Sega Model 3

[Motoracer5]

I found something interesting

https://www.youtube.com/watch?v=wriw0CyHUwg

https://en.sega-dreamcast-info-games-pr ... g-Nid5YjgE

Sega Model 3 (1996): Processor 66Mhz / Audio RAM 1Mb / Vídeo 8Mb RAM 66Mhz.
Sega Dreamcast (1999): Processor 200Mhz / Audio RAM 2Mb / Video RAM 8Mb 100Mhz

How about comparing what it takes to run Model 3 emulation on PC?

Processor 3000+Mhz on a far more advanced CPU / 2000MB system RAM / Video 1000MB RAM 900MHZ

Looks like dreamcast might have at best 4% of what is needed. Also the openGL for Dreamcast is way too old

[Ian Micheal]

This because people MIX up atomiswav thinking it's emulation which it's not cant compare or think anything like it or even close can run dreamcast cant even run a c64 emulator in single cycle mode with 1541 disk drive lol.. Not joking here it does not have enff cpu horsepower to do it..

C64 emulator is cpu bound not 3D grafix bound.. 200mhz is not enff

This whole crazy thing was because of Atomiswav which is just a dreamcast..

[Motoracer5]

This because people MIX up atomiswav thinking it's emulation which it's not cant compare or think anything like it or even close can run dreamcast cant even run a c64 emulator in single cycle mode with 1541 disk drive lol.. Not joking here it does not have enff cpu horsepower to do it..

This whole crazy thing was because of Atomiswav which is just a dreamcast..

It seems lots of people expect for too much of the Dreamcast. Some of the requests for PC ports I have seen are insane

[Roareye]

Sega Model 3 (1996): Processor 66Mhz / Audio RAM 1Mb / Vídeo 8Mb RAM 66Mhz.
Sega Dreamcast (1999): Processor 200Mhz / Audio RAM 2Mb / Video RAM 8Mb 100Mhz

Even if those stats are true (I have no reason to believe they aren't), they only show that technically the DC could run Model 3 ports.

The main issue is how you would deal with the ports. Emulation would require more processing power and RAM than the DC has available (especially when you add the requirements of the Model 3 on top).

You could do a straight port, changing the game slightly to adapt it for the console. However to do a straight port, you'd need the source code of the original game. Only SEGA has that (and even then, sometimes source code gets lost) and I doubt they'd give the code up to fans. Even with the source code, it'd take a team to actually port the game - there's a reason most ports take an approximately 20 man team over the course of 1 or 2 years.

The last option is to build each game's engine from scratch or via reverse engineering. This has worked in the past - Christian Whitehead built his own 2D Sonic engine and eventually got to make official ports of Sonic 1, Sonic 2, Sonic CD and then make Sonic Mania. But that's a 2D engine, and a rare occurance. And reverse engineering is also a very lengthy and legally shaky endeavour.

I, like anyone, would love to see the Model 3 games ported to DC. Daytona USA 2 deserves to receive a solid home port. It'd be great!
However the first task would be to find a team not only competent at the task, but also willing to give months (if not years) of their time to make it a reality. That's a big ask.

[Ian Micheal]

Also reverse a code bin is still copyright.. a decompilation still falls under copyright
it's not a clean room reverse engineering (where you just try to recreate what you see on the screen) but they've taken the copyrighted executable and transformed it into code..

So even after all that i doubt any team with the talent to do it would do it.. for something that would get prolly nothing for it but a black mark on there name..

Course a straight port with all the src and translated to C as well no doubt these games have vast amounts of asm that suits there cpu and file systems.. Your right up to 2 to 3 years.. also depending on if these said persons cared enff to do it..

Just because a system specs is lower dont mean it's easy job at all The way the pvr works and system 3 is not the same at all might as well be double dutch to it.. You might be shocked to find it having a hard time running it with out vast amounts of optimizing to suit the sh4 pvr2dc chip set.

The others like gta iii and sm64 you know they did it for fame and bragging rights .. System 3 i doubt they care.

[Alex Maafck]

What about model 1 or model 2 emulation? Or any of the pc ports Sega released such as Sega Touring Car for windows? It’s such a shame how the Saturn port turned out. Of course I could contribute nothing except to fund such a project. Just wishful thinking. Those Atomiswave ports have us all spoiled now..

[Ro Magnus Larsson]

Sega Model 3 (1996): Processor 66Mhz / Audio RAM 1Mb / Vídeo 8Mb RAM 66Mhz.
Sega Dreamcast (1999): Processor 200Mhz / Audio RAM 2Mb / Video RAM 8Mb 100Mhz

Even if those stats are true (I have no reason to believe they aren't), they only show that technically the DC could run Model 3 ports.

The main issue is how you would deal with the ports. Emulation would require more processing power and RAM than the DC has available (especially when you add the requirements of the Model 3 on top).

You could do a straight port, changing the game slightly to adapt it for the console. However to do a straight port, you'd need the source code of the original game. Only SEGA has that (and even then, sometimes source code gets lost) and I doubt they'd give the code up to fans. Even with the source code, it'd take a team to actually port the game - there's a reason most ports take an approximately 20 man team over the course of 1 or 2 years.

The last option is to build each game's engine from scratch or via reverse engineering. This has worked in the past - Christian Whitehead built his own 2D Sonic engine and eventually got to make official ports of Sonic 1, Sonic 2, Sonic CD and then make Sonic Mania. But that's a 2D engine, and a rare occurance. And reverse engineering is also a very lengthy and legally shaky endeavour.

I, like anyone, would love to see the Model 3 games ported to DC. Daytona USA 2 deserves to receive a solid home port. It'd be great!
However the first task would be to find a team not only competent at the task, but also willing to give months (if not years) of their time to make it a reality. That's a big ask.

The Model 3 /DC specs I published are right. Of course not enough. What you say about emulate M3 in DC I was talking with one of Supermodel creators (Bart), and he thinks about the same probabilities for DC, it is more accurate to think in port each game than think in the M3 emulator. In porting games the main issue is 1)Source code (as always) 2) the additional RAM memory M3 games uses in ROMs.

About time and effort I guess young people like Bart does not lack of that time

[Ro Magnus Larsson]

Also reverse a code bin is still copyright.. a decompilation still falls under copyright
it's not a clean room reverse engineering (where you just try to recreate what you see on the screen) but they've taken the copyrighted executable and transformed it into code..

So even after all that i doubt any team with the talent to do it would do it.. for something that would get prolly nothing for it but a black mark on there name..

Course a straight port with all the src and translated to C as well no doubt these games have vast amounts of asm that suits there cpu and file systems.. Your right up to 2 to 3 years.. also depending on if these said persons cared enff to do it..

Just because a system specs is lower dont mean it's easy job at all The way the pvr works and system 3 is not the same at all might as well be double dutch to it.. You might be shocked to find it having a hard time running it with out vast amounts of optimizing to suit the sh4 pvr2dc chip set.

The others like gta iii and sm64 you know they did it for fame and bragging rights .. System 3 i doubt they care.

About time, a talented team and effort I guess young people like Bart does not lack of that time, maybe we can encourage them to advance in this! they love SEGA and they are in this now

[Ro Magnus Larsson]

Bart Trzynadlowsky, the creator of Supermodel emulator was talking with me and I asked him about both systems in the possibility to port M3 to Dreamcast, he 's saying to me this in his Youtube channel:

-I think PowerVR was indeed more powerful than Model 3, and certainly more flexible. Maybe there is some metric that Model 3 excelled at but unless someone shows me otherwise, the public specs (wherever they come from) indicate the Dreamcast's PowerVR exceeds Model 3 performance in vertex transforms and fill rates.

Would be nice to run a synthetic benchmark someday once I can more easily transfer my own custom Model 3 programs onto the boards.

In the 90's, graphics progress was extremely rapid. Every piece of 3D hardware was obsolete within months! The Pro-1000 occupied an entire motherboard-sized board. It was therefore not a true "GPU" because it consisted of multiple large custom chips. PowerVR was a single chip, if I remember correctly, and therefore a true GPU and an example of the powerful implications of Moore's Law.

(...) in the how the two system naturally want to manage assets. E.g., on Model 3, the renderer has access to a huge amount of VROM whereas meshes need to be explicitly loaded in from disk on Dreamcast.
-I would expect that Dreamcast's dynamic lighting capabilities were far better than Real3D. For example, how would one do lightmaps on Real3D? I presume something like it was possible on Dreamcast. I was thinking the other day whether Quake could be ported to Model 3. It could easily handle the geometry but what about the dynamic lights?

-The hot air ripples on the desert stage are interesting because they're so ridiculously cheap and easy to do on tile generator hardware, which has a per-line scroll mode. But if all you have are textured polygons or a frame buffer to draw to directly, the effect becomes much more complicated to pull off.

-I'd love to see a synthetic benchmark constructed for Model 3 and Dreamcast Smile It is within the realm of feasibility because I have gotten code running on Model 3. Just need a better mechanism to transfer it than burning EPROMs lol.

RAM and ROM: RAM and ROM re only one part of the issue. Model 3 games have access to all of their ROM immediately and the CPU (and graphics processor) can read from it directly. A big part of the porting effort is reorganizing the games so that they can be broken up into chunks that can be loaded infrequently from disc into Dreamcast RAM. For example, in a fighting game like VF3, you can read in the level geometry and opponent character model between stages. It might be trickier with something like Sega Rally 2 where on the Model 3, the replay at the end of the championship race can go through all the levels without any loading time (maybe the levels were small enough that this was true on Dreamcast too, I don’t know). But the bigger issue is that emulation has ~10x the overhead of a port. Supermodel doesn’t translate Model 3 code, it interprets each instruction one by one. It takes more than 10 instructions (ballpark estimate) for Supermodel to execute a single instruction of the emulated PowerPC CPU. On top of that, Model 3 graphics hardware works differently than PC 3D hardware or Dreamcast, so there is considerable overhead parsing and preparing the 3D vertex data for rendering and making sure the graphics hardware renders it in a way that matches Model 3 behavior.
When porting a game, the graphics code would be completely rewritten to work directly with the target system.

May 2022

[NeoSnk]

It exists a SEGA Model 3 Emulator for Windows, Mac OS and Linux, it is called "Supermodel", it has very good compatibility, still in work but you can see yourselves the compatibility list:

Supermodel (SEGA Model 3 emulator comp. list).png

This M3 emulator is open source and has a website ( https://www.supermodel3.com/About.html ). The most ambitious game for me is Scud Race, a HQ racing game at the time.

I remember that in the 1998 Japan Dreamcast presentation it was announced that the console was an easy machine to port previous arcade games and it was designed that way specially to port SEGA Model 2 and Model 3 games, later nothing happened.
It would be the most successful console in video games history if Dreamcast could had officially emulated the previous SEGA Arcades boards!

Supermodel Technical info:
Supermodel was written by Bart Trzynadlowski and Nik Henson. It is based on the findings of the original Supermodel effort by Ville Linde, Stefano Teso, and Bart from 2003. The PowerPC emulator is originally by Ville Linde and the Sega Custom Sound Processor emulator was donated by ElSemi. Development began in January 2011.

Supermodel incorporates code from the following projects:

GLEW: The OpenGL Extension Wrangler Library, for managing OpenGL 2.x support.
zlib, for ZIP file support.
Musashi by Karl Stenerud, a Motorola 68K emulator.
YAZE-AG: Yet Another Z80 Emulator by Andreas Gerlich, for its Z80 emulator by Frank D. Cringle.
Amp by Tomislav Uzalec, for decoding MPEG Layer 2 audio.

If we ask about DC capability to emulate Model 3 maybe we should start comparing the two hardwares:
Sega Model 3: Processor 66Mhz / Audio RAM 1Mb / Vídeo 8Mb RAM 66Mhz.
Sega Dreamcast: Processor 200Mhz / Audio RAM 2Mb / Video RAM 8Mb 100Mhz

Not sure if more arcade games would had helped the Dreamcast. At that time Playstation was doing well because of games like Metal Gear and FF, maybe in Japan Arcade games could have helped more.