might say it's about the process; the skills you put in, the learnings along the way, and the satisfaction of making it all work. And I could say the same about working with original hardware.

cyclone3d wrote on 2020-12-11, 03:54:

Does the EX2 support ISA DMA? It says no LPC bus so I'm not so sure it does. And the MX says no ISA.

Asking the questions that matter! I just downloaded the Vortex86 EX2 Master Core Fact Sheet, and it says:

1ISA Bus Interface
2- AT clock programmable
3- 8/16 Bit ISA device w/ Zero-Wait-State
4- Generate refresh signals to ISA interface during DRAM refresh cycle
5- Support Max ISA Clock 33M
6- Support 7 channel ISA DMA
7- Support ISA IRQ x 11

cyclone3d wrote on 2020-12-11, 03:54:

I have an old SOC development board that has PCI and ISA slots. I forget what the SOC is. I'll have to take a look since I haven't messed with it in quite a while. It is like a fast 486 with SDRAM.

I wonder if you have an AMD Elan SC520... that's a fast 486 with SDRAM.

cyclone3d wrote on 2020-12-11, 03:54:

I also have a a System On Module board that has a Celeron 800Mhz? CPU which is PCI and ISA and pretty sure the module has the VIA onboard audio which supports DOS but the backplane doesn't have the traces / output run for the audio chip so I would have to mod it in order to use the onboard audio.

I don't think the VIA onboard sound is good enough for this project. I'm aiming for genuine OPL3 one way or another, but also ideally an all digital audio path. I can get that from a discrete YMF262 / YMF289 (without using the Yamaha DAC), from a PCI Yamaha YMF74x, or possibly from an ISA CMI8330. Seriously, check out this video - the CMI8330 is indistinguishable from a discrete YMF262. Having the FM integrated into the sound chip again saves money and reduces risk. The CMI8330 also has the bonus of SB16 and WSS compatibility, not just SB Pro.

Just checked.. it is an STPC Consumer-II.