Would you not say R & C would push Xenos? If not, how so?
I think R&C looks great! I also think that Xenos would render it with ease on it's own, and would have the added bonus of having a few extra dozen megabytes of memory free for other use.
scificube said:
In any case wouldn't it be a more valuable discussion to talk about potential solutions to problems than the hardware given the hardware is fixed and there is no way to change it now?
Our parent company hasn't blessed us with a Killzone style budget and timeline, so there are constraints as to how many of these solutions we can implement. It's even harder to explain to the suits why you managed to get something working so fast on 360, yet suddenly you're asking for lots more time to accomplish the same thing on PS3.
I'll give you a very simple example. In replay mode, the user moves the camera around anywhere in the game and zooms in/out to their hearts content. They decide to zoom in real close to a players face. This face has a very expensive pixel shader because it must look like the real life player. With the camera is zoomed in really close to the face, it means the overall scene vertex count is dramatically reduced, since the user is zoomed in on one thing in particular. This also means that the face is occupying the entire screen now, which normally is not a good scenario for an expensive pixel shader on older graphics hardware. However, Xenos, given that the main load is now pixel side, will automatically shuffle its shader units to handle more pixel load and less vertex load.
Next, the user zooms the camera out really far so tons of geometry is visible, but none of the very expensive pixel shaders are taking up much screen space anymore. Problem? Nope, because Xenons will again shuffle its resources automagically and dedicate more of them to vertex processing than to pixel processing.
Problem solved! Or is it?? Well it turns out no, because none of the above is handled on PS3. Got a view with < 500k verts but expensive pixel shaders? Too bad, the vertex units will idle. Got a view with a huge number of verticies but simple pixel demands? Too bad, the pixel units will idle.
Can this problem be overcome? To a certain extent it can be, but it's not easy. Along side typical geometry lod's now you have to write shader lod's. A pain in the ass which isn't 100% foolproof either.
Take alpha for another example. Alpha as most know is terrible on RSX. Fine, so we ditch it and go with alpha coverage. It is after all claimed to look as good as alpha to the common user. Well guess what, that's not always the case! You can end up in situations where, depending on the art, alpha coverage results in an image with more 'noise' compared to just going with regular alpha. More specifically, imagine a chain linked fence that covers the entire screen with lots of people behind it wearing all kinds of different clothing. Zoom back to just the right amount and suddenly that fence looks more like noise than a fence, compared to the alpha version. What's the solution here? We can't use alpha because its a full screen draw that would kill RSX. We can't render to a smaller buffer because we need to preserve the detail. So after taking time to try different things, we just settle on living with noise at certain camera distances.
Anyways, I can go on and on, but that's just two examples on many such graphics related issues that arise when you try getting a Xenos pushing game over to RSX. The ultimate question is will people have time to resolve them all in their budget and time constraints. The further question is if you happen to have a stupidly large budget and timeframe, will you even be able to recoup your costs? I know people here love to talk about Killlzone, but that's because you don't have to foot the bill (how much was it again, > 40 million? ). Believe when, when that game finally ships bean counters everywhere will be watching it very carefully. If enough of these huge budget titles fail, it will have definite ramifications on the industry.
