Analysis of resources

[tenken]

Hi, this is my first post after lurking for about two years (I hate to get into forum arguments and usually avoid posting!). Anyway I need to calm my burning curiosity while hoping that this thread WILL NOT TURN INTO A CONSOLE WAR ZONE. Let's concentrate on the technical aspects. Being neutral on consoles (good games on all of them), and just a lover of new tech I always wondered why a lot of members hardly expect any graphical differences even though the ps3 seems to have quite a bit more resources to be exploited on the paper specs.

Let's use the example of say a fps game where computational power seems to spread out nicely among the ai, physics, graphics and sound game code. The SPEs of cell could be divided among the gaming code and there could be situations where about three or more SPEs could be used for graphics with the rsx. The interview with Mike Acton sums up nicely all the tasks we could expect them to do. The 360 doesn't have such resources at its disposal since the xecpu cores may be used to push the other gaming code and for decompression as was shown in early documents. Of course there is also physics improving the look of the game in motion which also on the SPEs side.

So basically is it some of the abilities of xenos such as good dynamic branching on the pixel shaders and vertex texturing (what sort of graphical effects could we see from these techniques by the way) etc. that lead to some taking the no difference position? Maybe the graphical tasks being done by the SPEs are not enough to show any big graphical differences? Or is it that the ps3 does not have available bandwidth to exploit the resorces sufficiently?

I am not so knowledgeable in the deep technicalities and will like to have some decent explanations. I hope this is not flame bait. Honestly I see it as a tri-core or more processor specialised in graphical tasks with a good gpu against a good gpu with some pretty neat abilities with minimal help from its processor. I know its developers talent that count in the end but i just want a little technical discussion. Especially welcomed to Acert, Mintmaster, Nao and Fafalada since you guys seem to know a lot about producing graphics. Thanks.

 

[cthellis42]

Hi, this is my first post after lurking for about two years (I hate to get into forum arguments and usually avoid posting!). Anyway I need to calm my burning curiosity while hoping that this thread WILL NOT TURN INTO A CONSOLE WAR ZONE.

Screw YOU, man! Jaguar forever! Woooooooooooooooooooo!!!!!!!!



(Sorry, was asking for it. -_^ )

 

[Shifty Geezer]

I always wondered why a lot of members hardly expect any graphical differences even though the ps3 seems to have quite a bit more resources to be exploited on the paper specs.

There's several possible reasons I can think of, though I don't know which one(s) particular forum members subscribe to...

1. Good looking graphics will be limited quite a lot by pixel shaders, and Cell won't help with that so well, leaving Xenos and RSX as similar performers
2. Games will mostly be multi-platform or running on cross-platform engines that limits performance toward the lowest common denominator. If Cell could add 50% to the graphics side of things, it likely won't be used as such.
3. The graphics possible on XB360 are high enough that to make a noticeable improvement, you'd need considerably more rendering power. If you consider XB vs. PS2, though the XB has more graphical power, to the average Joe the two platforms are similar in appearance. Things like texture res and MSAA don't always make for a big visual difference even if they're using lots more resources to be achieved. It depends how people measure a graphical difference as to whether there is a big one between PS3 and XB360 or not!
4. The possibility that to use Cell requires talent/experience/investment that most developers won't/can't provide, so even if Cell can add a lot to the visual appearance of the games, that potential will go unused (similar to argument 2 above). Think of PS2 and how top-end devs got great looking games more comparable to XB, whereas the majority couldn't milk the hardware so effectively.
5. Other limits. Power is just one part of the equation. If your system is short on BW say, you can't get the data moving fast enough to use that power. XB360 has some architectural features that we're hoping will result in efficient image quality improvements, while it's not clear on PS3 how those same IQ systems can be implemented and whether if they can (like 4xMSAA) whether that's in a way most devs can use or only the good ones. Likewise if XB360 has better data streaming from optical drive, perhaps it'll see generally improved texture quality over PS3. What advantages one platform has over another is the glut of discussion on this forum, and you'll find nobody can agree
6. The diplomatic response, where a forumite expects PS3 to look better but doesn't want to say so.
7. The fan--- argument, where one platform will likely look better than another but they don't want to admit it

Of course these points can be flipped over, and instead of starting with the question 'why won't PS3 games look better?' you can start with 'why won't XB360 games look better?' and use the same arguments...'engines will likely be more efficient and easier to write on XB360 so why won't it look better than PS3?'

I suppose if you were really interested where people stood, you could ask a poll 'which platform will on the whole look the best visually, XB360, PS3, or neither' though I think that may have been done already. Perhaps a year ago and maybe people's opinions have changed. If you were to have such a poll, you'd need to clarify whether 'on the whole' means a mean average of best+worst, or a median of the most common look to games. Or add another couple of options 'both will look similar but ..... will have better looking exclusive AAA games.'

 

[nAo]

1. Good looking graphics will be limited quite a lot by pixel shaders, and Cell won't help with that so well, leaving Xenos and RSX as similar performers

Since GPUs shade quads and not pixels it's really important that one submits to them relatively big and isotropic (in screen space) primitives in order to not a waste a LOT of your computational power, at the same time this also improve color and z buffer compression ratios.
So it's not just about quantity (high poly count) but also about quality (well-formed triangles), and CELL can potentially help a lot in this department, thus it can also improve pixel shaders performance

Marco

 

[Fafalada]

# Good looking graphics will be limited quite a lot by pixel shaders, and Cell won't help with that so well, leaving Xenos and RSX as similar performers

As far as I can see, good looking graphics are currently very commonly limited by ugly, aliased, lowres shadows. And off the top of my head, I can think of at least 3 ways how Cell could potentially help out with that.

Though what nAo said is probably more relevant - shadow artifacts are just one of my current pet peewees.

 

[cthellis42]

As far as I can see, good looking graphics are currently very commonly limited by ugly, aliased, lowres shadows. And off the top of my head, I can think of at least 3 ways how Cell could potentially help out with that.

Though what nAo said is probably more relevant - shadow artifacts are just one of my current pet peewees.

With comments like that, Faf, I'm counting on you to make a next-gen spiritual successor to Ico.

 

[assen]

The SPEs of cell could be divided among the gaming code and there could be situations where about three or more SPEs could be used for graphics with the rsx.

This is highly debatable - I personally think that due to the difficulty of dividing gameplay code amont the SPEs, even more SPEs will be typically used for graphics/physics/procedural content.