MIST vs MISTER?

[orange808]

13 minutes ago, Keatah said:

Nobody is really duplicating the TIA on a transistor/capacitor/resistor/diode level. So we're stuck with approximations.

 

Not enough gates on the DE-10 Nano to be that accurate.

 

The source for the MIST and MiSTer cores are on github.  If you peek under the hood, you'll see just how generalised the emulation/simulation is for multiple machines.

 

Yet, the emulation for most machines is pretty solid.

 

Of course, the Atari VCS is so tricky because it relies heavily on timing.  (Something we all know.) 

 

Anyhow, porting the MIST core to MiSTer is an option, but there's rumors that better TIA and RIOT simulation is cooking with the MiSTer 7800 core.  So, porting over from the future MiSTer 7800 core may be the direction the official a2600 MiSTer goes.

 

[ijor]

43 minutes ago, Keatah said:

Nobody is really duplicating the TIA on a transistor/capacitor/resistor/diode level. So we're stuck with approximations.

 

I don't see any discrete diodes or many capacitors at the schematics. TIA is not a complex mixed signal chip like, say, SID. It is basically purely digital except three rather simple analog features:

 

- Sound volume

- Paddles

- Color phase generator

 

Nothing that seems too difficult to model accurately with an FPGA. And even if not done very accurately, it doesn't seem that it could actually break a game, in the worst case. Btw, exactly the same three analog features are present in the 8-bit chipset with almost the same logic, one in CTIA/GTIA and the two other ones in POKEY.

 

Do I miss something? What kind of approximations are you talking about?

[ijor]

45 minutes ago, orange808 said:

Not enough gates on the DE-10 Nano to be that accurate.

 

Are you sure? More complex devices were modeled with SPICE like simulation on smaller FPGAs.

But again, why would you need something like that?

 

Quote

The source for the MIST and MiSTer cores are on github.  If you peek under the hood, you'll see just how generalised the emulation/simulation is for multiple machines.

Yet, the emulation for most machines is pretty solid.

 

It varies a lot depending on the specific core. Some are cycle exact, some are close, some don't even attempt to be very accurate.

Edited February 14, 2020 by ijor
Typoe

[foft]

The sid also has the fun feature where with multiple wave types active the transistors drive against each other, somewhat randomly ending up high/low. This is handled with tables captured from original chips usually, but it varies by chip. So there is no one true sid sound.

Edited February 14, 2020 by foft

[orange808]

On 2/13/2020 at 8:45 PM, ijor said:

 

It varies a lot depending on the specific core. Some are cycle exact, some are close, some don't even attempt to be very accurate.

Cycle accuracy is more about timing than accurate hardware behavior.

 

Also, I peeked in on the project today.  Looks like paddle and spinner support is coming down the line.  :)

[ijor]

3 hours ago, orange808 said:

Cycle accuracy is more about timing than accurate hardware behavior.

 

You don't have to tell that to me. I am literally obsessive about cycle accuracy. And I do mean exact timing not just accurate "behavior". Did you see my cores?

[Keatah]

Timing is important. Behavior is important. All of it is important. And that's why people dislike software emulators, especially the early ones on early PCs of the 486-PentiumIV era.

 

They got a bad rap for belching, burping, and stuttering their way through games. With today's i7 machines and years of development behind us that is no longer a problem.

[orange808]

23 hours ago, Keatah said:

Timing is important. Behavior is important. All of it is important. And that's why people dislike software emulators, especially the early ones on early PCs of the 486-PentiumIV era.

 

They got a bad rap for belching, burping, and stuttering their way through games. With today's i7 machines and years of development behind us that is no longer a problem.

Without going to far into this, I will point out a catch-22.

 

1.  People complained when MAME and Higan stopped working on "playable" frame rates and focused on producing accurate emulation.

2.  People complain about emulators that were clearly designed to hit frame rate on real machines.

 

There's no way to win.  :)

 

Nesticle certainly didn't stutter it's way along.  In fact, it was able to output the right refresh rate.  Try that with your garden variety Retroarch core and a crummy cheap LED flatscreen.  You can always underclock the emu to get an even 60Hz, but that's not right.  You can also employ some kind of frame rate conversion if you like tearing or dropped frames.  Nesticle just changed the signal and kicked out analog video to a CRT.  No problem.

[slingshot]

I found this topic, and I felt I must write something here...

I've spent a lot of time (and it's really a lot) with MiST. I've also developed a lot of code on it. And I still like it.

MiST is already 6 (or 7?) years old. Yet still capable of many thing. MiSTer of course is more powerful. Not just the FPGA has much more space, but the ARM side is stronger. However it has an unfair advantage: For every DE10-nano purchase, Intel pays the bill. The FPGA chip itself is more expensive than the whole MiST! The HDMI license for producing this board in small quantity would be also huge. Of course, as a user, probably nobody cares. But this basically stopped custom board development, since nobody can compete with this price. The realistic price would be probably around 5-600$. Also this board is not intended to sell to end-users. If some kind of fair-trade commission would exist, they surely could sue the sellers for selling under cost price, but I'm not a legal man to discuss this.

 

But even MiST is old, and less powerful, it still has some advantage (at least what I feel as an advantage):

- Fully open source design, the whole board can be built at home (well, if you have very good soldering skills).

- Single board, USB HUB included.

- No cooling is needed, the FPGA nor the ARM are warmer than the room.

- Some cores are still better (I feel many MiSTer cores goes to a more features direction instead of accuracy - also the huge number of available resources are encouraging developers to write lousy code).

- Better SDRAM placement, the GPIO pins on the DE10-nano are not really designed for driving a fast SDRAM.

 

And for some advantages to a developer:

- No Linux, it's much more fun to hack on the firmware.

- Faster compile time of the cores.

- Simpler framework.

- No code gets rewritten just because the coding style of the original.

 

I don't want to start a war, it's just my personal opinion, and why I'm still on MiST (however a fair amount of my code are also ported to MiSTer). Maybe I like retro hardware on retro hardware

 

About the A2600: the original core from Retromaster is closely designed around the authentic TIA hardware. The lfsr counters, event handling, and so on. I've added some missing features, and now the core is almost 100% perfect on MiST. I cannot find a single glitch in games. Maybe the RIOT timer is not perfect in some circumstances. Then this code is ported to MiSTer. I wonder why the people are complaining there, don't know what went wrong. However about Rysha's A7800 core: I saw the code. The TIA there is from some university project, which is surely crap. It's not the Retromaster's TIA. I'm sure that has to be replaced. Also I don't know why that code is used. Maybe because it's Verilog, Retromaster's TIA is VHDL.

Edited February 24, 2020 by slingshot
typo

[ijor]

6 hours ago, slingshot said:

I found this topic, and I felt I must write something here...

I've spent a lot of time (and it's really a lot) with MiST. I've also developed a lot of code on it. And I still like it.

 

For those that don't know him, slingshot is probably now the main MiST developer. Thanks to his talent and dedication the MiST is still very much alive

 

Quote

- Fully open source design, the whole board can be built at home (well, if you have very good soldering skills).

 

To be completely fair, I'm not sure this is 100% accurate. The board design files, AFAIK, were never published, right? Guess this was a Lotharek requirement and perfectly understandable.

 

[tuf]

Still interested in picking up a mist for the ST emulation.  Can't be beat.  

[slingshot]

6 hours ago, ijor said:

To be completely fair, I'm not sure this is 100% accurate. The board design files, AFAIK, were never published, right? Guess this was a Lotharek requirement and perfectly understandable.

 

Yes, only board schematics and a PDF of the PCB is published, but this didn't prevent two identical clones, and one with a Cyclone IV FPGA already. Too bad not with a bigger model.

 

And to be fair, your contribution raised the accuracy of the 68000 based cores to a new level (which are not just ST and Genesis, but several arcade cores are using FX68K, too).

[eightbit]

Jumping back in on this and bumping the thread as I thought it was quite interesting.

 

I used to own a MiST and sold it a while ago as the cores were not really mature and I already have a real Amiga and an ST. I wanted to play other things (game consoles, other computers) with it. But, a lot was to be desired.

 

So, fast forward to recently (last week actually) I started thinking about revisiting FPGA gaming and computing. I saw that the old MiST was at v1.4 and is still sold by Lotharek. I almost purchased it but checked on core status beforehand. Well, not much has changed. People are just no longer working on cores for it. I am not going to say all work stopped, but at least the few cores I was interested in have not been updated in a year or longer.

 

On contrast, the MiSTer project is in full swing. Every other day I check and some update has happened to a core...or a new core was introduced. Lots of activity means a brighter future in my eyes. So, I gave up on the MiST purchase idea and went for the MiSTer.

[Keatah]

Emulators, hardware FPGA and software-based stuff, are never quite 100% perfect, so ongoing development is a required feature. And yes that gives MiSTer an advantage.

[foft]

51 minutes ago, Keatah said:

Emulators, hardware FPGA and software-based stuff, are never quite 100% perfect, so ongoing development is a required feature. And yes that gives MiSTer an advantage.

I’d say this is one of the main disadvantages of emulators and fpga’s. The original ASICS have many bugs, but bugs that never change. Well except when other chips are made, eg sid versions, psg versions. Even then these are a limited number of versions to code against.

 

What I’m trying to say is no-one is going to code against a bug in only core version 32-35 of the flag 800xl core, there are just to many logic gate patterns.

 

i actually started the 800xl core as a less buggy one, though too much software requires the hardware bugs to work. Poor software practice really, but this is a large aspect of the fun of the 8-bits.

[foft]

4 hours ago, eightbit said:

On contrast, the MiSTer project is in full swing. Every other day I check and some update has happened to a core...or a new core was introduced. Lots of activity means a brighter future in my eyes. So, I gave up on the MiST purchase idea and went for the MiSTer.

It’s really great to see the activity, very exciting to read. I actually only get the mister out every & months, spend 70% of time updating, play a game or two and put it away! Also the multitude of systems magnifies the paradox of choice, I find this with many of the multi system systems.

 

More static, more mature, less time updating and more time using.

[Keatah]

I can understand that. Well developed emulators can and do emulate those bugs. Some buggy chips or versions are selectable via an option. Tick box or command-line option. And undocumented CPU op-codes are almost always handled by every emulator.

 

Constantly improving an emulator means more of those bugs are being preserved. Early on it didn't used to be that way. But now it is.

 

Making the "perfect" emulator doesn't mean squashing out chipset revision bugs to create the theoretically perfect representation. Rather the reverse, duplicate the operation of the device including errors and one-off corner case issues.

Edited April 10, 2020 by Keatah

[Keatah]

I spend about an hour or so every month reviewing and conducting select updates to emulators. I consider it the virtual equivalent of dusting the cartridge shelf, or rewiring the ABCD switchbox, or cleaning the contacts and ports on a console. Might as well liken it to the occasional controller repair, or cable replacement, too

 

One way or another you WILL NOT be skipping maintenance. Regardless of what path you choose to tread.

Edited April 10, 2020 by Keatah

[Colleton]

I bought a nice Mister last summer via the FB group.  Very nice device, the arcade emulation is excellent and the 8-bit Atari emulation is absolutely perfect.  The only downside that I've seen is that the ST emulation doesn't work for me. 

[ijor]

4 hours ago, Colleton said:

I bought a nice Mister last summer via the FB group.  Very nice device, the arcade emulation is excellent and the 8-bit Atari emulation is absolutely perfect.  The only downside that I've seen is that the ST emulation doesn't work for me. 

 

My core? You don't get any video? If so, then your monitor probably doesn't support the video mode my core uses by default ...

 

Connect the MiSTer temporarily to some other monitor or TV. Most do work. Invoke  the menu, change the video mode to 60Hz and save the configuration. Now when connecting to your other monitor it should work.

[slingshot]

On 4/10/2020 at 5:26 AM, eightbit said:

So, fast forward to recently (last week actually) I started thinking about revisiting FPGA gaming and computing. I saw that the old MiST was at v1.4 and is still sold by Lotharek. I almost purchased it but checked on core status beforehand. Well, not much has changed. People are just no longer working on cores for it. I am not going to say all work stopped, but at least the few cores I was interested in have not been updated in a year or longer.

 

On contrast, the MiSTer project is in full swing. Every other day I check and some update has happened to a core...or a new core was introduced. Lots of activity means a brighter future in my eyes. So, I gave up on the MiST purchase idea and went for the MiSTer.

I don't know what did you check, but lot of code which end up on MiSTer still taken from MiST. It's just not promoted on that level.

[seastalker]

I already had a MiST and eventually got a MiSTer. I like them both, and have the MiST hooked up to a 90s VGA monitor and my MiSTer sitting where I can go HDMI to the flatscreen, connect to another VGA monitor, or use a converter for the big CRT tv.

[Hwlngmad]

Is the MiST worth getting at all considering all that is out there?  I have one for sale locally and am tempted to get it.  However, I really don't have any time to fiddle with things.  I am more of a 'plug and play' kind of guy admittingly.  Would it be worth me looking into as I currently don't have anything set up to play Amiga, ST, or C64 games.  Thanks and any feedback is appreciated.

[newtmonkey]

I've got a MiST and I use it primarily to play Amiga and ST games.  Great cores for those computers.  If the price is right, I'd say go for it.  You just drop the core files on an SD Card, put your disk images in subfolders, plug in a keyboard, mouse, and joystick and you're good to go.

 

It supports both USB and traditional "Atari" 9-pin joysticks.  For video, it's just got a VGA port (no HDMI or DVI), but you can also get a custom RGB SCART cable if you want to plug it into a TV.

 

It does have some decent PC Engine and Megadrive cores (no CD support) if you want to play some console games on it, but it falls far behind MISTER when it comes to console gaming.

[youxia]

Depends on the price I guess. If it's really cheap and you really want to mess around only with these micros then go for it. But consider that MiSTer+32SDRAM (this runs everything apart from some NeoGeo games) costs ~160$ and offers more platforms.