Render resolutions of 6th gen consoles - PS2 couldn't render more than 224 lines? *spawn

sheathx013

Newcomer
PS2 can only render at 448 interlaced at best, anything higher is upscaled after being rendered at the PS2's lower 224 lines.

This is often ignored in comparisons. 480p rendering (like most Dreamcast games) and even a 640x480 PC game is a significantly higher rendering resolution than what the PS2 is capable of.

This is covered in detail in Sony's 2003 doc "How far have we Got" and every detailed comparison over the years. Typically when the lower PS2 rendering resolution is brought up it is dismissed as unimportant.
 
PS2 can only render at 448 interlaced at best, anything higher is upscaled after being rendered at the PS2's lower 224 lines.

This is often ignored in comparisons. 480p rendering (like most Dreamcast games) and even a 640x480 PC game is a significantly higher rendering resolution than what the PS2 is capable of.

Er.....you do know there are true 480p games on PS2 right?

And I don't mean a few low end games, I'm talking tens, if not hundreds of high-end AAA games.

The Linux kit for PS2 even allowed you to output native 1024x768 via a VGA cable.
 
Last edited:
The Snowblind Engine ran at 2x supersampling.
Assuming as we are that no doubt brilliant exception to the rule is not a multipass doubled scene downsampled to 640x448, it is still an exception to the rule. As poorly documented as the engine for Champions of Norath, Baldars Gate 1-2, and, ugh, I forget what else is, I wouldn't think Dreisbach's solution was immune to the limitation.

Similarly, I wouldn't think that Omikron on Dreamcast apparently, according to emulators, rendering at higher than 800x600 but displaying at 480p is something that could just be applied willy nilly to all Dreamcast games without significant performance hits or (more likely) ground up rewrites. I tend to wonder instead how much better Omikron might have been if the models were enhanced instead.
 
Assuming as we are that no doubt brilliant exception to the rule is not a multipass doubled scene downsampled to 640x448, it is still an exception to the rule. As poorly documented as the engine for Champions of Norath, Baldars Gate 1-2, and, ugh, I forget what else is, I wouldn't think Dreisbach's solution was immune to the limitation.

Similarly, I wouldn't think that Omikron on Dreamcast apparently, according to emulators, rendering at higher than 800x600 but displaying at 480p is something that could just be applied willy nilly to all Dreamcast games without significant performance hits or (more likely) ground up rewrites. I tend to wonder instead how much better Omikron might have been if the models were enhanced instead.
Actually super sampling 2x on the DC is strictly 1280x480.

Edit: some one was able to get the frame buffer of 2 super samples dc games. It's 1280x480. Massive but I guess dc doesn't do vertical ss
 

Attachments

  • tex-0008out.png
    tex-0008out.png
    463.9 KB · Views: 16
  • tex-0036out.png
    tex-0036out.png
    706 KB · Views: 15
Actually super sampling 2x on the DC is strictly 1280x480.

Edit: some one was able to get the frame buffer of 2 super samples dc games. It's 1280x480. Massive but I guess dc doesn't do vertical ss

Don't think it took up any extra vram for the buffers either (SS was done on the on-chip tile buffer), and I believe the additional rendering cost was somewhat less than 2x the cost of of rendering 640x480 (I think @TapamN has talked about this before?)
 
I thought there was a caveat there for DC as well. The mass assumption by the gaming media and youtubers that every DC game was anti-aliased alone is cringe worthy.

I was mistaken about the "How Far Have We Got" Performance Analyzer Sony document discussing resolution, apparently it only mentions pixels and height maps, or I've forgotten the discussion around these statements equating to resolution limitations. There was a "developer" comment that seemed legitimate enough:

"Achieving 640x480 was not really possible in a game because it took up so much VRAM for the 2x back buffers and the front buffer. The chip AA just didn't work. At all. And the way you got around that was by rendering to double height back-buffer, and then drawing a full-screen quad (well..a series of 32x32 quads because it was faster that way) at half height to use the texture sampler reduce the jaggies. So yeah....the guys at Melbourne House (who were awesome BTW) are talking about that technique. It was actually an advisory from Sony right from day 1 "Hardware AA is broken. Don't use it. Do this instead."

It was nowhere near as good as proper AA though, but good enough for most things. It also meant you could do some neat tricks. See the bloom/fake HDR in B2 and B3 as an example of that.

The PS2 was never really meant to render at high res. It was supposed to render at a lower resolution than the DC, but use HW AA to smooth out the jaggies. Doh!"

He was supposed to be a big wig developer at Criterion, but groups like these can have so many amazing double accounts who knows.
 
I thought there was a caveat there for DC as well. The mass assumption by the gaming media and youtubers that every DC game was anti-aliased alone is cringe worthy.

I was mistaken about the "How Far Have We Got" Performance Analyzer Sony document discussing resolution, apparently it only mentions pixels and height maps, or I've forgotten the discussion around these statements equating to resolution limitations. There was a "developer" comment that seemed legitimate enough:

"Achieving 640x480 was not really possible in a game because it took up so much VRAM for the 2x back buffers and the front buffer. The chip AA just didn't work. At all. And the way you got around that was by rendering to double height back-buffer, and then drawing a full-screen quad (well..a series of 32x32 quads because it was faster that way) at half height to use the texture sampler reduce the jaggies. So yeah....the guys at Melbourne House (who were awesome BTW) are talking about that technique. It was actually an advisory from Sony right from day 1 "Hardware AA is broken. Don't use it. Do this instead."

It was nowhere near as good as proper AA though, but good enough for most things. It also meant you could do some neat tricks. See the bloom/fake HDR in B2 and B3 as an example of that.

The PS2 was never really meant to render at high res. It was supposed to render at a lower resolution than the DC, but use HW AA to smooth out the jaggies. Doh!"

He was supposed to be a big wig developer at Criterion, but groups like these can have so many amazing double accounts who knows.

Well Sheath, I see you are still happy to throw out ad hominem attacks, this time against a member of another forum who is not here to defend themselves, and who has not been an active member of said forum for many years.
I'm curious, in the 10 + years since that was posted over at the Sega-16 forums, have you been successful in discovering and documenting counter sources to thoroughly debunk and rebut what Rusty posted there? There is a great database of excellent sources here at Beyond 3D, years of discussions on the pros and cons of the PS2 and its architecture. Maybe focus the efforts on disproving the source, instead of impugning the character of the person posting it. After all, do you have any evidence that Rusty was, as you called him a "double account/alt account".
 
Actually super sampling 2x on the DC is strictly 1280x480.

Edit: some one was able to get the frame buffer of 2 super samples dc games. It's 1280x480. Massive but I guess dc doesn't do vertical ss
You technically can do vertical AA (I've done full 2x2 SSAA), but it's not really worth it. You have have to render the top and bottom halves of the screen separately, resubmitting geometry for each half.

Another way to render above 1280x480, that I haven't tried, would be to generate the polygon lists for the PVR on the CPU. The data structures seem to support up to 2048x2048, it's just that the tile accelerator, the part the receives rendering commands from the CPU and breaks them up into tiles, only supports up to 1280x480. I doubt this way would have much of an advantage over double submitting, outside of very low poly scenes or specialized cases.

I've tried thinking of ways to trick the PVR into double rendering a 640x480 scene in a way to get vertical AA out of it, but it doesn't seem possible. It probably would have been simple to add a hardware feature to adjust how the TA operates and allow processing tiles as pseudo 64x32, 32x64, or 64x64, but that feature doesn't exist.

2xSSAA does take some extra RAM, but nowhere as much as doubling the framebuffer size, in the range of 100-200KB, depending on the scene. Each tile needs a list of what polygons it contains, and with more tiles you have to have more lists, which takes up space.

The performance cost of AA varies depending on the scene, but what I've seen in game-like situations is a cost of something like 40-60%. Long horizontal spans can be rendered more efficiently than short ones, and increasing the horizontal resolution makes the spans longer, so with the extra efficiency it's less than a 2x cost. Fillrate costs almost double, but polygon setup costs don't (they do go up some, but less than double), so some of the extra fillrate load gets hidden under existing stalls elsewhere. Large transparent quads would be closer to costing double, since they're low poly and less minified, the fillrate cost can't be hidden.

From what I've seen, SSAA with mipmaps almost always has better performance than no SSAA without mipmaps (assuming you can see into the distance with some texture minification, and not face-first into a bilinear upscaled wall), so games without mipmaps, like Shenmue or Sega GT, would see a GPU performance increase if you managed to fit in mipmaps somehow and enable AA.
 
This may or may not be relevant here, every game I've tested (and posted videos of) on the Neon 250 has shown that it had very little performance hit between 640x480 and 800x600, either 16-bit or 32-bit color. The performance hit to 1024x768 was significant, and indicated to me that the card started in the same generation as the G400 Max and Nvidia TNT. I still could have played Soul Reaver 2 at 1024x768x32bit on Neon 250 all the way through, but the performance dips were less obvious to me than at 800x600x32-bit. Even then, particular scenes caused Soul Reaver 2 to crash, so I got through those at 640x480x16.

The Power VR 2 AGP card known as Neon 250 did not, however, have the same drops in performance between 16-bit color and 32-bit color as Nvidia cards had, or the Voodoo 3k serious could not even achieve. I don't know why Matrox and Power VR (Videologic) were better at 32-bit color in particular in 1998-1999 cards, but it all seems to be about silicone budgets and RAM bandwidth to me.
 
PS2 can only render at 448 interlaced at best, anything higher is upscaled after being rendered at the PS2's lower 224 lines.

This is often ignored in comparisons. 480p rendering (like most Dreamcast games) and even a 640x480 PC game is a significantly higher rendering resolution than what the PS2 is capable of.

This is covered in detail in Sony's 2003 doc "How far have we Got" and every detailed comparison over the years. Typically when the lower PS2 rendering resolution is brought up it is dismissed as unimportant.
Pretty sure Gran Turismo 4 has both a 480i and 480p mode that is a true progressive scan mode. It would have to be able to do 480p to be able to pull off the 1080i mode it does where it renders 2 576x480 fields to make a 576x960i Interlaced signal.
 
I also remember that some other games also had a 480p option at the start of the screen.
 
Pretty sure Gran Turismo 4 has both a 480i and 480p mode that is a true progressive scan mode. It would have to be able to do 480p to be able to pull off the 1080i mode it does where it renders 2 576x480 fields to make a 576x960i Interlaced signal.
Still lower than 640x480

In 480i the internal framebuffer is 640x224

In 480p the internal framebuffer is 576*480 with reduced colour depth.

The PS2 as stated before was better suited for lower resolutions.

Output resolution ≠ internal resolution the fastest people understand it the better.

You can force 720p and 1080i in almost every single game on PS2 with homebrew apps, image clarity improves (a lot actually) but the internal resolution is still the same.
 

Attachments

  • gs_20240819115844-1.png
    gs_20240819115844-1.png
    232.5 KB · Views: 33
  • gs_20240819120043-1.png
    gs_20240819120043-1.png
    255.1 KB · Views: 32
Still lower than 640x480

In 480i the internal framebuffer is 640x224

In 480p the internal framebuffer is 576*480 with reduced colour depth.

The PS2 as stated before was better suited for lower resolutions.

Output resolution ≠ internal resolution the fastest people understand it the better.

You can force 720p and 1080i in almost every single game on PS2 with homebrew apps, image clarity improves (a lot actually) but the internal resolution is still the same.
those screenshots show that PS2 was an absolute beast still
 
If it's 960i why do you call it a 1080i mode?
Because that's what Sony referred to it as in the game itself.

Still lower than 640x480

In 480i the internal framebuffer is 640x224

In 480p the internal framebuffer is 576*480 with reduced colour depth.

The PS2 as stated before was better suited for lower resolutions.

Output resolution ≠ internal resolution the fastest people understand it the better.

You can force 720p and 1080i in almost every single game on PS2 with homebrew apps, image clarity improves (a lot actually) but the internal resolution is still the same.
It's 576x480 because that's just about 1/3 horizontally of what 960p 16:9 would be. So that scales up better at 480 lines. To do 1/3 of 1920x1080 would require 640x540 which is beyond what you can do on an NTSC system. There's quite a few other 480p games on PS2, including some of Sega's own games. Many of them you need to hold the Triangle and X button at boot to enable it. To say PS2 is incapable of doing 640x480p resolution is just ridiculous and goes against actual hard facts.

I'm well aware that internal resolution does not always equal output resolution.

[Ad hominem removed]
 
Last edited by a moderator:
Because that's what Sony referred to it as in the game itself.


It's 576x480 because that's just about 1/3 horizontally of what 960p 16:9 would be. So that scales up better at 480 lines. To do 1/3 of 1920x1080 would require 640x540 which is beyond what you can do on an NTSC system. There's quite a few other 480p games on PS2, including some of Sega's own games. Many of them you need to hold the Triangle and X button at boot to enable it. To say PS2 is incapable of doing 640x480p resolution is just ridiculous and goes against actual hard facts.

I'm well aware that internal resolution does not always equal output resolution.

Though I mean saying internal resolution is irrelevant to output isn't really a good thing either. If we go by this the Dreamcast from its official sdk had some weird scaled modes where the resolution where supposed to be more than 480p. Pretty sure the highest on the sdk was 640x560 but iam sure internally it's just 480p, doesn't make it more.
 
Last edited by a moderator:
Back
Top