Could Dreamcast et al handle this/that game/effect? *DC tech retrospective *spawn

[Cloofoofoo]

Maybe that's what soul reaver 2 was using. They said they were confident they could match ps2 just fine!

Very very very much doubt it. Probably very custom work. Remember ninja wasn't available to us devs. The pc and DC version were developed by different devs than the PS2 version .

 

[Shifty Geezer]

No kidding. Ninja1 docs on sdk9 ( last usa sdk for DC that even had ninja stuff before they completely removed it ) were horribly outdated and didn't even reflect the ninja API calls( sc2 guy said by looking at the headers he realized it's more of ninja 1.5 since it had some improvement but it still wasn't anywhere near nj2). Also told usa devs to ignore ninja stuff and just focus on on using kamui2. ( Ironic since Sega themselves used nj heavily ) .

Was that...because they wanted to hamper Western games and have Japanese games look better?

 

[davis.anthony]

Ordered a DC today off eBay as I've not had one for a long time.

Can't wait to smash my bother in Power Stone 2 like I used to

 

[sheathx013]

I have recorded Soul Reaver 2 running on Neon 250 (November 1999 drivers) with a PIII 800. Believe it or not it runs above 60FPS most of the time. From comments I've seen online the PIII 800 might be LESS capable than the SH4's FPU, and the Neon 250's 3D side is potentially weaker than the Dreamcast's PVR-2 due to smaller tiles. That said, in Win98SE I'm seeing equal or higher performance with the Neon 250 in this setup than we saw in launch titles like Expendable or Incomming, or WinCE titles like Soldier of Fortune, Half Life, Hidden and Dangerous and Sega Rally 2. It's too bad Sega GT was such an unoptimized PC port, I'd have loved to see it on this setup.

Anyhow, here's the AirForge link. I'm sure later drivers could have dealt with the slowdown layered transparencies cause with the Air forge, or the Light Reaver's artifacts on water and 2D objects.

You can see all of the Soul Reaver 2 Neon 250 videos at the bottom of this Playlist:

PowerVR - PCX-2, Neon 250

PowerVR Generation 1 and Generation 2 videos, focusing on miniGL and SGL compatibility. Neon 250 is new by Anthony Zxclxiv. Matrox m3D is my original from 19...

www.youtube.com

 

[function]

...the Neon 250's 3D side is potentially weaker than the Dreamcast's PVR-2 due to smaller tiles.

There's some old posts from SimonF who worked on the CLX (PowerVR2DC) where he talked about some of the differences:

AFAIK, PVR250/Neon250 had the same shading capabilities as CLX. The only differences I can recall are that

• CLX had a bigger tile size
• CLX Had better support for palette textures
• CLX Had some custom fading tricks that were often in arcade systems
• CLX acted as the memory controller for the DC
• PVR250 had VGA/2D engine and a PCI interface.

There might be others but it was a long time ago.

I think there was also some talk about some features being cut from Neon 250 hardware to keep die area down and make PC cards more cost competitive.

Other titbits I've found from looking at Simon's posts are that CLX could have clocked a lot faster, and that there were untapped features on the GPU (even by Sega):

That figure is a hard limit in the setup engine of the CLX2 chip. They only way to go faster is to overclock it! (Actually, AFAIK, from experiments the CLX2 had about 30% headroom but that meant the SH4 would also have to be overclocked by the same margin which was not possible).

To Marconelly,
I'm going to be blunt with you. AFAIK you are not, nor have ever been, an employee of PowerVR and so I don't expect that you know the internals of CLX. When I say that there were untapped features in the chip, I mean it. It takes time for developers to learn the intricacies of any system and there wasn't sufficient time. No, not even for Sega.

Modders actually found that they could overclock the SH4 in the DC upto something like 240 mhz without direct cooling and in one case something like 270mhz with. Would probably have affected yields though.

A DC with a 125 mhz GPU and a 250 mhz CPU would definitely have been a possibility though, especially with global launches across 1999 and not going early in Japan only (ironically the only place where the Saturn was still doing okay). Sony used a far more heavyweight cooling system with the PS2, and also something like double the power. It's fun to speculate about a Dreamcast designed to be used with a simple but effective cooling system like the Gamecube.

The DC CPU was mostly occupied with none transform and lighting work (I think CrazyAce said it was something like >80% typically) so a CPU upclock would have disproportionately allow for more polygons or more light sources. Possibly the reason that DoA2 has such high polygon counts is that being a 1v1 fighting game an a-typical high amount of CPU time can be used for transform and lighting.

The DC had amazing tech but Sega had no money. If Sega had been able to target a $300 launch system, and developers had pushed into its full featureset for 5+ years lifespan I'm certain it would have stood the test of time pretty well.

 

[sheathx013]

I've got the prior discussions bookmarked and saved in case they disappear. I keep meaning to do a write up of all of the benchmarks I've done with the Neon 250 and other 1999 cards, but I keep discovering new things and expanding the research instead. I think the dev kit scenario must have more nuances than east west relations as well. Like western devs and eastern devs giving different feedback along with the deal with Microsoft and WinCE affecting the western side. But I'd rather stick to the fact we know, which is PowerVR 2 does the same games at 60-75FPS, with dips down below 30 in certain situations. So a locked 30FPS Dreamcast game actually did have a ton of headroom.

 

[Sonicfan]

There's some old posts from SimonF who worked on the CLX (PowerVR2DC) where he talked about some of the differences:



I think there was also some talk about some features being cut from Neon 250 hardware to keep die area down and make PC cards more cost competitive.

Other titbits I've found from looking at Simon's posts are that CLX could have clocked a lot faster, and that there were untapped features on the GPU (even by Sega):



Modders actually found that they could overclock the SH4 in the DC upto something like 240 mhz without direct cooling and in one case something like 270mhz with. Would probably have affected yields though.

A DC with a 125 mhz GPU and a 250 mhz CPU would definitely have been a possibility though, especially with global launches across 1999 and not going early in Japan only (ironically the only place where the Saturn was still doing okay). Sony used a far more heavyweight cooling system with the PS2, and also something like double the power. It's fun to speculate about a Dreamcast designed to be used with a simple but effective cooling system like the Gamecube.

The DC CPU was mostly occupied with none transform and lighting work (I think CrazyAce said it was something like >80% typically) so a CPU upclock would have disproportionately allow for more polygons or more light sources. Possibly the reason that DoA2 has such high polygon counts is that being a 1v1 fighting game an a-typical high amount of CPU time can be used for transform and lighting.

The DC had amazing tech but Sega had no money. If Sega had been able to target a $300 launch system, and developers had pushed into its full featureset for 5+ years lifespan I'm certain it would have stood the test of time pretty well.

Doa dies have a 2v2 mode where it's possible to see all four on screen for a moment. Though, that's on the small practice stage. I can't wait to see if Esppiral can integrate the tag mode into the ps2 extra stages, too!

 

[Sonicfan]

I have recorded Soul Reaver 2 running on Neon 250 (November 1999 drivers) with a PIII 800. Believe it or not it runs above 60FPS most of the time. From comments I've seen online the PIII 800 might be LESS capable than the SH4's FPU, and the Neon 250's 3D side is potentially weaker than the Dreamcast's PVR-2 due to smaller tiles. That said, in Win98SE I'm seeing equal or higher performance with the Neon 250 in this setup than we saw in launch titles like Expendable or Incomming, or WinCE titles like Soldier of Fortune, Half Life, Hidden and Dangerous and Sega Rally 2. It's too bad Sega GT was such an unoptimized PC port, I'd have loved to see it on this setup.

Anyhow, here's the AirForge link. I'm sure later drivers could have dealt with the slowdown layered transparencies cause with the Air forge, or the Light Reaver's artifacts on water and 2D objects.

You can see all of the Soul Reaver 2 Neon 250 videos at the bottom of this Playlist:

PowerVR - PCX-2, Neon 250

PowerVR Generation 1 and Generation 2 videos, focusing on miniGL and SGL compatibility. Neon 250 is new by Anthony Zxclxiv. Matrox m3D is my original from 19...

www.youtube.com

Dude! This is so interesting! I wonder if this cide could somehow be transferred over to DC. I bet it would look better lighting and shadow-wise using DC hardware features

 

[Cloofoofoo]

Was that...because they wanted to hamper Western games and have Japanese games look better?

It's hard to tell to be honest. In practice letting western devs use kamui 1/2 only could be a good idea since they can interface alot closer to the hardware ( especially tnl since they HAVE to write it themselves) . Reality quality of code , budget, time or even motivation could greatly affect it. I mean one dev might take the time and right very custom sh4 tnl code or maybe some one else might just do sloppy c code and decide it's good enough( you ever read the tony hawk post mortem for dc interview ? They admitted to not using the CPUs fpu). Though the same can be said for naomilib or ninja at least you have a baseline and helper tools to speed up the process rather than start from scratch. And I've heard there was the darkness layer, the lowest interface for DC but I don't know much about or if any games used it.


Sega themselves pretty much used ninja and Naomi libs for their own titles even the biggest hits , so what does that tell you. On the ninja side you got sonic adventure 1/2, skies of Arcadia , space channel 5 1/2, blue stinger, d2, rez. On the naomilib side you got Daytona 2001, virtual tennis 1/2, crazy taxi 1/2, f355 challenge , DEAD OR ALIVE 2 , sega marine fishing and so on. Just stepping back and looking at this incomplete list you could see how it does seem like they were hoarding stuff for the Japanese . Makes no sense depriving the western side of even giving them the choice .

 

[Cloofoofoo]

I have recorded Soul Reaver 2 running on Neon 250 (November 1999 drivers) with a PIII 800. Believe it or not it runs above 60FPS most of the time. From comments I've seen online the PIII 800 might be LESS capable than the SH4's FPU, and the Neon 250's 3D side is potentially weaker than the Dreamcast's PVR-2 due to smaller tiles. That said, in Win98SE I'm seeing equal or higher performance with the Neon 250 in this setup than we saw in launch titles like Expendable or Incomming, or WinCE titles like Soldier of Fortune, Half Life, Hidden and Dangerous and Sega Rally 2. It's too bad Sega GT was such an unoptimized PC port, I'd have loved to see it on this setup.

Anyhow, here's the AirForge link. I'm sure later drivers could have dealt with the slowdown layered transparencies cause with the Air forge, or the Light Reaver's artifacts on water and 2D objects.

You can see all of the Soul Reaver 2 Neon 250 videos at the bottom of this Playlist:

PowerVR - PCX-2, Neon 250

PowerVR Generation 1 and Generation 2 videos, focusing on miniGL and SGL compatibility. Neon 250 is new by Anthony Zxclxiv. Matrox m3D is my original from 19...

www.youtube.com

Nice, very interesting. The guys who did the windows ce sdk demo "defense commander" pretty much stated that Dreamcast is equivalent to Pentium 3 500 and a decent grafx card( they tested tnt , voodoo3 , g400). This does pretty inline with those assumptions.

 

[Sonicfan]

I wonder what visual concepts used. If it wasn't one of the Japanese kits, they did a really great job on working around those limitations! I would have loved to see what they could do with full access! Maybe higher mesh players, actual 3d bg players (might have to be static) and use of rtt for the giant screens

 

[sheathx013]

Nice, very interesting. The guys who did the windows ce sdk demo "defense commander" pretty much stated that Dreamcast is equivalent to Pentium 3 500 and a decent grafx card( they tested tnt , voodoo3 , g400). This does pretty inline with those assumptions.

I'd bet in Windows CE a PIII 500 is about as far as performance can go. That seems about in line with what I've benchmark so far. That also happens to be the fastest PC processor around while we were waiting for the western launch. So they might have had nothing else to compare to.

I can drop the PIII 800 to 533mhz with a 66mhz FSB. But the Neon 250 still peaks at 75FPS in games that were limited to 30FPS on the DC, with more and longer dips.

 

[xaeroxcore]

No kidding. Ninja1 docs on sdk9 ( last usa sdk for DC that even had ninja stuff before they completely removed it ) were horribly outdated and didn't even reflect the ninja API calls( sc2 guy said by looking at the headers he realized it's more of ninja 1.5 since it had some improvement but it still wasn't anywhere near nj2). Also told usa devs to ignore ninja stuff and just focus on on using kamui2. ( Ironic since Sega themselves used nj heavily ) . And I am guessing naomilib is something us devs wouldn't be made aware it was even an option since it's not included in the sdk by default.

Ninja2 simply wasn't used by anyone simply because it came out too late, around September 2000 which means in a few months dc would be discontinued and most devs working on anything would likely scrap their plans. It's a shame , the feature and performance increase would have made even the common game capable of using most effects easily this time around and even increased polygon output. Not to mention the flexibility of the file format, you could pretty much animate everything in the modeling program and export light info/animation , bake vertex colors , set blending levels and materials ( bump mapping and env map included) . Have more limited framebuffer effect access but far less costly / more usable.

So this is one of the reasons why there is an almost generational disparity between the top graphical DC titles and most part of its catalogue...Sega themselves screwing things up Probably with Ninja 2 we would had ended having up more complex games on par to what was on multiplattform PS2 games at the time.

Also i wonder wich SDK was used to create Test Drive Le Mans and Headhunter?

In terms of visuals sure, in terms of performance I am pretty sure it would run most of the time at 30 FPS.

Even at 30 fps and the DC SR1 Raziel model, i'd love to play SR2 on my DC.

 

[see colon]

So this is one of the reasons why there is an almost generational disparity between the top graphical DC titles and most part of its catalogue...Sega themselves screwing things up

I think we would always have this gap, because Dreamcast existed in in a mid-generation timeframe. A fair portion of the Dreamcast library are ports from N64 and PS1.

 

[Cloofoofoo]

So this is one of the reasons why there is an almost generational disparity between the top graphical DC titles and most part of its catalogue...Sega themselves screwing things up Probably with Ninja 2 we would had ended having up more complex games on par to what was on multiplattform PS2 games at the time.

Also i wonder wich SDK was used to create Test Drive Le Mans and Headhunter?

Even at 30 fps and the DC SR1 Raziel model, i'd love to play SR2 on my DC.

Well even being stuck on ninja 1 there was nothing stopping Japanese third parties from making complex and even fairly detailed games. I think it's more of an investment issue and confidence issue of companies with low expectations of DC. Ninja2 would have certainly help push things along graphically but sega needed support heavily from companies people loved like Konami and namco( very good art from both companies) , neither invested heavily on dc. I'll give you an example below, some screen shots of a 1999 ninja1 game , d2 and gifs of a 2002 ninja1 game Sakura wars 4( thx for whoever posted these on discord). As you can see they could have always made nicely animated / frame buffer effect / detailed games had they wanted to invest time and money. There's no magic bullet, the console just need more time, effort and support.

I already told u western developers used kamui2 . I am 100% that's what those games used.

Ninja1 D2 1999:

Ninja1 Sakura wars 4 2002( uses radial blurs/motion blues , even REFRACTION)

 

[Nesh]

Well even being stuck on ninja 1 there was nothing stopping Japanese third parties from making complex and even fairly detailed games. I think it's more of an investment issue and confidence issue of companies with low expectations of DC. Ninja2 would have certainly help push things along graphically but sega needed support heavily from companies people loved like Konami and namco( very good art from both companies) , neither invested heavily on dc. I'll give you an example below, some screen shots of a 1999 ninja1 game , d2 and gifs of a 2002 ninja1 game Sakura wars 4( thx for whoever posted these on discord). As you can see they could have always made nicely animated / frame buffer effect / detailed games had they wanted to invest time and money. There's no magic bullet, the console just need more time, effort and support.

I already told u western developers used kamui2 . I am 100% that's what those games used.

Ninja1 D2 1999:

Ninja1 Sakura wars 4 2002( uses radial blurs/motion blues , even REFRACTION)

The PS2 was definitely much harder to program and during it's launch it's libraries were a mess to non existent. The DC's tools and libraries must have been at a much better shape. Despite that, some first PS2 games were competing easily with DC's latest games and some were doing stuff that the DC didn't match.

That shows that, there must have been huge power and capability disparity between the two.

Regardless, considering DC and PS2 being drastically different, with DC's much higher VRAM size, higher resolution output, and AA implementation there is a chance that the DC in some occasions might have exhibited advantages were the PS2 wouldn't be able to match.

It is very very expected that DC's potential were never exploited to their fullest and we might have seen a lot of surprises if it reached maturity properly. I doubt though that it would have ever reached parity with PS2's best.

 

[Sonicfan]

I think we would always have this gap, because Dreamcast existed in in a mid-generation timeframe. A fair portion of the Dreamcast library are ports from N64 and PS1.

True, but remember ps2 got a fair number of ports from 5th gen, but they updated them dramatically. I'm sure it's more of a budget/trust thing; 3rd parties weren't gonna put the effort into huge DC upgrades, unfortunately...

 

[Sonicfan]

The PS2 was definitely much harder to program and during it's launch it's libraries were a mess to non existent. The DC's tools and libraries must have been at a much better shape. Despite that, some first PS2 games were competing easily with DC's latest games and some were doing stuff that the DC didn't match.

That shows that, there must have been huge power and capability disparity between the two.

Regardless, considering DC and PS2 being drastically different, with DC's much higher VRAM size, higher resolution output, and AA implementation there is a chance that the DC in some occasions might have exhibited advantages were the PS2 wouldn't be able to match.

It is very very expected that DC's potential were never exploited to their fullest and we might have seen a lot of surprises if it reached maturity properly. I doubt though that it would have ever reached parity with PS2's best.

Yeah, i think it was cause it was simply easier to bruteforce ps2 to do higher meshes and fx. DC was easier to program, but the hardware was more nuanced if yiu wanted to get a ps2 level of detail out of it. The tools weren't there in the west for that to happen, plus the teams were all smaller, lower budgets, and trying to get as many games out as possible before PS2's launch all limited the scope of the DC's library...

 

[Shifty Geezer]

So, what's the homebrew scene like on DC? We've seen some incredible old-hardware stretching games like the Atari 7800 or ZX Spectrum. Is anyone making a DC pushing title just for fun? I suppose maxing an 8 bit machine is far, far less work than maxing out a complex console.

 

[Nesh]

Yeah, i think it was cause it was simply easier to bruteforce ps2 to do higher meshes and fx. DC was easier to program, but the hardware was more nuanced if yiu wanted to get a ps2 level of detail out of it. The tools weren't there in the west for that to happen, plus the teams were all smaller, lower budgets, and trying to get as many games out as possible before PS2's launch all limited the scope of the DC's library...

I wouldn't associate PS2 with bruteforce at all. You couldn't just dial up everything and expect it to work. PS2 was the opposite of that. The machine was designed around a premise which required developers to know how to manually squeeze it's capabilities, like using it's VU's, managing the bandwidth of tiny VRAM to stream in and out textures, etc. AA wasn't supported by the hardware by default either. It's effects weren't coming for free. They had to be manually handled and a lot of effects required smart use of geometry multipasses.

If there was one console where you could bruteforce everything was XBOX, where it had the processing power, the memory and the default hardware support to dial up everything.