MiSTer FPGA Forum

I mean it seems like PS1 / Saturn are doable so honestly if MiSTer can pull that off...I am cool with it!

I mean in ten years do I hope for "MiSTer V2: Hyper Edition" that can handle Dreamcast/PS2 era stuff? Def. But I can wait

I've been really enjoying playing with the MiSTer and the 8/16 bit cores; these are machines where timing is vital to the systems and at the same time impossible to implement in an emulator on modern PC hardware.

Later systems aren't a great match - the code written for an Xbox / PS2 / Gamecube doesn't rely on exact timings so you can achieve a fantastic result via software emulation (with powerful PC hardware). Then you have some insane benefits by applying new rendering techniques and these AI-upscaled textures and art..

Sigismond0 wrote: ↑Tue May 04, 2021 4:02 pm You're missing something here--emulators for N64, Playstation, Wii, etc. can alter the internal rendering resolution and actually run the game at native 1080p or 4k or whatever, plus they can load in HD texture packs. Completely separate from scaling native render resolution.

That is because those emulators are replacing the hardware render of those original systems with the emulation's own tools.
Example: an N64 emulation on a Original Xbox isn't using the 3D hardware and Software API render the N64 had, it's using the existing Xbox Rendering to render N64 Game Graphics. It can give you stunning visuals, but it's also completely inaccurate as you are just forking N64 assets to work on Xbox hardware.

You are pretty much using the Emulation Environment to fudge rom into barely working order, You aren't simulating that MIPS R4300i @ 93.75 MHz and that Reality Coprocessor GPU, you are making that N Six Four rom run on a 733 MHz Intel Pentium III and Nvidia GeForce 3-based NV2A GPU and betting on the 733 Mhz Pentium 3 to be fast enough to do what a 93.75 MHz MIPS R4300i is supposed to do along with all the API redirects. Hoping the GeForce 3 with redirects can do a Simulacrum of a different GPU with completely different architecture.

radicalbyte wrote: ↑Sat May 15, 2021 6:59 pm I've been really enjoying playing with the MiSTer and the 8/16 bit cores; these are machines where timing is vital to the systems and at the same time impossible to implement in an emulator on modern PC hardware.

It's possible to implement the system timings of a 8 or 16 bit machine in modern PC hardware, you just need crazy amounts of system resources to do everything in code.

Full speed accurate emulation of a Super Nintendo with low level emulation (using something like BSNES) requires a minimum Dual core 3 Ghz CPU with 4GB of Ram, a Graphics card with bare minimum VRAM of 512 MB and I would not bother with an OS older than Win 7. As you are trying to do is use the higher clock speeds of that modern CPU to do tasks that was done with multiple chips in the original system.

You right about one thing if you were referring to emulators such as ZSNES or SNES9X, you aren't getting the accuracy results.
The upside of FPGA, we don't need a CPU serval billion times faster than the original hardware had, and we can run all those special chips in parallel rather than series. A modern gaming PC can run 8 and 16 bit consoles just as well as FPGA, but you use alot of system resources to do it.

This topic went the way I expected just by looking at the title, discussing emulating super mega duper powerful systems like PS2. My question is rather something in between- will it be possible to emulate rare systems like 3DO and CDI?? I know they were all commercial failures but this is why these systems should rather be in top priority of preservation and some games/hardware may still be interesting and worth it

VegaVegas wrote: ↑Sun Jan 02, 2022 7:29 pm This topic went the way I expected just by looking at the title, discussing emulating super mega duper powerful systems like PS2. My question is rather something in between- will it be possible to emulate rare systems like 3DO and CDI?? I know they were all commercial failures but this is why these systems should rather be in top priority of preservation and some games/hardware may still be interesting and worth it

This is more interesting to me rather than going beyond the Saturn. There are plenty of systems that could benefit from having an FPGA implementation, like the Apple Lisa for example.

The Apple Lisa could take a while. It doesn't seem to have a lot of software. And honestly, it's only been in the last six months that Macintosh core has gotten good.

I've said this in other threads, but the limitation is becoming people more than the FPGA. The DE-10 is probably pretty well matched to what individual hobbyists can actually do. Even if you postulate an infinite-size FPGA, going much further than has already been done would require Herculean amounts of labor; the original Cell processor took like $700 million to design. An FPGA reimplementation would be easier, because the original chip has already been made, but it could be still be a $100 million project, and that's assuming that Sony didn't swoop in and shut the whole thing down.

Software emulation is likely to be the only workable method for anything much past the PS1, no matter how good FPGAs get.

Malor wrote: ↑Wed Jul 13, 2022 2:52 am I've said this in other threads, but the limitation is becoming people more than the FPGA. The DE-10 is probably pretty well matched to what individual hobbyists can actually do. Even if you postulate an infinite-size FPGA, going much further than has already been done would require Herculean amounts of labor; the original Cell processor took like $700 million to design. An FPGA reimplementation would be easier, because the original chip has already been made, but it could be still be a $100 million project, and that's assuming that Sony didn't swoop in and shut the whole thing down.

Software emulation is likely to be the only workable method for anything much past the PS1, no matter how good FPGAs get.

The CELL also ran at 3.2 ghz in the PS3 so even if you have an unlimited size FPGA, unless it clocks at 3+ ghz is isn't going to be enough. The DE10 tops out around what... 90-100 mhz?

That's also an excellent point. FPGAs just don't seem to go that fast.

Malor wrote: ↑Wed Jul 13, 2022 2:52 am I've said this in other threads, but the limitation is becoming people more than the FPGA. The DE-10 is probably pretty well matched to what individual hobbyists can actually do. Even if you postulate an infinite-size FPGA, going much further than has already been done would require Herculean amounts of labor; the original Cell processor took like $700 million to design. An FPGA reimplementation would be easier, because the original chip has already been made, but it could be still be a $100 million project, and that's assuming that Sony didn't swoop in and shut the whole thing down.

Software emulation is likely to be the only workable method for anything much past the PS1, no matter how good FPGAs get.

MiSTER is very clearly and tangibly better at being an 8/16 bit machine than software emulation. The project has been an eye opener for me.

However I don't believe that holds for the later era. The P1/PS2/PS3 generations are debatable - I personally prefer to emulate over even using original hardware. For anything after that.. we're talking about machines which are designed as and - importantly - treated as "fixed configuration" PCs. In those cases the delta between software and any theoretical hardware re-creation is smaller. Also emulation is likely to improve the games via higher resolution, better frame-rates and more advanced effects. You see that already with PS1/PS2 emulation; for machines designed in era where TV resolutions span 720P to 4320P...

What you're actually seeing in PS2 and arguably already in the PS1 and Saturn is how ugly 3D used to be. It was positively dreadful compared to the 2D pixel pushing that was perfected before it. Mind you, that process took a decade or two and similar timespans are at play for 3D. Only in the current generation or two do we see really convincing 3D graphics that stand on their own without being 3D for the sake of it. We're decades removed from the likes of Descent on PC's or the PS1's library.

We see emulation trying to spruce up the horrific quality of yesterday's visuals in 3D by increasing resolution, frame rate, texture detail etc. When was the last time we saw that happening with Commodore 64 games? What we're seeing there instead is an attempt to turn those 4K TV's into CRT's with tricks like scanlines and shadow mask emulation essentially recreating 80's display tech. For 3D we're pushing the old games into 4K territory that they were never intended for... Intriguing..

Bas wrote: ↑Wed Aug 03, 2022 5:08 pm What you're actually seeing in PS2 and arguably already in the PS1 and Saturn is how ugly 3D used to be. It was positively dreadful compared to the 2D pixel pushing that was perfected before it.

Opinions, heh. Guess we should all uninstall that PSX core pronto, lest the visual dreadfulness of games such as Grandia, Einhander, Xenogears, GranTurismo, R-Type, Vagrant Story, and numerous other damages our eyeballs

There is discussion here about what a new board can do, but the way to conceive that board seemed to be as a simple upgrade of what it's done now but with more "headroom", so it include the FPGA + ARM CPU. It seems manufacturers start to build SOC with FPGA, CPU and GPU.
Of course these CPU and GPU parts are sure to be pretty slow... for now.
So the question i am asking, as a complete profane, is what the FPGA can offer here to build hybrid emulator for later hardware with these new all-in-one SOC ? I mean it was already discuss here that the asynchronous nature of PC is a problem for old hardware and less of a problem for newer one, where software emulation can "shine". But i remember reading here, i think from Aberu, that there is still part of newer hardware that would benefit from fpga because there are "things" really hard to emulate in software, for example building the PS2 CPU inside a fpga and doing software emulation for the gpu.
I am asking myself what could be done if the cpu and gpu "raw power" of these SOC is used together with the FPGA part that act to accurately emulate part of these machine that couldn't be well or easily emulated with software emulation.

Bas wrote: ↑Wed Aug 03, 2022 5:08 pm We see emulation trying to spruce up the horrific quality of yesterday's visuals in 3D by increasing resolution, frame rate, texture detail etc. When was the last time we saw that happening with Commodore 64 games? What we're seeing there instead is an attempt to turn those 4K TV's into CRT's with tricks like scanlines and shadow mask emulation essentially recreating 80's display tech. For 3D we're pushing the old games into 4K territory that they were never intended for... Intriguing..

We do this with mid-to-late-90s games:

https://www.youtube.com/watch?v=uSzGUEEbk4U

Exactly my point @radicalbyte.. crummy 3D apparently needs an upgrade.

@akeley my point wasn't to bash the PS1 library but it is objectively true that early 3D had a lot of catching up to do to late 2D games. The clunky animations, lack of facial expressions, clipping errors, low polygon count etc. were prevalent across platforms. If you pit that against mature 2D titles like the Metal Slug series on Neo Geo there's not even a comparison. Except (!) where 3D really does add to the gameplay, but that simply wasn't the case in far too many titles. They were made a certain way because the snazzy GPU was there and it was expected. Not because the game needed it. Funny that you should mention R-Type. That got a whole slew of fluffy but superfluous 3D cut scenes and animations but the game itself looks like the TurboGrafx version: very decidedly flat 2D.

There's a hardware context here. Those PS1 games look much better with interlace fields the system was designed around.

My concern is more what will happen if/when Terasic discontinues the DE10-Nano or decides not to sell it to the general public anymore. I was wondering about that today. I am happy with the modest emulation abilities for 8, 16, and basic 32-bit systems of this FPGA chip, but it would be very disruptive to lose the supply of boards. I understand they were significantly less expensive when the project started so maybe it was an easy choice, but I somewhat wish MiSTer were based on a custom PCB design.

How hard would it be to move to a custom design and what would happen to legacy support? Would there be a fork of the project at that point?

thisisamigaspeaking wrote: ↑Thu Aug 04, 2022 3:17 am How hard would it be to move to a custom design and what would happen to legacy support? Would there be a fork of the project at that point?

Custom design was the norm before MiSTer, but it's easier said than done.

MiST was a great choice back in the day as it was the cheapest option and easily available, but it remained expensive (about same as a DE-10 Nano last I checked) and it's currently out of stock given there are no Cyclone III FPGA chips available to make it (https://lotharek.pl/productdetail.php?id=45)

Then you had something like the FPGA Arcade Replay, which is older than MiST but suffered from supply issues (MiST creator shared that the project started due to not being able to get one). A Replay 2 is now in the works, but Im not sure about ETA for it.

Those two are/were not truly open hardware though. The only project like that was the ZxUno (CC license for the hardware) but that had a very tiny FPGA (9K LEs). AntonioVillena, who was part of the design team, used to have them in his store; recently he had an updated version (ZxDos) with 16K LEs.

None of these cards has HDMI built-in, which also meant you need an external upscaler to use on modern TVs, or suffer bad upscaling/shimmering on VGA. Before starting on MiSTer, Sorg spent quite a bit of time to find a good upscaling solution for MIST, but most had issues. As for me, back in the day I would use an XRGB Mini upscaler; but that was a premium option. Today you could probably use an OSSC instead or try to mod a GBS 8200 if you're a tinkerer.

TL;DR it's hard to beat the value proposition of the DE10-Nano, even at a slightly higher price. If that goes away, unless a similar card is available, then it would take more hardware to do something similar (and I'd expect everything to be pricier, easily in the $500-$800 range for a full setup)

Edit: to end on a more positive note, at least there are plenty of options as long as cores themselves are open source

Let's not forget one EXTREMELY critical factor - the DE10-Nano is very heavily subsidized by Intel University...
It's doubtful the MiSTer project would be where it is without that subsidy

Personally speaking, I am very surprised that we have gotten the PSX and 32X core on the MiSTer. I really thought they would never happen.

And the Saturn core is just icing on the cake.

The fact that we can have so many computers and consoles and arcade cores in a tiny box already seems like magic to me.

Even if we never see a successor to the MiSTer, the current version still has years of untapped potential if you ask me.

Newsdee wrote: ↑Thu Aug 04, 2022 4:05 am TL;DR it's hard to beat the value proposition of the DE10-Nano, even at a slightly higher price. If that goes away, unless a similar card is available, then it would take more hardware to do something similar (and I'd expect everything to be pricier, easily in the $500-$800 range for a full setup)

Edit: to end on a more positive note, at least there are plenty of options as long as cores themselves are open source

I think it already ran me $800+ with Checkmate case and MT32-Pi (not to mention that I had a Pi 3B on hand). AliExpress price for complete system is pushing $500 with no MT32-Pi. I'm wondering if some of these costs couldn't be consolidated with a single (mini-ITX?) board solution. Not counting the fancy case of course, but if it's mini-ITX there are a ton of cases on the market.

And of course the barebones DE10-Nano + memory module option can't be discounted, still quite cost effective.

I'm definitely glad I went this route rather than a Vampire V4 Standalone, no question there, I didn't want to be tied to some proprietary code, although I lust after the specs still.

Re: your edit, that's what I hope, is that with the open source cores in the future when it is time to move on, it can be done reasonably painlessly. Open source is the way. Would love to see the PCB open source as well, assuming an FPGA chip can be chosen that will have some legs as far as availability. Won't the cost on FPGA chips drop at some point?

Bas wrote: ↑Wed Aug 03, 2022 8:46 pm
@akeley my point wasn't to bash the PS1 library but it is objectively true that early 3D had a lot of catching up to do to late 2D games. The clunky animations, lack of facial expressions, clipping errors, low polygon count etc. were prevalent across platforms. If you pit that against mature 2D titles like the Metal Slug series on Neo Geo there's not even a comparison. Except (!) where 3D really does add to the gameplay, but that simply wasn't the case in far too many titles. They were made a certain way because the snazzy GPU was there and it was expected. Not because the game needed it. Funny that you should mention R-Type. That got a whole slew of fluffy but superfluous 3D cut scenes and animations but the game itself looks like the TurboGrafx version: very decidedly flat 2D.

It's in no way "objectively true" - it's just an opinion, based on a really bad comparison and a slew of rationalizatons. It's no more true than somebody saying that "pre-NES games were too simple graphically to be counted as art" (true story). Following this logic I could also say that Asteroids, Donkey Kong, or Defender look crummy/horrific/dreadful, borrowing from your hyperbole, because they lack zillions of colours, sprites, parallaxes, and whatever else the pinnacle 2D had. And, of course, it would be as ridiculous as the argument about the early 3D. Conversely, it could be also applied as a judgement from a modern gamer regarding that very pinnacle compared to modern standards. Just ask any kid without the nostalgia bias which 2D games he would prefer.

As it is, the fact that 3D at the time was at its proto stage doesn't matter, because a skilled artist can utilize any tools at his disposal to great effect, and so was the case with the early 3D games. I mean, seriously, trying to tell me that games with exquisite art design such as the ones I mentioned earlier (which R-Type was the weakest of, but even so) look "dreadful" is simply absurd, even allowing for "not disputing the taste" trope. Not all games had great looks, of course, just like not all 2D games had, of any given era, but that's normal. They don't need upgrading any more than caveman paintings do. Sure, it's being done because we can, and because some people enjoy this look, but it in no way it is a proof that it needed to be done. In fact, I think it often hinders rather than helps because it makes the games look unnatural - but that in turn is a function of them being displayed on modern displays, which indeed do make them look "horrific" in the first place, but that's another story.

Overall, the "early 3D sucked/aged badly/is dreadful - 2D rocks!" is just one of the many modern myths and misunderstandings about retro gaming. What makes it stand out is how dreadfully unfair it is to the people who made these games, and the games themselves. I just hope this is something that won't stick past the modern trend and be remembered as fact say, 100 years from now, hence my occasional counterpoint when I see it.

I guess my main point, without triggering sensitivities here, is that there was an era in which 2D games were at their pinnacle while 3D was in its pre-infancy. Marketing forces then dictated an exaggerated push to 3D long before the hardware was ready to match that of 2D to create immersive experiences. The result was games with clunky graphics.

I'm arguing not from taste but from fact. The fact being that we see many emulators nowadays trying to improve and paper over early 3D's deficiencies by pulling them up to today's tech standards while the reverse is true for 2D where we actually see efforts to push today's tech down to the period-appropriate display devices' characteristics. There are obvious exceptions both ways, but the trend is there.

thisisamigaspeaking wrote: ↑Thu Aug 04, 2022 5:12 am And of course the barebones DE10-Nano + memory module option can't be discounted, still quite cost effective.

True, prices are now closer to 500 already (or more) with a fancy case. But to be fair something like the Checkmate is reusable; you could always upgrade it to a different internal board later on. Whereas a MiSTer PCB case will only ever be for the DE10 (but is cheaper).

I see much talk of MiSTer on other channels (YT, Twitter, etc) mostly treating it as a packaged product and total cost. I understand that's a mental shortcut that many people need to make otherwise they will think you have to solder PCB and compile Linux kernels to use it. But that point of view completely misses the bigger picture, which is that the DE10-Nano can be just one component in a wider build. Without much more effort than building an IKEA cabinet, you have a lot of options in terms on scope and cost.

For example, if you have an HDMI arcade cabinet (say a Chinese Vewlix clone or a modded Arcade1up) then the bare DE-10 with a cheap OTG USB hub and 32MB of RAM is more than enough to run most of the cores. And sure those cabinets and mods can add cost, but like the Checkmate case, they're not restricted to MiSTer per se.

And if you own a full-size CRT arcade cabinet like a Japanese Candy Cab, then the cost of a DE10 and one of the MiSTer-specific JAMMA adapter IO boards is peanuts compared to the cost of a JAMMA PCBs for any famous game.

Re: your edit, that's what I hope, is that with the open source cores in the future when it is time to move on, it can be done reasonably painlessly. Open source is the way. Would love to see the PCB open source as well, assuming an FPGA chip can be chosen that will have some legs as far as availability. Won't the cost on FPGA chips drop at some point?

One major issue about custom PCBs Is the soldering required for newer and bigger FPGA boards. The Cyclone III is just SMD and can be done by a hobbyist at home with a reflow oven, but the Cyclone V is BGA and that needs to be done at a factory. Antonio had an interesting approach for his NeptUno - he took a Cyclone 4 breakout board and built the rest around it (although that's still only VGA, and I'd be more worried about supply of that board vs. a DE10...)

akeley wrote: ↑Thu Aug 04, 2022 8:52 am Overall, the "early 3D sucked/aged badly/is dreadful - 2D rocks!" is just one of the many modern myths and misunderstandings about retro gaming.

It's not a myth, not even a little bit. Early 3D was terrible. Late 2D was usually excellent.

It took about ten years for 3D games to catch up to where 2D games had been. Doing 3D was extremely difficult, and getting a 3D engine up and running *at all* was a major feat. With the extreme amount of effort required to create and animate even crude 3D, game quality took a huge hit. I think of it as the lost decade, where game design barely moved forward, as everything had to be re-invented from the ground up for the new paradigm.

I remember being annoyed at the time at how poor so many of the 3D games were. The PS1 library has a ton of stinkers. And even the good ones are often super-painful for modern gamers to play, because the interface standards hadn't been worked out yet.

Yeah, early 3d games lacking was the consensus even at the time. Most articles from that era that talked about 3D games focused on their potential.

Consoles were always going to be behind and compared to 'then' current PC hardware. Doom changed home and arcade gaming forever for me in 1994.

What the PS1 core does and the ao486 can't do, is show off that early development in 3d gaming well on the MiSTer.

I think there's a natural break at PS1 and Saturn cores for the DE-10.

Plenty of Arcade cores yet to be built. I'd like to see some more input options developed, to get the right control set up to work the core as intended.

To everyone saying early 3d sucked compared to 2d: Daytona USA.

That's really all that needs to be said on this idiotic topic. It's a myth that early 3d was bad. Get over yourselves and submit to Daytona.

Blamo wrote: ↑Sat Aug 06, 2022 6:47 am To everyone saying early 3d sucked compared to 2d: Daytona USA.

Elite, The Sentinel, BattleZone, Stellar 7, SkyFox, Carrier Command, Star Fox...

And then there's Mario64...

Blamo wrote: ↑Sat Aug 06, 2022 6:47 am To everyone saying early 3d sucked compared to 2d: Daytona USA.

That's really all that needs to be said on this idiotic topic. It's a myth that early 3d was bad. Get over yourselves and submit to Daytona.

Yes, amazingly enough, the PS1 did have some good games on it. But as a percentage, it was probably lower than on any other console, because 3D was so difficult to program.

For every standout folks list, there will probably be twenty or thirty that sucked. There were a lot of PS1 games, so there were bound to be some good ones. But if you were living through the era, you were probably burning a lot of cash on losers.

I think you might see "fantasy" machines/consoles get created .. someone taking bits and parts from a number of preexisting cores and creating something new that never existed before .. or say taking an existing core and adding all kinds of enhancements to it that radically change it in many ways.. I think we haven't even really touched that aspect of what the Mister makes possible..