Originally posted by: Stoneburner
Originally posted by: BenSkywalker
Ben, What games have used the SPE's on the cell?
KZ2, GT5, UC, UC2- pretty much anything exclusive is going to use them along with any port that uses multiple cores on the other systems. How heavily they are used has come down to the particular title, UC2 being the only one that I am aware of that is using all of the cores.
I haven't seen updated information on this point in a while, but for all the theoretical power of the cell, games on the PS3 basically use the general processor and the GPU with the SPE's doing nothing. SUpposedly they were capable of doing physics calculations but I've read nothing concrete on that point.
Any game that uses more then two cores on the PC wouldn't run at all if it only used the PPE on Cell. Cell's cores are powerful, but they aren't 3x more powerful then a multi core PC chip per core. As far as confirmation on the SPEs handling physics, it has been explicitly documented since around a year prior to the launch of the PS3. You aren't likely to see updates as it is a given and very simplistic to do(it is very straightforward).
So the PS3 despite high raw computational power ends up not being a greatly efficient gaming platform as it has a relatively weak general processor and very weak graphics processor. NO amount of overhead on the PC and tweaking for PC can bridge the gap that has been growing since the ps3 was released.
This seems to be where people are having a disconnect. PC games have been in a decline in terms of visual quality since Crysis. The high point for PC graphics was 2 years ago. The technology can increase by orders of magnitude over what we have now, if devs don't use it all it is going to give you is more AA/AF, not better game graphics(just cleaner).
I even compared rainbow 6 vegas 2
As I have mentioned, repeatedly, every port is going to look better on the PC. Every. Single. One. If a game is capable of running on the PC, it isn't using everything the consoles offer and vice versa. The only way to compare the capabilities is by looking at the platform exclusives- for any of the systems.
You also have to take into account that pretty much every game on both the consoles and PC looks better on the PC.
And that won't change with this generation for certain. Anything that can run on the PCs is going to have an edge if you have higher end hardware as the code base needs to be able to run on lower end PCs. Anything that can run across multiple platforms is never going to push any particular platform to its' peak potential.
You are using conditional language for all of this. UC2 "would" use all core? The point isn't use of cores, the point is effective use of cores and effective use of computational power. A lengthy article (chip architect i believe) explained that Cell was a massively powerful processor for specialized computational needs BUT NOT for gaming. Following up articles that I've read have stated developers find coding to use the cell's SPE's.
Also, the theoretical computational power of cell being theoretically used as a graphics processor is WORTHLESS when you consider the fact that the computational power of the latest ati/nvidia gpu's dwarfs cell AND is specialized for graphics.
So empirically and theoretically, PC >>>> console.
Also, I don't own a PS3 but it's my understanding that equivalent xbox360/ps3 games look better on Xbox360. 360 has the more powerful gpu.... not a big surprise.
Cell actually is very powerful for graphics (particularly ray-tracing) and beats out conventional architectures by a long shot. Not so much due to the power of its cores, but due the speed of the caches (or scratch ram, whatever it's called) that the cores have available. Graphics rendering is all about the GB/s, and conventional architectures don't provide the bandwidth, even in cache, to support it.
But the thing is, graphics chips are even better at graphics. The ps3 is hampered with a tragically weak graphics chip, especially in the area of vertex processing (where the cell excels). But on a PC, Directx10 (maybe even Dx9) can be used to do just about everything cell is being used for in games, and faster. Not all games do though, I'd bet money on GTA4 offloading much of it to the cpu. And to be honest, the dynamic landscape of gta4 would be difficult to implement in dx9 hardware, and since they didn't bother to support dx10, well it explains the poor performance (even though high end quads have no trouble with the game at this point, unless you crank view distance up well beyond what the consoles are doing).
Ghostbusters on the PC had the same problem, it wasn't even taking full effect of DX9 hardware. Well, the same is likely true to the console versions, they're pretty ugly.
Right now, we're in a similar situation as we were last gen. The PC has the APIs and the hardware to handle so much more, but the consoles are holding it back. The last gen of consoles kept things at the DX6/7/8 level (depending on what console it was ported from), this gen is keeping things at the DX8/9 level. The huge FLOPs potential of the console cpus is causing devs to program games as if they were still dealing with fixed function hardware, and ignore the vertex shaders of the gpus. On the PS3, this is because the Cell is more powerful at vertex shading, and on the 360 it's because it supports features beyond DX9 (and tessellation is even beyond dx10), so when the games are ported to the PC they use cpu routines to accomplish the same effects.