• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Was the Dreamcast actually powerful at launch? Or the beneficiary of no competition?

Was the Dreamcast a powerhouse at launch?

  • No

    Votes: 111 11.3%
  • Yes

    Votes: 867 88.7%

  • Total voters
    978

Fafalada

Fafracer forever
So the number of inference maths calculations per pixel I would still expect to be around the 20 maths ops you mentioned
Yea but that was with GS compute, where flexibility is really limited, as it could only really execute 8bit component multiply-add, albeit a lot of them.
So not really applicable to FSR2 which just runs general shader compute (you'd need VU code to emulate that), or DLSS which does run 8bit ops, but also needs other math.

Plus with these super low resolution targets, what would even be the point? Unless we're talking tech demo stuff like that Quake engine on Spectrum that still runs in low single digits
framerates - from that perspective, sure a lot 'might' be possible but we're well into the realm of different kind of extremes (not really topical to this thread).
 

PaintTinJr

Member
Yea but that was with GS compute, where flexibility is really limited, as it could only really execute 8bit component multiply-add, albeit a lot of them.
So not really applicable to FSR2 which just runs general shader compute (you'd need VU code to emulate that), or DLSS which does run 8bit ops, but also needs other math.

Plus with these super low resolution targets, what would even be the point? Unless we're talking tech demo stuff like that Quake engine on Spectrum that still runs in low single digits
framerates - from that perspective, sure a lot 'might' be possible but we're well into the realm of different kind of extremes (not really topical to this thread).
The point would be that the IQ oer pixel at 160x120 native could be much higher, and because it is such a small resolution frees up memory, memory bandwidth and could allow for doubling the OPs per native pixel for fx, and increase the level of tessellation, and geometry density per pixel given the gain in redundancy(culling, overdraw, depth testing, blending) and allow for more textures, for multi-texturing, because the textures could then be a lower resolution when targeting just 160x120, (or whatever the actual 1/4 area equivalent of widescreen 512x384 was, maybe 256x144), effectively leading to results way cleaner and more advanced beyond what was achieved at the time, because the reconstruction alleviated the memory limitation.

The inferenced reconstruction could also be trained with baked in high quality AA and texturing AF, maybe even quality world/model space AA using an accumulation buffer, that would really improve the IQ, and even if the reconstruction introduce 30% noise, the end result should still be far better than anything native PS2 did at 512x288 and potentially lead to the same image quality but at double the frame-rate, so long as the reconstruction could work in the processing resources saved, from rendering lower native. In the same way we might ask could the Dreamcast of done this if it had similar processing versatility?
 

Alexios

Cores, shaders and BIOS oh my!
Reconstruction benefits diminish the lower the original resolution is and that's when talking about it in modern terms like 1080p or 720p, I don't imagine much benefit with an input res that low.
 

Fafalada

Fafracer forever
The point would be that the IQ per pixel at 160x120 native could be much higher
I get that but you're downgrading the base resolution so low that it will degrade the end-IQ significantly anyway - no matter what upscaling you apply to it.
Also don't forget reconstruction still has to write out higher resolution - so this would take significant amount of frame time in of itself to run on PS2 class hardware. We're talking 100x less compute resources compared to a PS3. 1000x compared to a PS4 Pro.
Even at 320x200 - resolution delta to 1080p is 'only' 27x.

and allow for more textures, for multi-texturing, because the textures could then be a lower resolution when targeting just 160x120
Not if you want them to look good when upscaled. This is actually one of the hidden gotchas with reconstruction techniques - texture resolution must still be high, and worse - it requires negative mip-bias to ensure you don't lose important details in reconstructed image. Direct consequence is that 1080p rendering for DLSS/FSR2 is actually more expensive than 1080p Native - before you upscale either of them to 4k.

The inferenced reconstruction could also be trained with baked in high quality AA and texturing AF, maybe even quality world/model space AA using an accumulation buffer
That's just asking for magic. All current models do is a 'smarter' version of history resolve (keeping multiple frames of subsamples to reconstruct from). They aren't creating data that doesn't exist to the picture (spatial AI upscalers attempt this - and they tend to all look terrible most of the time, and not at all temporally stable).
 

PaintTinJr

Member
I get that but you're downgrading the base resolution so low that it will degrade the end-IQ significantly anyway - no matter what upscaling you apply to it.
Also don't forget reconstruction still has to write out higher resolution - so this would take significant amount of frame time in of itself to run on PS2 class hardware. We're talking 100x less compute resources compared to a PS3. 1000x compared to a PS4 Pro.
Even at 320x200 - resolution delta to 1080p is 'only' 27x.
I get your point about the compute scale in raw term comparisons, but the inferenced target of a upscaled, sharper, more fx PS2 game is still orders of bandwidth much lower than even an early PS3 game, say comparing PS2 VF4Evo to VF5b. And with my estimate of maybe 20 per pixel maths ops for inference, that would be about 22MFLOP/s - assuming the game is 30fps interlaced(256x144) like MGS3- of the 600MFLOP/s (if I have my numbers correct for the correct unit), and if that is too conservative, even at 80 per pixel that is still only 1/6th, which still makes me think it isn't completely beyond the realms of possible on processing grounds.
Not if you want them to look good when upscaled. This is actually one of the hidden gotchas with reconstruction techniques - texture resolution must still be high, and worse - it requires negative mip-bias to ensure you don't lose important details in reconstructed image. Direct consequence is that 1080p rendering for DLSS/FSR2 is actually more expensive than 1080p Native - before you upscale either of them to 4k.
Personally, I'm surprised that the techniques even with training and motion vectors are that poor, when in effect the regression inference equations are just substituting a set of lookup values into tables and returning a superior rendition of the data. But your point is well made, and I guess that would kill all the memory texture savings of a 256x144 native render. Hopefully as it would be order less work than the native 512x288 to begin with, by being 1/4 of the pixels, and that saving multiplied by the ops per pixel, I think that could still be okay, so long as the extent the increase in native render input as the source for inference was less than 3x normal native 256x144
That's just asking for magic. All current models do is a 'smarter' version of history resolve (keeping multiple frames of subsamples to reconstruct from). They aren't creating data that doesn't exist to the picture (spatial AI upscalers attempt this - and they tend to all look terrible most of the time, and not at all temporally stable).
Fair enough, my own use of regression was for a texture compression paper I abandoned writing at the launch of the gen with everyone going the rate adaptive way. so I've assumed the reconstruction technique were far better than what you describe, and were at least as effective as the range of X1 and XR chips in Sony's flagship TVs, that I assume are all spatial AI upscalers, which would definitely mean better algorithms would be need first for some of the gains to happen.

Thanks for your replies, really great info for me to chew on for theorising about a PS2 maybe, or probably not being able to use FSR2 for some gains :)

/edit

I went and looked deeper at the implementation details of the techniques and hadn't realised they are all effectively supersets of the TAA technique, and that TAA is needed on the source input too, prior to FSR, etc, so even more compute overhead.
 
Last edited:
And goes down to like 15 fps for it so maybe it shouldn't have been 480p🤷‍♂️

No man, you are making a wrong comparison and this will lead you to have a wrong understanding. CT2 is not the reference for GTA 3 on Dreamcast, they are 3d games but not 3D games of the same category.

Do you know a game called Banjo Kazooi? In this game there is an open 3D area, however when entering it the camera moves away, filming in 3/4 but after you approach a closed area, a house, a cave, the camera will adjust in 3D mode on the character's back . The technical reasons for this are to relieve the cpu, the same principle can be better understood in the port of Silicon Valley and Glover from the N64 to the ps1 where the camera remains distant and at 3/4.

GTA 3 (2001)'s competitor on Dreamcast is Omikron (2000). If you cannot distinguish the difference between Omikron and Crazy Taxi 2 then you will not have understood that not every 3d game is a proper 3d game.



 
Last edited:

Alexios

Cores, shaders and BIOS oh my!
No man, you are making a wrong comparison and this will lead you to have a wrong understanding. CT2 is not the reference for GTA 3 on Dreamcast, they are 3d games but not 3D games of the same category.

Do you know a game called Banjo Kazooi? In this game there is an open 3D area, however when entering it the camera moves away, filming in 3/4 but after you approach a closed area, a house, a cave, the camera will adjust in 3D mode on the character's back . The technical reasons for this are to relieve the cpu, the same principle can be better understood in the port of Silicon Valley and Glover from the N64 to the ps1 where the camera remains distant and at 3/4.

GTA 3 (2001)'s competitor on Dreamcast is Omikron (2000). If you cannot distinguish the difference between Omikron and Crazy Taxi 2 then you will not have understood that not every 3d game is a proper 3d game.

I was just referring to your 480p praise, why derail? Camera angles that reduce lod or what is rendered on screen or whatever have absolutely nothing to do with it, it's not like I compared GTA2 to III and said all it is is a camera angle switch, lol. I did discuss CT2 in relation to GTA but not there, here:
Same line of thinking I've been following since the 00s, maybe it would have ended up as a very differently developed game if tailored to DC's architecture but any GTAIII-like seems more than doable when looking at Crazy Taxi (or better yet Crazy Taxi 2 which VASTLY improved the scenery pop in) and its blazing fast, unwavering 60fps gameplay. Sure, it doesn't have all GTA gameplay elements but they clearly had room for a lot of optimization to add more, like going for 30fps (or GTAIII's 20 or less at times), reducing the traffic density and the gameplay speed too. They did cool hair too 🤷‍♂️

The hair bit was obviously not about that one bald character but others like Slash & various NPCs, nice use of alphas for such a minor detail given the game's style and the era not being exactly known for great looking hair in games so it made an impression to me. Anyway I was commenting on this:
To add something to the topic other than Monkey Balls before I go to bed, found this tweet chain interesting:

Omikron was a janky ass unfinished if ambitious game nobody wanted anywhere, of course a port is even more so. You don't even drive in it. I'd say it's more like Deus Ex: Invisible War, trying to ape a city with tech & design not really suited to it & running like ass on PCs of the era too unlike GTAIII+

Hell, it's the first and jankiest of David Cage/Quantic Dream. Their next game that also happened to get on PS2 was where they first went with his QTE/cut-scenes-as-gameplay style, probably realizing their tech, talent and know how just aren't suited to conveying what Omikron seemed to attempt...

Random idiotic youtuber clickbait isn't a way to make a point, should I expect more late replies only to things you searched a gotcha for? You ignore my PSA about Dreamcast footage, your video is with the Dreamcast mod causing this in Crazy Taxi as I linked before (but Omikron is shit anyway) lol:
vs normal Dreamcast

Don't be so adamant about stuff you have no first hand experience or knowledge of and search for gotchas on google with confirmation bias making such mistakes (if it wasn't deliberate as it's from the same channel I linked before and the mod is in the video title, serial port streaming is just slow).

You might as well compare the Shenmues with GTA, lol, Omikron is far more similar to them with having fighting and adventure elements too (thankfully no crappy FPS in them) and they clearly have framerate issues as well, but at least they justify it with how they look and are generally better made.

Anyway, nobody said every DC game is great or every 480p game on it runs flawlessly to think digging up a bad and/or bad performing one makes any good argument. There are bad games everywhere, I commented on a PS2 crown jewel you praised for its rare 480p negating all of Dreamcast 🤷‍♂️
 
Last edited:

RobRSG

Member
Maybe it's possible on the Dreamcast with 30fps, reduced draw distance and some cuts in alpha effects.Each game has a reason, the dev's choice to make Gungriffon this way was to explore long draw distances without fog at 60fps. In the later stages there is fog, but with an artistic purpose. Gungriffon Blaze is part of the initial wave of the ps2 with a Dreamcastish look.

We cannot underestimate 60fps, the Dreamcast cannot reach 60fps even in some advanced 5th generation games.

shadowman 30fps (looks amazing)
Soul Reaver 60 (with frame drop)
star wars ep 1 race 30fps
Tony Hawk pro Skater 30fps
Re-Volt 30fps with code to reduce draw distance
Rayman 2 60fps ( almost no geometry increase)
Dino Crisis (3D engine) 30fps
RE3 and AitD 60fps
Vigilante 8: 2nd Offense 60fps

As we can see, some games have had an upgrade in polygon count, becoming full Dreamcast games, the character Tony Hawk has 1,500 tris and Soul Reaver Raziel 1,222 tris. Shadowman was also upgraded so 60fps was not possible.

Floigan Bros
Illbleed (with frame drop)
KAO the Kangaroo
Outtrigger
Sonic Adventure 2
Spawn: In the Demon's Hand
Super Magnetic Neo
Unreal Tournament 60fps fake
Soul Fighter
Pen Pen TriIcelon

As we can see, some games despite being complete Dreamcast have one or another intelligent trade off Sonic Adventure 2 holds 60fps better in the on-rails stages, Kao The Kangaroo has only one enemy at a time in the scene, Soul Fighter has shadows but collisions are bad, Spawn also doesn't have shadows like in Shadowman, alpha simulations are simple, Outtrigger is basically Spawn. Sonic 2 looks the best

the best and most impressive Dreamcast games are 30fps so these early ps2 games running at 60fps was a huge achievement in those days.
Nice fiction you wrote here.
 

Fafalada

Fafracer forever
I did discuss CT2 in relation to GTA but not there, here:
Which is - to be fair - a really poor comparison. CT is 'very' light game simulation wise (eg. one rigid-body on screen 95% of the time - outside of occasionally hitting boxes/chairs that drop the framerate badly anyway). Not to mention even several DC circuit racers had larger geometry (and maybe texture) budget for their levels than the city in CT is.

GTA was far from a particularly optimized experience (or visually impressive) - but when you're porting you don't get to rewrite the entire game for efficiency. Could it run on DC? Sure - they squashed it down to PSP which had a lot of similar constraints, but it definitely didn't get away unscathed in the process, it was ported years after last PS2 release (so much more modern tooling) and there's a rather large gap to the console version(s) still in terms of visuals and simulation fidelity alike (can't remember if framerate was better or not).
 

Alexios

Cores, shaders and BIOS oh my!
Which is - to be fair - a really poor comparison.
No. I didn't say it's 1:1 which you ignore to repeat it in a more negative way. I'll take the dev's words over yours. I didn't say the game would be identical (even said "GTAIII-like") had it been made with DC's architecture (too) in mind but it would end up better than a port, which may be possible too.

Once again CT isn't even the best form of it, CT2 is vastly improved with longer draw distances and still 99% of the time flawless framerate (liar, CT doesn't "badly drop" often as videos show and it's not like your precious PS2 does it - CT or GTA for that matter - better so it may be a software issue).

GTA seems to have survived the technical conversion to PSP about as good as Crazy Taxi did since you mention that, perhaps with a more severe performance drop (from its original 60fps) in the latter vs more cuts in other visual elements in the former (overall complexity), roughly on par in the end.
 
Last edited:

Fafalada

Fafracer forever
No. I didn't say it's 1:1 which you want to ignore to basically repeat it in a more negative argument.
I didn't argue negatively (hence the PSP GTA games analogy) - I just said CT is a bad comparison as it tells you literally nothing other than 'CT can run on DC'.

And once again CT isn't even the best form of that, CT2 is vastly improved with longer draw distances and still 99% of the time flawless framerate
It's still apples to tomatoes comparison. Just like that poster with whatever the other game was that you already shot down.
It's equivalent to making an argument that PS2 had open-world games running at some 7x geometric complexity of GTA3 (20x of CT) at 60fps. That doesn't say PS2 should have run GTA at 60 or better visuals - as obviously it didn't.

GTA seems to have survived the technical conversion to PSP about as good as Crazy Taxi did since you mention that
Yea cheap ports vs real budget and all that. GTA games on PSP were new games built for the system, probably with more development budget than all CT ports combined, and also made more money than all console versions of CT combined.
 

Alexios

Cores, shaders and BIOS oh my!
Your random arguments about imaginary games of 20x complexity have nothing to do with me. The what made money discussion, if that's what you wanna talk about, was over long ago and it wasn't the Dreamcast, lol. Idk what you're talking about, certainly nothing related to anything I've said🤷‍♂️

Again taking GTA dev's word (btw he's the only programmer - technical director in fact - in that discussion, indirect or not, as King is clearly a big business guy and of course business would have decided at the time if they would continue on DC or not and the answer was obvious) over this new rant.

CT(2) seems to have a solid engine allowing for a large, detailed, streaming cityscape with high quality, varied textures, a large draw distance and many interactive objects with physics (whether arcadey or not the physics are in, if anything for a more realistic setting they'd need some toning down) whether that's the traffic or most decoration objects in the game and while the game is basic in some ways (there's no complex AI from what we can see as it doesn't even have enemy drivers a la Driver, though the traffic AI is one of the things they had to reduce for PSP multiplayer so maybe it's bloated and ripe for optimizations too) there is a LOT of room and things to reduce/remove to then add such elements for the DC CPU & GPU to process. To put it simply. If you don't see how all this can relate to a GTA-like open world game then you're not the authority you and PaintTinJr think.

Of course nobody's saying you can just import GTA's map and gameplay in Crazy Taxi's engine and play, it's just an indication you could have a "GTAIII-like" work on Dreamcast as I said before, not even necessarily in this engine (though that is very much plausible too, again they have many needed game elements already done for it and other games in the past did change and add gameplay in such ways, for example Driver 2 added the on foot/car switching sections over Driver for the original PlayStation and that was with many general/visual improvements rather than more concessions on top as optimization and ever improving know how simply go a long way), rather than a sandbox open world action game that looks and runs good for the era being impossible on Dreamcast (as other posters previously proclaimed you can't even stream game data on Dreamcast to make a large city possible and only the PS2's DVD drive enabled that, when even the first Crazy Taxi employs such methods as we literally just saw what happens when the streaming is affected for any reason with that hardware mod video) because it didn't happen but some shitty games did happen, lol.

Feel free to disagree I guess but I don't see any real argument against any of this, just some weird fixation on semantics and mistranslating my arguments to things I didn't really say (I know CT isn't GTA and didn't say otherwise). Hell, you're the one who said a GTAIII port is very possible on the Dreamcast as on PSP, I wasn't talking ports of any finished game, just a theoretical GTAIII that also had Dreamcast's hardware in mind during development or a different game altogether from Sega or another developer that could take advantage of the system to provide solid open world action.

In the end you agree I and Obbe are right but find fault in using existing titles to reach this conclusion & ignore Obbe was the GTAIII technical director because you reach the same conclusion not by observing what's been achieved but because, Idk, you read a manual/messed around with an SDK?

As for implying with no evidence the Crazy Taxi port (a port obviously costs less than a ground up game as you say yourself, duh) was some botch job and it could have fared much better, weirdly the Power Stone ports from DC to PSP from a different developer from a different porter did similar🤷‍♂️

The difference to Crusher's arguments regarding PS1 ports on the Dreamcast is that we have in fact seen Dreamcast do better than the worst examples he emphasises both in some of the better ports and with new games, so there we do have the ability to speak based on evidence. Where are yours?
Fafalada said:
We're comparing ports that were made in as little as 3-6 months in some cases to a fully budgeted release on the other side. If you think that 'doesn't' impact the results...
There's a pretty big gulf between more money/time = better results vs a "botched" port and any port obviously wouldn't ever need as much money/time as a new GTA game, even with many preexisting assets, so, maybe at some point you can stop being so damn disingenuous and only go after the people speaking positively of DC if you eventually end up conceding yes GTA DC is possible and yes it's right to shoot down Crusher's weird af arguments as I did even less here (again I wasn't speaking of a GTAIII as is port yet you pretend I said CT is just like GTAIII) yet you attacked me over it, lol?
 
Last edited:

Fafalada

Fafracer forever
Your random arguments about imaginary games of 20x complexity
They're not imaginary - I just didn't want derail more with further lists.

As for implying with no evidence the Crazy Taxi port
We're comparing ports that were made in as little as 3-6 months in some cases to a fully budgeted release on the other side. If you think that 'doesn't' impact the results... did I mention I have this bridge in lower Manhattan available that I've been meaning to sell for a good price?
Like this thread spent an unhealty amount of time debating the merits of cheap PS1 ports to DC and why they're not using mipmaps and god knows what - 90% likely it's because they were done by one dev on a shoe-string timeframe. I have 1st hand details on several such ports to DC and why they turned out like PS1 running on bleemcast.
 
Last edited:
The Dreamcast is an enigma, it allows devs to multiply the polygonal budget by 5x in Mortal Kombat Gold while turning off the dynamic lighting present in the ps1, n64 and arcade.
Allows you to improve Tomb Raider by adding shadow and fog but removes Lara's ability to leave footprints in the sand
improves lighting in Dave Mirra BMX but removes shadows, some transparent waters in Rayman 2 have become matte, certainly some changes are artistic.

If Cloofoo posts here to extract the Shadowman model from the Dreamcast version, I think the developers did an excellent job there.
 
Last edited:

Fat Frog

I advertised for Google Stadia
I know it hurts but it's our duty to share the following link again, again and again 🤣😎:

The GTA3 architects chose to make that GTA/Crazy Taxi comparison.

You can write books on the subject but...

That's not our fault. 😘


If you think the GTA3 tech guys are stupid, then have faith, courage, go on twitter and tell them they are talking shit and don't have a clue about the Dreamcast 🤣
 

PaintTinJr

Member
I know it hurts but it's our duty to share the following link again, again and again 🤣😎:

The GTA3 architects chose to make that GTA/Crazy Taxi comparison.

You can write books on the subject but...

That's not our fault. 😘


If you think the GTA3 tech guys are stupid, then have faith, courage, go on twitter and tell them they are talking shit and don't have a clue about the Dreamcast 🤣

I'm more interested in his reasoning about the DC having a "very weird but powerful graphics card" because everything in all the original dev files and current tools say it is very vanilla Opengl 1.4 or less. Only the fog tables and the framebuffer formats were mildly different from a low/mid performance PC GPU of the time.
 

RobRSG

Member
What in my text is fiction?
If you don't have arguments, don't quote me.
Like the parts you suggest the Dreamcast couldn’t handle PS1 ports at 60FPS, because they were close to the limit of the hardware. Honestly, my brain was not ready for it.

You should write a book, it would sit well with the Da Vinci Code and other books aimed at conspiracy nuts. 😅
 
Last edited:

Fat Frog

I advertised for Google Stadia
Doesn't make it any less stupid. Besides that thread is literally two GTA3 guys indirectly arguing with each other how DC could/could not do GTA.

Anyway - because reading is apparently hard - this is what I said, on this page:
Obbe was the Technical director of GTA 3, i doubt we'll find more relevant opinion on the subject...

Hard to read ? Did i mention you before, pawl ? 😁
But now i will:
Go on twitter and tell Obbe his comparison is stupid.
It's pointless to discuss with me ( i admit i just like to tease people 😆) but i would be curious to see a real discussion between GTA3's technical director and a few smart asses from GAF.(I don't necessarly think about you, pawl, but all of you... 🤣 If some of you feel confident enough to discuss with Obbe, you'll have my respect for real 🤝😘)
 
Like the parts you suggest the Dreamcast couldn’t handle PS1 ports in 60FPS, because they were close to the limit of the hardware. Honestly, my brain was not ready for it.

You should write a book, it would sit well with the Da Vinci Code and other books aimed at conspiracy nuts. 😅
This is normal because everything new is strange, you obviously weren't used to reading something like this, but like most Dreamcast fans, you were used to thinking of the console as an untapped power and of games like DOA2 as an example of what the console can do and not (which it actually is) the limit.

Understand, the games you call 'ps1 games' are not ps1 games. Dreamcast's Shadowman is not a ps1 game nor is it an N64 game, do you realize that?the Dreamcast can make ps1 games at 60fps as long as those games don't turn into a Dreamcast game like Shadowman did.
 

RobRSG

Member
Understand, the games you call 'ps1 games' are not ps1 games. Dreamcast's Shadowman is not a ps1 game nor is it an N64 game, do you realize that?the Dreamcast can make ps1 games at 60fps as long as those games don't turn into a Dreamcast game like Shadowman did.
Dino Crisis running at 30 because it is an advanced PS1 game? Do you really think Dino Crisis ran at 30 due to the lack of Dreamcast technical prowess? Suddenly, Code Veronica doesn’t exist anymore?

Rayman 2 looks way better than its N64 and PS1 counterparts, still outputting in 60FPS. Maybe you don’t remember how this game looks in person. Heck, the Dreamcast version also completely destroys the PS2 version.

It falls all back to the “artistic choice” argument Xbox used recently for some its exclusives. If the devs wanted, they could have ran all those games in 60FPS. And Shadowman, THPS, Star Wars Racer comfortably sit into that category, as they don’t look impressive at all when stacked against other quality Dreamcast titles.

See, as long as keep writing fun and absurd stuff like that, I’m happy.
 
Last edited:

PaintTinJr

Member
Obbe was the Technical director of GTA 3, i doubt we'll find more relevant opinion on the subject...

Hard to read ? Did i mention you before, pawl ? 😁
But now i will:
Go on twitter and tell Obbe his comparison is stupid.
It's pointless to discuss with me ( i admit i just like to tease people 😆) but i would be curious to see a real discussion between GTA3's technical director and a few smart asses from GAF.(I don't necessarly think about you, pawl, but all of you... 🤣 If some of you feel confident enough to discuss with Obbe, you'll have my respect for real 🤝😘)
Can you not tell by Fafalada's contributions that they are or were a lead developer that actually implements all the technical stuff? Your comment IMO is both juvenile and disrespectful, and likely to discourage actual industry people to contribute so freely.

Someone being a technical director for a company can mean many things, and without the context doesn't automatically carry the developer knowledge you assume it does, or put them on par with someone that's job it is to actually code real solutions on the hardware.

Based on his description of the DC hardware in his tweet, it doesn't sound like that "technical" director had first hand knowledge of the hardware, and as his comment was about "rendering" - in comparison of CT and GTA3 - and not running the game systems plus rendering, his tweet doesn't actual say what you think it says to begin with to support your assertion.

Him making a comparison of scene complexity for "rendering" at half the CT frame-rate on DC for GTA3 assets is fine, but as Fafalada is eluding to, the games are completely different and in different genres and have vastly different system demands on a piece of hardware. AFAIK GTA games are CPU and memory use heavy, but in the Dreamcast situation AFAIK from what has been said here, the "T"(transformation) part of the T&L pipeline was done on the DC CPU and not the GPU, meaning rendering alone for CT is probably taking CPU resources that will be needed, and that would increase if the frame-rate halved through increased GTA3 geometry, and so would have a negative knock on effect on the GTA game systems running on the DC CPU with the resources left over.
 
Last edited:

Fat Frog

I advertised for Google Stadia
Your comment IMO is both juvenile and disrespectful, and likely to discourage actual industry people to contribute so freely.
Disrespectful and juvenile ?

YES, I AM !
BLABLABLABLAH

(you seriously wrote a proper answer just for me ??? 🤣 I'm just a Fat drunk loser having fun on the internet 😜.
Spoiler: I didn't read the rest of your message 😇.

Nonetheless, i would be pleased to read your discussion with Obbe here:

Writing novels on GAF with drunken bastards like me is pointless, don't you think ? 🥰
 
Last edited:

RaySoft

Member
It was stronger than Model 3...at home...for $199! That was enough for me.
The Model 3 boards had three revisions of the original boards during it's use. Step 1.5, 2.0 and 2.1 where Step 2.0 had some mayor spec upgrades where some where even better than the Dreamcast. So if you played any of the games listed below, some could probably feel better on the arcade than on the Dreamcast.

Games on Model3 Step 2.0 and 2.1: (2.1 was a minor upgrade from 2.0)

Step 2.0

Step 2.1

 
Last edited:

Esppiral

Member
nice model


If the model used in the final game is something to go by, there is some trickery involving how it is displayed

You get a hint here,

pbk2ltV.jpeg


also I tried to recreate her a few days ago and failed miserably lol.
UYVrnHw.jpeg


I wish someday it leaks.
 
Last edited:

Fat Frog

I advertised for Google Stadia
Alexios Alexios
Speaking of which, I think the only EMOTION the PS2 gave to Sega programmers is frustration 🤣:

zqten7D.png

Shenhua (Shenmue 1 in engine intro/9000 polygons. It's less than Shenmue passport Nozomi/14000 polygons)
1999 (emulator screenshot i guess, there are hair glitches...)

6 YEARS LATER

StqdEs4.png

Haruka (Yakuza in engine cut scene)
2005


Kutaragi told us the only limit of the PS2 was our imagination... Where are our billion polygons ?

Sega was amazing on Gamecube, Xbox with F-Zero GX, Panzer Dragoon Orta, Jet Set Radio Future...

But with Yakuza we waited 6 fucking years for that...a less blocky Kiryu Kazuma (thanks modern tools), overall more graphical consistency but fixed cameras(don't brutalize the PS2), no magic weather.

PS: Calm down, big boys. It's just a troll 🥳 i know Shenmue also has low poly character versions like FFX and so many games.
I also know that Douglas from Silent Hill 3, the ultimate reference, has probably an insane amount of polygons...

Let me check... wait... only 8000 polygons for the mighty douglas ? No link with the unlimited power of the PS2 ????
As we already said it's the power of the Playst... of modern tools 😜
 
Last edited:

frankefrum

Neo Member
I’m old enough to remember the Dreamcast era perfectly. I would have been 16 years old at the time and basically in the height of my teenage video game obsession. I still game regularly, but at that age you just have more free time to devote to a hobby.

Up to that point we had seen 2D gaming the majority of our life. It became very artistic with the likes of the SNES, Genesis, PS1, and Neo Geo. When the PS1, Saturn, and N64 started introducing us to 3D gaming as the dominant dimension at home, it was very tough looking. Trust me, that era did not age well with the exception of a few games like Mario 64. The polygons were few, ugly, low res, and clipped like crazy.

However, in 1996-1997 there was one place where you could experience ground breaking tech and impressive 3D and that was your local arcade. I was lucky enough to live an hour and a half from a very big arcade in Tallahassee Florida and they had the pinnacle of arcade gaming performance at the time…Virtua Fighter 3. I convinced my parents to make a trip to this arcade while we were visiting Tallahassee for a Florida State football game. The first time I saw this game in person I honestly couldn’t believe what I was witnessing. It was gorgeous, stunniny, remarkable graphically that is hard to describe these days. It’s extremely rare that a game’s graphics leave me jaw dropped, but that’s what it was like and you could only find it at the biggest, baddest arcades that imported it from Japan.

Then, on November 27, 1998 the Dreamcast debuted in Japan. Not long after I would have experienced Virtua Fighter 3 in an arcade. The port from arcade to home was almost flawless (the Dreamcast’s texture memory wasn’t quite as good as the arcade tbf). I was a savvy gamer back then and was willing to import a Dreamcast to get the experience at home a year before the USA release on 9/9/1999. I bragged to my friends and had the come over to witness the miracle of Dreamcast. We played Sonic Adventure which was a breathtaking experience at the time and actually a damn fun 3D Sonic game if ever one existed. We also enjoyed an arcade perfect port of Marvel vs Capcom 2. Mind you, arcade ports were never perfect back then. Hell, go look at Street Fighter 2. The developers had to make sacrifices back then and did little tricks like removing a couple 2D sprite elephants from Dhalsim’s stage to get it to work properly lmao!

All of that said, you know what really impressed? As I mentioned earlier, the arcade was the best place to experience the bleeding edge tech. This is foreign to us today because arcades are basically dead, but that’s how things went for us. Namco was one of the best in the business and Soul Calibur was one of their best looking arcade games (definitely not on par with the Virtua Fighter 3 hardware, but impressive nonetheless). Well, when Namco ported Soul Calibur to the Dreamcast, they completely upgraded the graphics to a degree that once again stood jaw absolutely dropped drooling over something that felt like it was brought to us from the future. This level of graphical eye candy simply had never happened at home on consoles.

Eventually, we got the PS2 and it was pretty damn impressive in its own right. It was always a bitch to program for (unlike the Dreamcast which was extra easy according to developers at the time). So it seemed like the PS2 highs were extra high but there was a lot of trash on that console too.

At any rate, to answer your question, the Dreamcast was way ahead of its time imo. From November 27, 1998 to March 4, 2000 (the PS2 debut) it was unequivocally the best experience one could have at home on a console. You know what else? The games were extremely fun and very unique. It had a certain Japanese flair and charm that has never been replicated. It is easily one of my top 3 favorite consoles of all-time even though it wasn’t exactly a financial success for Sega.
 

Esppiral

Member
Accordingly to a dude in thread, being an enhanced PS1 game which is further enhanced on the Dreamcast, should have ran at 30FPS.
And be the maxium the DC had to offer.

I guess the recent "next gen" patch for fallout 4 that runs like dogshit on ps5 and Series X is representative of what those consoles can do, I mean following that logic....
 
Last edited:

PaintTinJr

Member
Alexios Alexios
Speaking of which, I think the only EMOTION the PS2 gave to Sega programmers is frustration 🤣:

zqten7D.png

Shenhua (Shenmue 1 in engine intro/9000 polygons. It's less than Shenmue passport Nozomi/14000 polygons)
1999 (emulator screenshot i guess, there are hair glitches...)

6 YEARS LATER
Are you sure the original Dreamcast version does in-engine rendering for that FMV/Cinematic? and that your not getting mixed up with a change to in game cinematic for the remaster?

Most consoles and games of that era used offline full motion videos captures that were rendered offline and filled up the CD/DVD storage for a game.
 
No intellectual dishonesty ok. I made it very clear (before mentioning the list of Dreamcast games with no shadows ) that the 30fps of action games like shadow man is due to a high degree of enhancement, hell even the ps1 runs fighting games like Tobal 2 at 60fps . I never said about fighting. games
 
Are you sure the original Dreamcast version does in-engine rendering for that FMV/Cinematic? and that your not getting mixed up with a change to in game cinematic for the remaster?

Most consoles and games of that era used offline full motion videos captures that were rendered offline and filled up the CD/DVD storage for a game.
He called himself a troll, so I recommend that you do not interact with him, there is no point in talking to someone who is not receptive.
 
And be the maxium the DC had to offer.

I guess the recent "next gen" patch for fallout 4 that runs like dogshit on ps5 and Series X is representative of what those consoles can do, I mean following that logic....

Straight facts. Also i do get some of points of G Geometric-Crusher and PaintTinJr PaintTinJr but dudes, you can write off all that you want, but Esppiral Esppiral being just one person with no budget it´s taking one of the most demanding and advanced games on DC and further than anythingwe have seen befotre and still 60fps...So just go and wonder what AM2 or another studios would had done if the console stood commercially alive and selling well at least another FY. DC just had tecnhically one generation of AAA games, most of them started since 99 or with technology created on that year or before (like the case of 2K sports games or Shenmue 2). Obviously an project AAA which would had started from zero in late 2000 or even 2001 would had take the console beyond to what we have just seen...I insist, just look the Atomiswave games like Demolish Fists, Faster Than Speed and Force Five...Yes, all DC elite games still stomps them, but they make interesting things which help them to look better than the equivalent budget games of actual DC lifespan, incluiding all 32/64 bits ports. You guys insist on Shadowman, but deliberately ignore Soul Calibur which is a real example of a 32 bit game turned for real into a proper 6th gen and Dreamcast game.
 

PaintTinJr

Member
Straight facts. Also i do get some of points of G Geometric-Crusher and PaintTinJr PaintTinJr but dudes, you can write off all that you want, but Esppiral Esppiral being just one person with no budget it´s taking one of the most demanding and advanced games on DC and further than anythingwe have seen befotre and still 60fps...So just go and wonder what AM2 or another studios would had done if the console stood commercially alive and selling well at least another FY. DC just had tecnhically one generation of AAA games, most of them started since 99 or with technology created on that year or before (like the case of 2K sports games or Shenmue 2). Obviously an project AAA which would had started from zero in late 2000 or even 2001 would had take the console beyond to what we have just seen...I insist, just look the Atomiswave games like Demolish Fists, Faster Than Speed and Force Five...Yes, all DC elite games still stomps them, but they make interesting things which help them to look better than the equivalent budget games of actual DC lifespan, incluiding all 32/64 bits ports. You guys insist on Shadowman, but deliberately ignore Soul Calibur which is a real example of a 32 bit game turned for real into a proper 6th gen and Dreamcast game.
For my part, you keep missing the point I've made repeatedly, and that is the access to the GPU is highly abstracted and was at a level of Opengl that wasn't great for anything better than what has already been made, and an extra year wasn't going to make a big technical difference.

As for Esppiral's mod work, he's just using existing engine capabilities from what he said, and even that is showing some teething difficulties for being compared with retail software because in his last video the opening flyby shows a engine frustum limitation where the camera incorrectly culls geometry from the large Egyptian Jackal(?) statues.
 

RobRSG

Member
As for Esppiral's mod work, he's just using existing engine capabilities from what he said, and even that is showing some teething difficulties for being compared with retail software because in his last video the opening flyby shows a engine frustum limitation where the camera incorrectly culls geometry from the large Egyptian Jackal(?) statues.
Yeah, we should believe the Shadowman duo instead. 😂

What a bunch of jokers.

Keep the good work, Esppiral Esppiral !
 

PaintTinJr

Member
Yeah, we should believe the Shadowman duo instead. 😂

What a bunch of jokers.

Keep the good work, Esppiral Esppiral !
No one is saying to give up. but are you saying you can't see the missing model geometry at 11secs below the "E" of "select" on the right handside?

PvhISYg.png


It might be a polygon winding issue, or something else instead of frustum culling the front near plane, but it clearly isn't how he intend to render.
 

RobRSG

Member
No one is saying to give up. but are you saying you can't see the missing model geometry at 11secs below the "E" of "select" on the right handside?

PvhISYg.png


It might be a polygon winding issue, or something else instead of frustum culling the front near plane, but it clearly isn't how he intend to render.
I can see that and I take no issue, it is WIP by a single man.
 

Esppiral

Member
No one is saying to give up. but are you saying you can't see the missing model geometry at 11secs below the "E" of "select" on the right handside?

PvhISYg.png


It might be a polygon winding issue, or something else instead of frustum culling the front near plane, but it clearly isn't how he intend to render.


What the actual fuch are you saying, this is getting tiresom really, I am pretty sure you know what Backface culling is, that model is only meant to be shown from the front, and it is set to cull the backwards geometry so that is what happens when the model is facing the wrong direction, for your information that model is directly taken from Soul Calibur 3 from the ps2 and that is how the model is set up... limitations he says... jesus christ....

The model has been repositioned in the latest version so it is not noticeable, (LIKE ON PS2) but this is NOT a limitation...

This is the original model.
TMc5rm3.png


The back face of the model is NOT modelled, ( maybe a limitation of the ps2)

FnMBjTs.png
.

When you set the model to backface cull (just like the ps2, why would you render polygons that are not visible?) this is how it looks, exactly like on Dreamcast or ps2, or maybe is a limitation of blender, I am not sure....

H6iN1uE.png
 
Last edited:

PaintTinJr

Member
What the actual fuch are you saying, this is getting tiresom really, I am pretty sure you know what Backface culling is, that model is only meant to be shown from the front, and it is set tu cull the backwards geometry so that is what happens when the model is facing the wrong direction, for your information that model is directly taken from Soul Calibur 3 from the ps2 and that is how the model is set up... limitations he says... jesus christ....

The model has been repositioned in the latest version so it is not noticeable, (LIKE ON PS2) but this is NOT a limitation...
Hey, no need to get so defensive, I was only pointing it out that it is a visual bug and my initial thought when I first saw it, was that it was frustum near plane clipping.

You said it is back face culling, but if the model has been built correctly and transformed correctly in the scene, then those polygons should be front facing when they are incorrectly being culled, and the front of the statue that is facing away at that point will have its polygon's backfacing at the time, and should be getting culled.

Even if you don't like my input, you might find recalculating outside faces and normals in blender on the model will fix that issue, allowing you to place it anywhere and view it from anywhere. But again, it might be something else, like too many overlaping layers exceeding the DC GPU layer sorting level count, or just a missing texture, etc.
 
Last edited:

Alexios

Cores, shaders and BIOS oh my!
Are you sure the original Dreamcast version does in-engine rendering for that FMV/Cinematic? and that your not getting mixed up with a change to in game cinematic for the remaster?

Most consoles and games of that era used offline full motion videos captures that were rendered offline and filled up the CD/DVD storage for a game.
Of course it's in-engine real time 3D like most if not all of Shenmue 1 & 2 cut scenes, hence the higher res rendering/potential glitches/other improvements in emulators, hence knowing the polycount of this or that model as people have extracted it in various ways and got the data, etc., do you and Crusher know ANYTHING about the system you're dissing and its existing games before moving on to its theoretical max via ports or later generation titles that didn't happen or whatever you wanna pretend you know?

Dude, he's just explaining to you that's how the model was ripped straight out of PS2, either because they didn't model the back of it at all knowing that in the game it comes from they'll never have a camera angle from the back like you get in the temporary set up esppiral got importing it as wip in a wholly different game on Dreamcast, or because the model ripping was incomplete/faulty in some way which gives the same result for him to work with in the end. Culling of non visible polygons was hardly optimized back then so it makes sense Namco simply didn't put those polygons in the game at all knowing full well they won't ever show it from the back, to maximize the qualtiy and detail of what's actually visible. It's got nothing to do with Dreamcast in any case. MANY games did this back then, famously racing titles (where you often couldn't ever reverse no matter how you tried and even if you did it was fine to have glitches) and lightgun games that are on rails only rendering what's visible by their camera angle to maximize the detail of visible things. It's not that har to get before pinning it all on the system he imported the model to, lol. Christ, you two are confirmation bias the poster.
 
Last edited:

Esppiral

Member
Are you sure the original Dreamcast version does in-engine rendering for that FMV/Cinematic? and that your not getting mixed up with a change to in game cinematic for the remaster?

Most consoles and games of that era used offline full motion videos captures that were rendered offline and filled up the CD/DVD storage for a game.

100% realtime, Shenmue does not use prerrendered videos, this are some models from those scenes.

wn0KBPW.jpeg


HseDgNk.jpeg


rFJaTAf.jpeg


Ryo's head is from the passport not the body, the rest are 100 the ingame models.

Hey, no need to get so defensive, I was only pointing it out that it is a visual bug and my initial thought when I first saw it, was that it was frustum near plane clipping.

You said it is back face culling, but if the model has been built correctly and transformed correctly in the scene, then those polygons should be front facing when they are incorrectly being culled, and the front of the statue that is facing away at that point will have its polygon's backfacing at the time, and should be getting culled.

Even if you don't like my input, you might find recalculating outside faces and normals in blender on the model will fix that issue, allowing you to place it anywhere and view it from anywhere. But again, it might be something else, like too many overlaping layers exceeding the DC GPU layer sorting level count, or just a missing texture, etc.

The model are perfectly fine, I just posted and image showing you the same behavior on blender when you set the backface culling, that part shouln't be facing the camera, nothing to do with the Dreamcast GPU....
 
Last edited:
Top Bottom