Support NeoGAF

[polybius80]

lot of games from that generation uses IK, primal ghost hunter and rogue squadron and I think halo too, but just because it use IK doesn't mean it uses everywere, most games only use it to adjust to the floor because its needed the most there

I think for the fireflies what they are using is a mask to tell GPU were "not to draw" shadows or not to shade the floor, the floor its obscured but the fireflies light simply make it not to obscure that part, because its very sharp its probably very simple, 1 bit per pixel probably

there are many trick used that generation to make "impossible" effects



deferred light was used in that generation, very rare, but was used

here is a demo for xbox



and of course, shrek the first game I think to use it




for more information about deferred rendering on xbox and ps2

Deferred Lighting and Shading - ppt video online download

Presenters Rich Geldreich, Ensemble Studios, Matt Pritchard, Ensemble Studios, John Brooks, Blue Shift, Inc.,

slideplayer.com

 

[Esppiral]

Lol at that guy thinking that Uncharted was the first game to use inverse kinematics, it shows you how short sighted many people is.

Shenmue, a 1998 game, uses inverse kinematics and OOt does it too, games that shipped waaay before Uncharted were even draft.

Also plenty of games from the same period of time as Uncharted uses IK

 

[ehnox]

Lol at that guy thinking that Uncharted was the first game to use inverse kinematics, it shows you how short sighted many people is.

Shenmue, a 1998 game, uses inverse kinematics and OOt does it too, games that shipped waaay before Uncharted were even draft.

Also plenty of games from the same period of time as Uncharted uses IK

Forgot to post the source

Zelda: Wind waker Tech and Texture Analysis *picture heavy*

So After this little thread: http://www.polycount.com/forum/showthread.php?t=102634 i felt like the opinion was more yay than nay.

polycount.com

There is even a picture of the guy XD

 

[ehnox]

Another game I find particularly impressive in the graphics department, this one a PS2 exclusive




This game has a Bugs Life vibe to it some how

 

[-Minsc-]

I seem to have traveled back in time to somewhere between the years 2001-2006. Those were exciting times.

 

[Romulus]

The Getaway on ps2 was another graphical gem.

 

[V4skunk]

The Getaway on ps2 was another graphical gem.

Mgs3 and Transformers.

 

[Enzo88]

What an amazing generation.
The DC did so much in such a short time, so many innovations, sadly it died early and most of its games were ported.
PS2 had the largest and most complete library, but my heart, somehow, in that gen belongs to the Gamecube, that console gave me the best time with its 15/20 must play games.

 

[SonGoku]

Another game I find particularly impressive in the graphics department, this one a PS2 exclusive




This game has a Bugs Life vibe to it some how

Toy Story graphics redeemed!

 

[pawel86ck]

I have finally played ghosthunter, character model faces are very detailed compared to other PS2 games, but I still think Splinter Cell 1 on PS2 has much better lighting.

Splinter Cell 1 PS2

And of course Splinter Cell 1 on xbox has even better lighting

Guys in your opinion what's more important when it comes to graphics, polygons or lighting?

 

[polybius80]

I have finally played ghosthunter, character model faces are very detailed compared to other PS2 games, but I still think Splinter Cell 1 on PS2 has much better lighting.

Splinter Cell 1 PS2

And of course Splinter Cell 1 on xbox has even better lighting

Guys in your opinion what's more important when it comes to graphics, polygons or lighting?

I remember the flashlight in ghost hunter is very well done and produce shadows

stolen is very competitive




but hitman blood money has better light overall, the way it interacts with objects is better than splinter cell, SC may be more intense in shadows but hitman has more open and interesting scenes that look better with the light system and variations in color, but its good to point that SC has to use more bland scenes for most of its levels because it makes sense for the story, most of the levels are boring office building and military bases, they have interesting scenes here and there but in hitman BM every level is interesting and beautiful in its own way

I think light is more important , SC use very low polygons models and scenes are generally simple but the shadows and effects(specially chaos theory on xbox) make the game very impressive, I remember chaos theory was used a lot on PC as benchmark

 

[K.N.W.]

Mgs3 and Transformers.

I would like to post MGS3 and Transformers' screnshots, but they both are rendering in interlaced mode, and so they would look really bad on screenshots. But I'll post some The Getaway screens later!!

 

[EightBit Man]

One of the games that inspired (the excellent) The Getaway:

Probably my favorite racing game ever, and a large part is due to it's capturing the cities at night so well. Unbelievable at the time.
Unfortunately it didn't do well:

(from Retro Gamer, Issue 124)

 

[Romulus]

but hitman blood money has better light overall, the way it interacts with objects is better than splinter cell, SC may be more intense in shadows but hitman has more open and interesting scenes that look better with the light system and variations in color, but its good to point that SC has to use more bland scenes for most of its levels because it makes sense for the story, most of the levels are boring office building and military bases, they have interesting scenes here and there but in hitman BM every level is interesting and beautiful in its own way

I think light is more important , SC use very low polygons models and scenes are generally simple but the shadows and effects(specially chaos theory on xbox) make the game very impressive, I remember chaos theory was used a lot on PC as benchmark

I think chaos theory has equal lighting, but where it really pulls away is textures imo.

 

[Esppiral]

Lol no, The lighting in stolen is way inferior.

 

[V4skunk]

I would like to post MGS3 and Transformers' screnshots, but they both are rendering in interlaced mode, and so they would look really bad on screenshots. But I'll post some The Getaway screens later!!

Just post 60fps YouTube videos.

I have finally played ghosthunter, character model faces are very detailed compared to other PS2 games, but I still think Splinter Cell 1 on PS2 has much better lighting.

Splinter Cell 1 PS2

And of course Splinter Cell 1 on xbox has even better lighting

Guys in your opinion what's more important when it comes to graphics, polygons or lighting?

Still doesn't change the fact that GC pushed more polygons in real time and had exclusives that give or take matched the best on xbox.

 

[Romulus]

Just post 60fps YouTube videos.
Still doesn't change the fact that GC pushed more polygons in real time and had exclusives that give or take matched the best on xbox.

Theres no "fact" about polygon count comparsions. Facts can be proven. An exclusive dev claiming something means absolutely nothing, especially considering how flaky and broad his comments were.

 

[polybius80]

Just post 60fps YouTube videos.
Still doesn't change the fact that GC pushed more polygons in real time and had exclusives that give or take matched the best on xbox.

actually grand prix push as much triangles as factor 5 claim(18 millions) and do that with far more resolution(640x960) and solid framerate

 

[Fafalada]

I would like to post MGS3 and Transformers' screnshots, but they both are rendering in interlaced mode, and so they would look really bad on screenshots. But I'll post some The Getaway screens later!!

Actually both MGS3 and Transformers were rendered in full-res, they just output interlaced.
The distinction is important if you're arguing about performance implications - there's virtually NO PS2 games that run sub 60fps and Don't render a full height frame-buffer (I think there might have been 1 or 2 in first year).

The output being interlaced or not has no impact on performance - but it does compromise image capture quality.

 

[K.N.W.]

Actually both MGS3 and Transformers were rendered in full-res, they just output interlaced.
The distinction is important if you're arguing about performance implications - there's virtually NO PS2 games that run sub 60fps and Don't render a full height frame-buffer (I think there might have been 1 or 2 in first year).

The output being interlaced or not has no impact on performance - but it does compromise image capture quality.

That's great because I can hack MGS 3 to output in progressive and take screenshots. I didn't want to because I'm trying to be fair and post screens that show what the console truly draws (my MGS2 480P screens were a sweet exception). Sadly, Transformers doesn't get in-game with the 480P hack

 

[Fafalada]

That's great because I can hack MGS 3 to output in progressive and take screenshots. I didn't want to because I'm trying to be fair and post screens that show what the console truly draws (my MGS2 480P screens were a sweet exception). Sadly, Transformers doesn't get in-game with the 480P hack

Yea the hacks can't account for specific memory layouts that can end up corrupting the game or crashing etc. PCSX2 should really add an option to bypass that though as in emulator it should be trivial to do it 'safely'.

 

[Rolling_Start]

Ooh, first post and it's a retro post!

Still doesn't change the fact that GC pushed more polygons in real time and had exclusives that give or take matched the best on xbox.

There are a few misconceptions that are still pushed around the web to this day - GC's relative polygon performance is one.

Xbox could actually push far more than the GC, and at all stages of the polygon pipeline. From setup, to transform, to culling , to rasterising. And not just on paper, either.

GC was a fine system, but the silicon designs were older, smaller, and used less power. Taken for what it was the GC hardware was certainly nice and smartly selected by Nintendo. It was outright less capable than the Xbox though.

 

[Romulus]

Ooh, first post and it's a retro post!



There are a few misconceptions that are still pushed around the web to this day - GC's relative polygon performance is one.

Xbox could actually push far more than the GC, and at all stages of the polygon pipeline. From setup, to transform, to culling , to rasterising. And not just on paper, either.

GC was a fine system, but the silicon designs were older, smaller, and used less power. Taken for what it was the GC hardware was certainly nice and smartly selected by Nintendo. It was outright less capable than the Xbox though.

I agree that the xbox was more capable overall too, but what evidence is there about polygon performance on either side? All it is just talk from what I've seen. It's not like modern specs where we know if something is actual 4k for example

 

[Rolling_Start]

I agree that the xbox was more capable overall too, but what evidence is there about polygon performance on either side? All it is just talk from what I've seen. It's not like modern specs where we know if something is actual 4k for example

Evidence comes from both the specs and developers who've talked openly about their experiences with the system.

Stuff like triangle setup, peak transform and fill rate are much higher on the Xbox. In terms of what's deliverable, ERP on Beyond3D (a well respected, veteran developer) talked about the GCs poor triangle culling performance and the need to write a CPU culling routine. This was also the same guy who says his cancelled racing game for Xbox was able to push something like 25 million (lit, textured) pps in game. I believe him. It was certainly well beyond the "best case Factor 5 PR quotes" that keep re-surfacing.

Xbox also had the advantage of having true vertex shaders, meaning things like deformation and skinning could be run as shaders with minimal CPU input. That's something that the GCs relatively weaker CPU would have to spend cycles on (even assuming you weren't culling on the CPU). Games like the Factor 5 Star Wars Games are almost a "best case" for the GC - the models don't deform or need skinning, there is very little animation of the models beyond simple world space transformation, and the environments are likely to make coarse CPU based culling reasonably easy (if, indeed, the CPU was used for that).

Another developer would be Fafalada himself (this post was one of the things that after nearly 15years of on / off lurking helped finally encouraged me to sign up!):

Ok let's get this out of the way - none of these systems were comparable on any single-metric like 'power' - but it can be broken down in some 'reasonable' categories:

XB:GC
Pixel operations throughput
2.9 : 1
Textured Fillrate / Untextured fillrate (mostly things like shadow rendering)
1.4 : 1 / 5.7 : 1

Geometry/transform throughput
3-4 : 1

CPU
2-3 : 1

Memory / Usable graphics memory
1.3 : 1 / 2-3 : 1
Max Framebuffer memory*
32 : 1
* This really just to illustrate 'why' 720p patching is so easy on XB, it's a lot less to do with performance than available memory.

Optical disc speed
3.3 : 1

Numbers with a 'range'(like CPU) are based off of observed 'real-world' stats more than on-paper specs. The precise ones are paper-spec based, but generally reflect in real-use too.
Bandwidth figures are mostly irrelevant to compare directly because the above figures directly correspond to available bandwidth (all peak figures are achievable with existing memory setups in these consoles).
Real-world it was uncommon to see cases where XB didn't outperform the other two by at least 2:1 'overall'. Specialized workflows existed where you could see that advantage disappear, but it was never lopsided.

PS2 compares better on paper, but matching certain things was a lot more cumbersome on it, and it had the weakest CPU, which was the biggest single limiter when comparing 3rd party games.
On the flipside there were a few things only PS2 could really do that gen (like geometry shader acceleration), so it was entirely possible to build software difficult to port the other way too.

Everything he's saying here is in line with everything else I've picked up over the years.

As I said, there's not really any area where the GC is at anything but a disadvantage against the OG Xbox in terms of polygon throughput and polygon processing flexibility.

[BTW I'm definitely not a developer, but many, many years ago I did a bit of OGL and D3D programming while in eduction. I have friends who've worked on various platform. I sometimes dick around with Unity et al but always give up before getting far with a game idea. I'm one of those losers. ]

 

[Romulus]

Evidence comes from both the specs and developers who've talked openly about their experiences with the system.

Stuff like triangle setup, peak transform and fill rate are much higher on the Xbox. In terms of what's deliverable, ERP on Beyond3D (a well respected, veteran developer) talked about the GCs poor triangle culling performance and the need to write a CPU culling routine. This was also the same guy who says his cancelled racing game for Xbox was able to push something like 25 million (lit, textured) pps in game. I believe him. It was certainly well beyond the "best case Factor 5 PR quotes" that keep re-surfacing.

Xbox also had the advantage of having true vertex shaders, meaning things like deformation and skinning could be run as shaders with minimal CPU input. That's something that the GCs relatively weaker CPU would have to spend cycles on (even assuming you weren't culling on the CPU). Games like the Factor 5 Star Wars Games are almost a "best case" for the GC - the models don't deform or need skinning, there is very little animation of the models beyond simple world space transformation, and the environments are likely to make coarse CPU based culling reasonably easy (if, indeed, the CPU was used for that).

Another developer would be Fafalada himself (this post was one of the things that after nearly 15years of on / off lurking helped finally encouraged me to sign up!):



Everything he's saying here is in line with everything else I've picked up over the years.

As I said, there's not really any area where the GC is at anything but a disadvantage against the OG Xbox in terms of polygon throughput and polygon processing flexibility.

[BTW I'm definitely not a developer, but many, many years ago I did a bit of OGL and D3D programming while in eduction. I have friends who've worked on various platform. I sometimes dick around with Unity et al but always give up before getting far with a game idea. I'm one of those losers. ]

It's good info, and most likely true, but again its impossible to know for sure when you have developers claiming numbers. There are so many viables in game engines and what they're trying to achieve.
What does boed well for xbox, is there are a fair amount of multiplatform devs that have a similar sediment that the xbox was far and away more powerful. Are they all lying about the same thing? Probably not.

 

[Rolling_Start]

It's good info, and most likely true, but again its impossible to know for sure when you have developers claiming numbers. There are so many viables in game engines and what they're trying to achieve.
What does boed well for xbox, is there are a fair amount of multiplatform devs that have a similar sediment that the xbox was far and away more powerful. Are they all lying about the same thing? Probably not.

Yep, sometimes you have to go with the weight of odds. Especially when all multi-plat devs say one thing, and none say anything else. Ever.

Also, a little understanding of the stages of the 3D pipeline is never unhelpful. Xbox is significantly ahead at all times.

But GC is still really cool. Fanboyism leads to a battle for a absolute supremacy, when a more nuanced appreciation means you can still really like something that isn't the supreme performer.

I've probably been a fanboy in the past, tbh. It doesn't make you appreciate the thing you really like more, it just deprives you of liking other things as well.

 

[pawel86ck]

That's great because I can hack MGS 3 to output in progressive and take screenshots. I didn't want to because I'm trying to be fair and post screens that show what the console truly draws (my MGS2 480P screens were a sweet exception). Sadly, Transformers doesn't get in-game with the 480P hack

You can hack MGS3 to output 480p, at least it will be easier to take high quality screen captures. If your screen capture card has poor deinterlace processing then 480i picture will look way worse than it should.

I have found out 480i vs 480p difference is very small on static picture if you will send both 480i and 480p over component and if TV has good deinterlace processing. On my old samsung tv there is really no difference between 480i vs 480p on static picture, only during motion I could tell 480i looks worse. However on my new panasonic TV 480i looks extremely blurry compared to 480p even when I send 480i over component.

Anyway back to screenshots

Ghost Hunter looks better with higher resolution and in game gamma adjustment (with default gamma settings black was grey). Textures arnt the best (If only PS2 would support S3TC texture compression ) but game looks good overall compared to other PS2 games, and it's actually very fun to play.

Maybe GC wasnt much faster compared to PS2, but it had S3TC and it made a difference for sure when it comes to texture quality. The same game Time Splitters 2 on PS2 and GC.

GC

PS2

GC

PS2

GC

PS2

 

[polybius80]

You can hack MGS3 to output 480p, at least it will be easier to take high quality screen captures. If your screen capture card has poor deinterlace processing then 480i picture will look way worse than it should.

I have found out 480i vs 480p difference is very small on static picture if you will send both 480i and 480p over component and if TV has good deinterlace processing. On my old samsung tv there is really no difference between 480i vs 480p on static picture, only during motion I could tell 480i looks worse. However on my new panasonic TV 480i looks extremely blurry compared to 480p even when I send 480i over component.

Anyway back to screenshots

Ghost Hunter looks better with higher resolution and in game gamma adjustment (with default gamma settings black was grey). Textures arnt the best (If only PS2 would support S3TC texture compression ) but game looks good overall compared to other PS2 games, and it's actually very fun to play.

Maybe GC wasnt much faster compared to PS2, but it had S3TC and it made a difference for sure when it comes to texture quality. The same game Time Splitters 2 on PS2 and GC.

GC

PS2

GC

PS2

GC

PS2

ps2 can use more texture compression algorithms too but they are programed in the system not on hardware so it depends the game if it uses them or not, some games have better textures in GC but you will find examples were the opposite is true like the need for speed games(videos posted earlier) and harry potter games( like the videos I posted earlier) and sphynx(emulator screens)

S3 have advantage of 2:1 compression over regular CLUT on Ps2 so it can use more color(resolution usually is the same), but also GC's cache is 1 MB were in PS2 it can use 2 MB of edram as texture cache, also PS2 has more ram and faster access, gc has a bit more texture cache access, its difficult to put one over the other, but an inexperienced developer will normally make better textures on GC out of the box and with more experience improve PS2 textures, in those screen form time splitters 2, GC seems to use the same back exact texture but its combining the texture with a "detail texture" similar to other games like area 51 to improve detail probably only used when close to the walls

check wall textures when camera is close and there are detail textures on top of the normal texture


also area 51 is a way more complex game than TS2 or 3

 

[K.N.W.]

You can hack MGS3 to output 480p, at least it will be easier to take high quality screen captures. If your screen capture card has poor deinterlace processing then 480i picture will look way worse than it should.

The TV has nothing to do with capture quality. I'm taking screens directly form PS2's framebuffer (no capture device), if the picture is interlaced there's nothing we can do, since the framebuffer contains only half of the picture, and the missing lines are gone forever.

I have found out 480i vs 480p difference is very small on static picture if you will send both 480i and 480p over component and if TV has good deinterlace processing. On my old samsung tv there is really no difference between 480i vs 480p on static picture, only during motion I could tell 480i looks worse. However on my new panasonic TV 480i looks extremely blurry compared to 480p even when I send 480i over component.

That's because, if you think about it, 480P and 480I have the same exact amount of pixels, they are just delivered in two separate frames in interlaced mode(at 30 FPS at least, in 60 FPS mode you get 60 indipendent frames) . But, then, both at 30 and 60 FPS, while the game isn't moving, the horizontal pixel lines match (your TV merges them to create a 480P image) and look almost as sharp as progressive mode. (This is also the reason why setting the game to upscale to 480P in GSM (not changing the internal rendering resolution via hack) results in a nice picture, on par with 480P, in many games, since many of them internally render a full frame before it gets scaled to an interlaced field).

Nice screens BTW, but they look emulated

ps2 can use more texture compression algorithms too but they are programed in the system not on hardware so it depends the game if it uses them or not, some games have better textures in GC but you will find examples were the opposite is true like the need for speed games(videos posted earlier) and harry potter games( like the videos I posted earlier) and sphynx(emulator screens)

I totally regret not picking up this game when I found it. One of the deciding factors was the extreme detail density in the "bullshots" available online.... "There's no wayyyy it's gonna look like this!!11!1!1"

 

[pawel86ck]

The TV has nothing to do with capture quality. I'm taking screens directly form PS2's framebuffer (no capture device), if the picture is interlaced there's nothing we can do, since the framebuffer contains only half of the picture, and the missing lines are gone forever.



That's because, if you think about it, 480P and 480I have the same exact amount of pixels, they are just delivered in two separate frames in interlaced mode(at 30 FPS at least, in 60 FPS mode you get 60 indipendent frames) . But, then, both at 30 and 60 FPS, while the game isn't moving, the horizontal pixel lines match (your TV merges them to create a 480P image) and look almost as sharp as progressive mode. (This is also the reason why setting the game to upscale to 480P in GSM (not changing the internal rendering resolution via hack) results in a nice picture, on par with 480P, in many games, since many of them internally render a full frame before it gets scaled to an interlaced field).

Nice screens BTW, but they look emulated

Because I'm using emulator like I have mentioned in my 1'st post. My previous screenshots from Ghost Hunter were in 640x448, but it's much harder to see texture quality and other details on such small pictures.

If you taking screenshots directly from PS2, so 480i should be not processed in any way and therefore look much more blurry compared to playing these PS2 games on TV. Like I have said, the same 480i picture will look sharper on tv (of course assuming your tv has good deinterlacing).

On emulator many games in PCSX2 renders perfectly sharp picture even without progressive mode, while other looks like a blurry mess, so I'm guessing you are right and many of these games render 480p internally.

I have to say the latest PCSX2 build (developer build) is MUCH improved compared to official release. Finally all effects are rendered in God Of War and other games using hardware mode. Right now I still prefer to play PS2 games on my real PS2 and small CRT because controls feels just much better (thumbstics emulation arnt perfect in PCSX2 and it's MUCH harder to aim in the same game compared to real PS2) but in the future maybe they will improve controls emulation and then I will have no reason to play them on PS2.

 

[pawel86ck]

ps2 can use more texture compression algorithms too but they are programed in the system not on hardware so it depends the game if it uses them or not, some games have better textures in GC but you will find examples were the opposite is true like the need for speed games(videos posted earlier) and harry potter games( like the videos I posted earlier) and sphynx(emulator screens)

S3 have advantage of 2:1 compression over regular CLUT on Ps2 so it can use more color(resolution usually is the same), but also GC's cache is 1 MB were in PS2 it can use 2 MB of edram as texture cache, also PS2 has more ram and faster access, gc has a bit more texture cache access, its difficult to put one over the other, but an inexperienced developer will normally make better textures on GC out of the box and with more experience improve PS2 textures, in those screen form time splitters 2, GC seems to use the same back exact texture but its combining the texture with a "detail texture" similar to other games like area 51 to improve detail probably only used when close to the walls

check wall textures when camera is close and there are detail textures on top of the normal texture


also area 51 is a way more complex game than TS2 or 3

I dont know if ps2 can indeed use some clever software texture compression algorithms like you say, but pretty much every game that I have played on PS2 had blurry textures. Even textures in dreamcast games were much better (quake 3) compared with PS2 ports, so I'm guessing it's has something to do with hardware texture compression, because Dreamcast has less RAM compared to PS2.

It's funny you mentioned area 51, because I have played it on my xbox for the first time just today. For some reason I always thought it's a very bad game, but I like it. Area 51 at least on xbox looks surprisingly good, definitely better compared to Time Splitters 2-3, but of course time spliters are 60fps games.

Here's some PS2 screenshots,

Even PS2 version looks good, but these detail textures looks just wrong in hardware mode, so only native resolution and software renderer works correctly.

 

[Esppiral]

Dreamcast has 8mb of ram for video+hardware compression so you can effectively store the equivalent of 24mb of textures IIRC, ps2 has half that, 4mb with no hardware compression, that's why Dreamcast games has better textures than ps2 or GameCube games.

 

[Panajev2001a]

Dreamcast has 8mb of ram for video+hardware compression so you can effectively store the equivalent of 24mb of textures IIRC, ps2 has half that, 4mb with no hardware compression, that's why Dreamcast games has better textures than ps2 or GameCube games.

GameCube supported S3TC for its texture “cache” and PS2 also used its 4 MB of monster scale speedy eDRAM to stream textures in and out of main RAM while keeping a local buffer (for off screen render targets too), you had a cruder form of compression (palettised textures), could use the IPU (MPEG2 decoder) to decompress in realtime, you could be selective and only upload the exact mip map levels you needed, or do on GS decompression if you had enough frame time free and were super short on RAM:

A Comparison of Several Texture Compression Schemes /PS2,XBOX .../

http://www.hemiware.com/technology.htm Features: Texture Compression - VQ for PS2, S3TC for GAMECUBE, Xbox and PC PS2 Features: · VQ compression (2 bits per pixel, full 32 bit output) · Auto-generate mipmaps - uses GS bandwidth to save main bus bandwidth · Path 3 parallel...

forum.beyond3d.com

You had more main RAM than Dreamcast, a much bigger external storage solution, and a lot of bandwidth between main RAM and graphics processor. It was not easy, but there were tons of ways to tackle issues you may have been facing.

 

[Rolling_Start]

Dreamcast has 8mb of ram for video+hardware compression so you can effectively store the equivalent of 24mb of textures IIRC, ps2 has half that, 4mb with no hardware compression, that's why Dreamcast games has better textures than ps2 or GameCube games.

IIRC, developer comments suggest something like 2.5 ~ 3 times the memory needed to provide similar texture quality to the DC on the PS2.

This makes sense when you consider that the common formats on PS2 were 4-bit (16 colour) and 8-bit (256 colour) CLUTs, and on DC 2 bit VQ (4-bit was support too) textures with the odd CLUT when transparency was required.

Typical DC games had around 5.5 MB for textures (texturing had to be done from video memory) so something like 14~ 16 MB on PS2. The thing is, you still wouldn't get exactly the same results. 4-bit and even 8-bit colour means using a minimal number of hues *per texture* to avoid garish contrast, which explains the PS2 "look" (you'll all know it even if you don't know you know it!) of lots of variety within the scene, but individual textures frequently looking rather monotone. You can test this out yourself by taking a PS2 screen grab and checking the colours in an image manipulation program (even good 'ol MS paint will do).

VQ on the other had the drawbacks of compression artefacts and not supporting transparency, iirc. Was undoubtedly a fantastic asset for the DC though.

Anyway, PS2 was a polygon monster and spending 16 MB of ram on textures would have starved you of memory for highly detailed meshes. So lower quality textures made sense.

 

[cireza]

GameCube supported S3TC for its texture “cache” and PS2 also used its 4 MB of monster scale speedy eDRAM to stream textures in and out of main RAM while keeping a local buffer (for off screen render targets too), you had a cruder form of compression (palettised textures), could use the IPU (MPEG2 decoder) to decompress in realtime, you could be selective and only upload the exact mip map levels you needed, or do on GS decompression if you had enough frame time free and were super short on RAM:

A Comparison of Several Texture Compression Schemes /PS2,XBOX .../

http://www.hemiware.com/technology.htm Features: Texture Compression - VQ for PS2, S3TC for GAMECUBE, Xbox and PC PS2 Features: · VQ compression (2 bits per pixel, full 32 bit output) · Auto-generate mipmaps - uses GS bandwidth to save main bus bandwidth · Path 3 parallel...

forum.beyond3d.com

You had more main RAM than Dreamcast, a much bigger external storage solution, and a lot of bandwidth between main RAM and graphics processor. It was not easy, but there were tons of ways to tackle issues you may have been facing.

Still in the end, texture work always appear largely better and cleaner on Dreamcast.

 

[pawel86ck]

GameCube supported S3TC for its texture “cache” and PS2 also used its 4 MB of monster scale speedy eDRAM to stream textures in and out of main RAM while keeping a local buffer (for off screen render targets too), you had a cruder form of compression (palettised textures), could use the IPU (MPEG2 decoder) to decompress in realtime, you could be selective and only upload the exact mip map levels you needed, or do on GS decompression if you had enough frame time free and were super short on RAM:

A Comparison of Several Texture Compression Schemes /PS2,XBOX .../

http://www.hemiware.com/technology.htm Features: Texture Compression - VQ for PS2, S3TC for GAMECUBE, Xbox and PC PS2 Features: · VQ compression (2 bits per pixel, full 32 bit output) · Auto-generate mipmaps - uses GS bandwidth to save main bus bandwidth · Path 3 parallel...

forum.beyond3d.com

You had more main RAM than Dreamcast, a much bigger external storage solution, and a lot of bandwidth between main RAM and graphics processor. It was not easy, but there were tons of ways to tackle issues you may have been facing.

One guy from beyond3d forum suggest MPEG2 decoder wasnt useful in real games.

Was GC more or less powerful than PS2? *spawn

In terms of raw processing GC was ahead of PS2 in most areas. Hardware T&L on flipper was 8 Gflops, plus 2.5 Gflops on the CPU, 10.5 Gflops total, compared to only 6.2 Gflops for PS2s entire chipset, and even at that you had to pull some magic tricks to get that out of it. The pixel side of...

forum.beyond3d.com

 

[Panajev2001a]

Still in the end, texture work always appear largely better and cleaner on Dreamcast.

We do not have a huge selection of titles overlapping to make a great call (as the Dreamcast died a bit early), but having double the memory and better HW compression lowers the bar implementation wise so I can see how even moderately early titles would compare favourably.

When I see games like MGS2, GT3, and FFX I see quite good variety and quality which only got better as time went on (see FFXII for example).

 

[Panajev2001a]

One guy from beyond3d forum suggest MPEG2 decoder wasnt useful in real games.

Was GC more or less powerful than PS2? *spawn

In terms of raw processing GC was ahead of PS2 in most areas. Hardware T&L on flipper was 8 Gflops, plus 2.5 Gflops on the CPU, 10.5 Gflops total, compared to only 6.2 Gflops for PS2s entire chipset, and even at that you had to pull some magic tricks to get that out of it. The pixel side of...

forum.beyond3d.com

Other developers in the same thread disagreed (assumption on having to decode from jpeg to palettised textures for example), but then again it was a staple of how Sony designed consoles: longer lifecycle and versatile HW that could punch way above its weight, but Had a very steep learning curve.

 

[Fafalada]

ps2 has half that, 4mb with no hardware compression, that's why Dreamcast games has better textures than ps2 or GameCube games.

On a "per frame basis" PS2 had 32-34 MB of usable graphics memory. GCN had 24-25MB, DC had less than 8 (after subtracting required space for Geometry and Framebuffers, it was closer to 4-5MB IIRC).
DC gets some effective space back since it could use 2bit textures (4bit minimum for GC/PS2), but it's still pretty lopsided overall.

Maybe GC wasnt much faster compared to PS2, but it had S3TC and it made a difference for sure when it comes to texture quality. The same game Time Splitters 2 on PS2 and GC.
GC

PS2

This isn't an example of using higher-res textures - it shows the ease/strength of GCs multi-texturing relative to PS2.
The base color-map is identical between the two, GC has an additional texture stage that adds detail/noise texture (which is a small, tiled texture, not adding a lot of data in memory).

 

[cireza]

When I see games like MGS2, GT3, and FFX I see quite good variety and quality which only got better as time went on (see FFXII for example).

I don't find late PS2 games to have much better texture work. I am still very impressed when I look at Skies of Arcadia, Sonic Adventure 2, Phantasy Star Online, Quake 3, Shenmue or Grandia 2 on Dreamcast. Textures are very high quality, more varied, more colored, more detailed. And this is a quick selection of games, but this is true for all games globally.

 

[Rolling_Start]

I don't find late PS2 games to have much better texture work. I am still very impressed when I look at Skies of Arcadia, Sonic Adventure 2, Phantasy Star Online, Quake 3, Shenmue or Grandia 2 on Dreamcast. Textures are very high quality, more varied, more colored, more detailed. And this is a quick selection of games, but this is true for all games globally.

FFXII is a good example of PS2 texturing, as it was a single platform, 3rd gen game. Lots of variety and very skilfully made.

It also demonstrates the PS2 trait of potentially high variety within a scene, but a monotone look for individual textures (with 16 colours you're very limited). This is possibly what you're referring too. It's not something the PS2 could move away from so long as it was heavily using 8 and particularly 4 bit CLUTS, as it did throughout its life.

DC could more effectively support subtle gradients *and* high contrast, differing hues within the same texture (as could GC and Xbox). It's one of the big things that leads to the difference in 'look' between PS2 and DC (that and the massively higher poly counts on PS2 ). You were never going to get the same kind of *pop* out of 4-bit CLUT heavy textures on PS2. It's just one of the hardware traits, just like you could never get FMV out of the DC that matched the PS2.

Seriously, get some screen grabs of PS2 games and check out the colours in an image manipulation program!

 

[cireza]

that and the massively higher poly counts on PS2

Not discussing the fact that PS2 is much more capable in other topics. Of course it is. Will take a closer look at some native screenshots.

 

[polybius80]

I dont know if ps2 can indeed use some clever software texture compression algorithms like you say, but pretty much every game that I have played on PS2 had blurry textures. Even textures in dreamcast games were much better (quake 3) compared with PS2 ports, so I'm guessing it's has something to do with hardware texture compression, because Dreamcast has less RAM compared to PS2.

It's funny you mentioned area 51, because I have played it on my xbox for the first time just today. For some reason I always thought it's a very bad game, but I like it. Area 51 at least on xbox looks surprisingly good, definitely better compared to Time Splitters 2-3, but of course time spliters are 60fps games.

dreamcast have very slow memory, its clear wasn't designed to update its texture buffer very often during framerate or at all, PS2(and GC) are designed to updtae its texture cache many times during frametime that is why it have a huge bandwidth in comparison and fast access to texture cache, I dont have the link but there is an interview to a developer that at the time was confused and then realized how it works so change it opinion and comments about how texture worked lot of dev have that problem with ps2 when they moved from DC , you cannot expect to have DC ports and then have better textures without changing how the texture streaming works, they are cheap ports most of the time to a very different architecture, some ports just adapt the DC game to PS2 and just use the extra power and ram, so improvements dont touch textures, for example unreal tournament have more complex scenes on ps2 and model variety while DC dont have those scenes and use 1 or 2 player models(repeated for all characters) and even omits female models


area 51 is a fantastic game, its never too late, I loved that game back in the day

Time Splitters games are very fun and yes they are 60 fps but are incredible simple(3 is the better) the scenes are so simple that they can use few textures and make sure they are good resolution while more complex games require more different textures, even then TS are very simple compared to other 60 fps shooters

there are other FPS that are hidden gems like darkwatch

 

[polybius80]

I don't find late PS2 games to have much better texture work. I am still very impressed when I look at Skies of Arcadia, Sonic Adventure 2, Phantasy Star Online, Quake 3, Shenmue or Grandia 2 on Dreamcast. Textures are very high quality, more varied, more colored, more detailed. And this is a quick selection of games, but this is true for all games globally.

you basically mentioned the best DC has to offer, except for quake3 I have those games and I disagree about textures

skyes of arcadia is a turn based combat game, scenes are relatively well textured but static, and enemies are not well textured and characters, compared ot other games is very simple

grandia 2 is a very good game and well textured, the ps2 was a very cheap port in comparison but even then it cannot be compared with later RPG like dark cloud 2 for example that game have more varierty of textures and good quality you cant compare trees and plants texture

while ports generally have better textures on DC is not always the case



the PS2 version even have better models on top of better textures and light specially in the angle with floor that look washed on DC and later ps2 games have much better textures like area51 and darkwatch

DC have very good texturing capabilities but a PS2 game well done show same or better quality with more variety plus more effects

 

[Rolling_Start]

I think the whole texture thing is a bit more nuanced than we sometimes appreciate.

4-bit CLUT requires twice the space for the same resolution of texture as 2-bit VQ (give up take lookup table data). So to match 5.5 MB of 2bpp VQ you're looking at roughly 11 MB of 4 bpp CLUT. Thing is, that only allows 16 colours per texture, which is horrible for some things, so realistically you're going to be using a mixture of 4 bpp and 8 bpp CLUTs so it's going to be more than that.

Now how you use 4 and 8 bit CLUT, and how much you use CLUT or 4bpp VQ on DC etc is going to affect the actual difference in memory used. But in the general case you're going to be looking at rather more than 11 MB of ram needed to match *resolution* of DC textures on PS2. But that's still going to leave you massively behind on colours (no way around it using CLUTs). And on PS2, unlike DC, the textures are stored in memory that also used for meshes, animation world data etc. So it's going to get squeezed.

In the case of DC -> PS2 ports you're sometimes looking at large textures relatively rich in hues and gradients, and that probably means greater pressure to use 8-bit CLUTs. And that ... probably means sacrificing resolution.

In games designed solely for PS2 the artists will be under pressure to create textures that wherever possible work okay with 4 bit, 16 colour textures. It might even pay to split what would be a single large texture on (DC, GC, XBox, PC) into group of smaller 16-colour textures if possible to save on memory. E.g. a single object skin might work broken up into a group of smaller ones with different colour groupings on each. And decal the heck out of things. PS2 has the fill for it!

For sure, PS2 benefited from being the market leader. If texture assets had ever been made with the Xbox in mind and then just thrown at the PS2 I think it'd have been carnage!

 

[cireza]

DC have very good texturing capabilities but a PS2 game well done show same or better quality

I am going to totally disagree here. I have not seen anything on PS2 even coming close to the games I gave as examples.

 

[Panajev2001a]

And on PS2, unlike DC, the textures are stored in memory that also used for meshes, animation world data etc. So it's going to get squeezed.

TBDR’s need to store (capture) the entire scene they need to render in that render pass in memory, post transform, before binning, performing the HSR step, and rendering It tile by tile. Hence F Fafalada ’s comment about real usable texture memory in VRAM, front and back buffers also excluded, being considerably less than 8 MB.

For sure, PS2 benefited from being the market leader. If texture assets had ever been made with the Xbox in mind and then just thrown at the PS2 I think it'd have been carnage!

Indeed, but then Xbox supported better compressed texture formats in HW and had double the main RAM... benefits of being a console released over a year later (for the 1999-2004 period, 1.5 years is almost 1.5-2 GPU generations ahead) and with a budget so high that it was good for MS to be able to move to the next generation early and not have to ramp up volume too too quickly and too high .

Then again what if PS2 has shipped with say an 8 MB eDRAM variation or the GSCube version with 32 MB of eDRAM ?

 

[pawel86ck]

area 51 is a fantastic game, its never too late, I loved that game back in the day

Time Splitters games are very fun and yes they are 60 fps but are incredible simple(3 is the better) the scenes are so simple that they can use few textures and make sure they are good resolution while more complex games require more different textures, even then TS are very simple compared to other 60 fps shooters

there are other FPS that are hidden gems like darkwatch

Time splitters 2 was my first game (along with Burnout 2 ) on PS2. Graphics in this game wasnt that bad in 2003 but back then I have already played xbox games like Splinter Cell 1, halo 1 or Enclave, so Time Splitters 2 wasnt that much impressive when it comes to graphics. But it was still extremely fun game overall, and it's visual style was really interesting. I'm playing xbox version right now and I cant stop playing it after so many years .

When it comes to area 51, I have also played PS2 version today. Graphics (textures, effects) looks the same as on xbox but peformance is just really bad, sometimes it looks like 15-20 fps and it's very hard to aim. Maybe Time Splitters 2 doesn't look as good, but at least performance is wayyyy better.

 

[Rolling_Start]

TBDR’s need to store (capture) the entire scene they need to render in that render pass in memory, post transform, before binning, performing the HSR step, and rendering It tile by tile. Hence F Fafalada ’s comment about real usable texture memory in VRAM, front and back buffers also excluded, being considerably less than 8 MB.

Yup, I was taking that into account.

If you wanted to support VGA you needed 600 KB each for front and back buffers (internally rendered at 24-bit, dithered down to 16-bit in the move off tile) with around 1 ~ 1.5 MB of memory for geometry depending on how many polygons you needed to accommodate at peak.

5.5MB for textures was figure given by a developer I saw, might have been the DC Quake 3 developers. Games that pushed more polygons per frame would have had less for textures, games that ran at 60 fps (Q3 was 30) probably more.

Interestingly, some areas of Shenmue limited the number of NPCs that could be present in some areas at the same time. It'd actually stream them in from the drive when it could. Probably to do with juggling memory contraints.

Games that didn't care about VGA support could get away with as little as 450KB per buffer (filtered down to half height, retaining 24-bit colour).

Indeed, but then Xbox supported better compressed texture formats in HW and had double the main RAM... benefits of being a console released over a year later (for the 1999-2004 period, 1.5 years is almost 1.5-2 GPU generations ahead) and with a budget so high that it was good for MS to be able to move to the next generation early and not have to ramp up volume too too quickly and too high .

Then again what if PS2 has shipped with say an 8 MB eDRAM variation or the GSCube version with 32 MB of eDRAM ?

My main point was that the PS2 was unique in it only using CLUTs for native texture compression. Textures that wouldn't have worked nicely with being crunched down to 16 colours (probably quite a lot) would have required 8 bpp CLUT to minimise degradation, which would have put enormous pressure on a system that already had lower texture resolutions.

An 8MB eDram GS would have been interesting (and probably a supersampling beast!), but it wouldn't have gotten around storing relatively large (memory wise) or low colour textures in that 32MB of main memory. Something like VQ texture compression would helped across the board -main memory space, texture quality within edram, texture colour variety.

PS2 is interesting not only because it was a phenomenon within the market, but because it was both visionary and anachronistic at the same time! I mean, a DVD drive and almost vertex shader like vector units combined with a mid 90's style rasteriser ... that was on steroids.

God, the current console market is boring in comparison ...

 

[Panajev2001a]

Yup, I was taking that into account.

If you wanted to support VGA you needed 600 KB each for front and back buffers (internally rendered at 24-bit, dithered down to 16-bit in the move off tile) with around 1 ~ 1.5 MB of memory for geometry depending on how many polygons you needed to accommodate at peak.

5.5MB for textures was figure given by a developer I saw, might have been the DC Quake 3 developers. Games that pushed more polygons per frame would have had less for textures, games that ran at 60 fps (Q3 was 30) probably more.

Interestingly, some areas of Shenmue limited the number of NPCs that could be present in some areas at the same time. It'd actually stream them in from the drive when it could. Probably to do with juggling memory contraints.

Games that didn't care about VGA support could get away with as little as 450KB per buffer (filtered down to half height, retaining 24-bit colour).

My main point was that the PS2 was unique in it only using CLUTs for native texture compression. Textures that wouldn't have worked nicely with being crunched down to 16 colours (probably quite a lot) would have required 8 bpp CLUT to minimise degradation, which would have put enormous pressure on a system that already had lower texture resolutions.

Oh indeed, as you say after, having additional HW based texture compression support in the GPU would have been better and made it easier for the developer: 16-256 colors (even with dithering) forced a lot more work on the developer and especially the artist that, especially as other consoles jumped on the horizon, would become more and more of an issue. You can work around limitations with CLUT, optimise you’re streaming around 8 bpp CLUT and uncompressed via various tricks (selective upload of the only mip levels needed... break down textures in smaller unique chunks, break complexity in multiple blended layers, one per pass) would help... but it is yet another thing you need to create, debug, and optimise. Time is a very limited resource.

An 8MB eDram GS would have been interesting (and probably a supersampling beast!), but it wouldn't have gotten around storing relatively large (memory wise) or low colour textures in that 32MB of main memory. Something like VQ texture compression would helped across the board -main memory space, texture quality within edram, texture colour variety.

PS2 is interesting not only because it was a phenomenon within the market, but because it was both visionary and anachronistic at the same time! I mean, a DVD drive and almost vertex shader like vector units combined with a mid 90's style rasteriser ... that was on steroids.

God, the current console market is boring in comparison ...

Oh indeed, you are not the first one talking about the modern market in these terms lol.
For me PS2 is a bit like The Barbican outside and inside: someone in the past looked at reality and trends and problems around them at the time and projected them in the future obtaining something a bit out of time (it is not 70’s, it is not modern).

The GS is one of the older parts of PS2 afaik and, as narrated in their own papers way back when, was the first item they completed specs wise and the EE had to be engineered to feed it properly: this is where I feel like they landed way past you would have expected given the GS (something clearly aimed to brute force away the problems they saw developers having in the PS2 generation: you got to admire how fast it was at some things it had no business doing... one of the few graphics processors where render state per frame let alone per mesh changes make it laugh ... flush buffers, change primitive type, change texture, etc... It would just keep going, while other GPU’s had to minimise draw falls and state changes a lot more not to lose speed... sorry I get carried away hehe).

The VU’s are way beyond VS of the time as they are fully programmable (so many moving pieces in PS2, quite amazing to learn about and make use of: DMAC, VU’s and their VIF, tags for the GIF, packets for the GS...) and packed with quite a lot of registers and local instruction and data memory for the time... very powerful and forward looking. It was VS + GS and then some since the GS did no vertex processing at all beyond setup and rasterisation. I still find its ability to render triangle strips of arbitrary length by hiding degenerate geometry with a single bit flag amazing .