Which console has best 2d performance? Not a battle.

Jabjabs

Regular
I've been thinking about this for a few days now and I've been wondering, just which machine would be the best at 2d?

Would PS2 with it's vector units be able to push through the data, would the GC with it's huge cache make for improved rendering times, and could xbox win through pure brute force and bandwidth?

In a world of primary 3d machines, it a question of how they could handle the older technology's.

By the way this isn't being made as a battle but just out of pure intrest, any information on this would be nice. ;)
 
Would depend on how much data one wanted to draw into the screen buffer. PS2 has highest fillrate (by far), but XB bandwidth is greater once the information won't fit in PS2's eDRAM. If the data isn't more than PS2 can stream it through its 1.2GB/s connection, it might still win. It can do almost 20MB/frame after all, which is about half its total memory capacity.

Of course, XB could fit a lot more 2D data than any of the other consoles, which would be the most limiting factor long before any bandwidth/fillrate limitations set in, in a real-world example...


*G*
 
then you have to consider the 2d of the nv2X series which were good, but not as good as other chips....


it's all subjective
 
To elaborate, the Guilty Gear X titles are the most impressive examples of 2D graphics I've seen and show that the Dreamcast/Naomi are quite capable there. Just with the memory upgrade a hypothetical Naomi 2 unit would bring, it would boost that even significantly more.
 
2D performance is mostly determined by raw single-textured fill rate. In that case all three consoles have more than enough fill rate to raster a 2D scene a couple of dozen times. The other relevant metric is memory capacity, which again all three consoles have more than enough of to run pretty much any 2D game that will ever be released in the future.

The previous "generation" of hardware was really the last one were 2D performance rank meant anything. The N64's carts, the PS's limited VRAM, and the Saturn's lack of functional transparency all were major limitations. The DC however had all the power any dev would ever need for 2D, and that likely will hold for all future consoles.
 
As a hypothetical entry, I'd say Naomi 2 handily.

Only in storing animation REALLY complex animation sequences in RAM and decent at managing opaque sprites...

The previous "generation" of hardware was really the last one were 2D performance rank meant anything. The N64's carts, the PS's limited VRAM, and the Saturn's lack of functional transparency all were major limitations. The DC however had all the power any dev would ever need for 2D, and that likely will hold for all future consoles.

That's fine if you just limit yourself to old fashioned sprite/map manipulation, but vector graphic animation is largely unexplored except in Flash, and IMO still offers quite a lot of potential (not to mention eating up computational resources)...
 
PlayStation 2 can pull ahead IMHO thanks to the use of the IPU to decompress MPEG2 compressed Textures ( as I-FRAMEs ): it would be able to fit the most 2D data in main RAM.
 
So you say 32MB (theoretically) of I-frame data would beat 64MB (just as theoretically) of 6:1DXTC?

I have no idea how well I-frames compress (on average; MPEG is a pretty heavy-handed compression scheme with "busy" patterns), but it'd have to be pretty darn well I guess to make up the difference.


*G*
 
I don't know which one should put out the best 2d. But the best 2d game so far this gen is viewtifull joe (sp?) The game is just amazing and its on the gamecube.
 
Actually, VJ is fully 3D. Capcom simply reproduced the "comic" look so convincingly that the character models are almost indistinguishable from their hand drawn counterparts.
 
So you say 32MB (theoretically) of I-frame data would beat 64MB (just as theoretically) of 6:1DXTC?

Easily... Even better than I-frame though would be to store an animation sequence as a stream of GOPS.

I have no idea how well I-frames compress (on average; MPEG is a pretty heavy-handed compression scheme with "busy" patterns), but it'd have to be pretty darn well I guess to make up the difference.

I-frames are similar JPEG pictures (different quantize matrix though), so quality can be dependent on the quantizer... In either case most of your sprites are going to be low color anyways so CLUTs would be more practical (not to mention you can do some pretty neat stuff by animating their palettes)...

I don't know which one should put out the best 2d. But the best 2d game so far this gen is viewtifull joe (sp?) The game is just amazing and its on the gamecube.

Just to agree with what Steve said, Viewtiful Joe is basically an excellent example of an NPR title...
 
Steve Dave Part Deux said:
Actually, VJ is fully 3D. Capcom simply reproduced the "comic" look so convincingly that the character models are almost indistinguishable from their hand drawn counterparts.

damn it looks so sweet though. I thought my 2d glory days would come back.
 
In regards to 2d rendering performance - all consoles basically have massive redundancy in the area, so whether one is faster then the other doesn't exactly mean much when they are all "too" fast so to say.

As a numbers excercise though, PS2 wins by default on 2d primitives with lots of transparent overdraw. If your scenes are opaque enough, NV2a could come close.
Ironically enough, if you used non-textured 2d primitives only (not sure if alpha blending would work or not) - I think NV2a could use it's AA fillrate(3.8GPx) - but bandwith limitations would likely still keep it in 2nd place.

So you say 32MB (theoretically) of I-frame data would beat 64MB (just as theoretically) of 6:1DXTC?
You can get 1-1.5bit/texel at good quality with I-Frames, and that'd give you up to 4:1 improvement over DXTC (and IPU can work 1bit alpha into those I-Frames also - so it's perfectly usable for in-game sprites).
 
Fafalada said:
I think NV2a could use it's AA fillrate(3.8GPx) - but bandwith limitations would likely still keep it in 2nd place.

Hm? How could it use AA fillrate when it gets that inflated AA number from the fact all subsamples of a polygon share the same color?

IPU can work 1bit alpha into those I-Frames also - so it's perfectly usable for in-game sprites).

How would you do transparency then? Can you unpack color data and alpha from separate sources into the same texture in eDRAM, or would you have to do 2D operations to merge the two?


*G*
 
Hm? How could it use AA fillrate when it gets that inflated AA number from the fact all subsamples of a polygon share the same color?
Well I said non-textured. Granted, gouraud would look a bit squareish too if you used it, but at least on constant color polys it'd work nice :p

How would you do transparency then? Can you unpack color data and alpha from separate sources into the same texture in eDRAM, or would you have to do 2D operations to merge the two?
99% of animated sprites are cutouts - 1bit alpha is all you need for that (and in fact it's actually a 3-value alpha, so you can do a bit softer edges too).
As for the flashy alpha effects you also see in 2d games:
a) they don't really have much use for pixel granularity alpha - even vertex interpolated alpha is often unnecessary.
b) they are usually not using animated textures either, so you can afford to use lower compression anyhow :p

Although to answer the other part of the question, you could always upload texture to eDram as split alpha/color if you wanted, yes.
And further speaking about multibit alpha - S3TC also drops to 8bit/texel if you use that, in which case I can make a very convincing argument that regular 8bit Clut will deliver better quality with handdrawn art as source texture material.
 
Back
Top