Hey all,

First off this is my first created thread on VGChartz so I hope you all go easy on me.

I'm a third year electrical engineering student at the University of Waterloo and I really am a computer hardware enthusiast but I must admit that there's still a HUGE amount for me to learn before I can even begin to make claims. I'd like to learn what you guys think (especially the more tech savvy amongst you).

Anyways, I was just sitting here and thinking of upgrading my PC and I started thinking about the differences in the 360's and PS3's tech. Now I've read that each Xenon core is actually of a similar architecture to the Cell's PPU. Of course the Cell also has 7 SPU's for an immense amount of parallel processing capability while the 360's CPU is good for general processing instructions. It's up for debate which is more powerful for gaming but the general consensus on tech savvy websites seems to lean towards the Cell's immense amount of "raw power" as it has shown off in more recent games with a ton of scripted events, fairly good AI, well implemented physics systems, etc. all at once, as well as things like Folding@Home. It's also capable of doing some pre-processing for the RSX if I've read and understood correctly, taking a load off of the RSX's work - reducing what most consider to be the bottleneck of the PS3's graphical prowess (that and the memory).

Now when I started thinking about all that I remembered that the 360's GPU (the Xenos) has a unified shading architecture. Essentially what this means is that all the hardware pipes can be used for shading or texturing. This differs from the RSX's dedicated pipes. Here's where I'd like you to shine some knowledge upon me. I started thinking that this would allow the 360's GPU to have more efficient utilization of its hardware and be capable of running at maximum efficiency and lower wait times for instructions as any instruction can be thrown at any free pipe unlike the RSX where instructions would have to wait for a certain type of hardware pipe in order to perform that instruction. I couldn't find much information on how this dual process natured Xenos architecture affected the processing times of individual tasks or anything. Would anyone know anything about that kind of stuff?

Also, would this "efficiency of processing" sort of make up for the Cell + RSX piggybacking processing style?

I love games and I love tech and I would love to learn a lot about both and be a bigger part of this community. VGChartz rules!

the 360's GPU is a really strong video game it was built from the ground up for gameing.

agree, thats kind of what i like about the 360 the most, yeah you can do other things on it but it was made for Friday night gaming. 

What you said is basically true. If I recall correctly, the PS3 GPU has a slightly higher clock frequency than the X360 GPU (i thought 750MHz vs 700MHz). Since the Xenon on the X360 has the unified shader architecture, every lane can handle every single graphical calculation. It can actually also be programmed to calculate physics or specific tasks, but was mainly build for graphical calculation (of course..) So you basically give it work, and it does it. The PS3 GPU however can only calculate specific tasks in specific lanes, which means, almost always some of the lanes have nothing to do, while others are being overthrown. This is the drawback of the PS3 GPU, but it can be helped with the SPE's of the cell. But since this is about the GPUs themselves and not the whole system, on average I think the PS3 GPU has an average efficiency of about 50-60% (if you program a game really well it might reach 70%) while the X360 GPU has about 90+% efficiency at all times. Besides this, the memory also plays a role. The X360 GPU shares its memory with basically the whole system. So if the GPU needs 350MB for example, they can program it that way. For the PS3 however, it's stuck with 256MB because of the split memory, and this limits the GPU. So yes, the GPU on the X360 is superior to the PS3 GPU in every single way.

they are ways of optimization RSX sony created.
deferred rendering, which was used in LBP and killzone 2 works quite well.
and offloading work to the SPU to let the RSX focus on graphics like textures.
the ps3 have 512MB XDR 3.2ghz memory also and that doens't left unused.
actually killzone uses around 420mb

where the GPU uses 200MB + around OS and other things 200mb around

also the 360 GPU have it own bootnecks like HDR. which can be done up to FP 16 like in uncharted.

or doing 1080p because 10MB is not enough to display a image 720p antialiasing + HDR fp16.

its not bandwrith speed limit but size. u need over 30mb to do so.

both system would really shine with more RAM.

I think it's actually 550 MHz for the RSX and 500 MHz for the Xenos.

As for the memory used by the two GPU's I think that'd really depend on what's happening on screen and what needs more memory at any given moment. As Jo21 pointed out, the PS3's GPU memory isn't ordinary. I don't know what makes it so different but if anyone has any links on the subject I'd like to take a look. :)

Thanks for the reply Jo21. Any idea on how much of an actual affect in terms of performance that offloading processing to the SPU's has to general performance?

Edit: Also, the 360 has extremely high speed EDRAM (10 MB) that is used as a means for "free" 2xAA without any performance hit on the CPU or the GPU. I recall reading a forum post about why Killzone 2 looks as it does (a "forced" art style) and the thread creator was very knowledgable on this kind of stuff. Anyone know which one I'm talking about? I can't remember his/her username. Anyways, they mentioned a processing technique that uses this extra 10 MB for something or other. Anyone have any idea about what this is?

This link is useful:

http://objectmix.com/graphics/144545-beyond3d-article-ati-xenos-graphics-processor-xbox-360-a.html

@The one above me: 500/550 can also be true. Don't remember. Was a long time since i looked into that stuff. Anyway, the difference in memory is simply that the PS3 has it split and the X360 i one chunk for the whole system.

Awesome! That's exactly the kind of stuff I'm looking for. Thanks!

far from it. The Xenos is rarely discredited on this site. While the RSX isnt far behind it, it is faster and is definitely better at using shaders.

I never said it was discredited. I've just never seen the point brought up that it theoretically runs at a greater efficiency than the RSX. Of course the effects of that can be overcome by using the Cell to offload some processing as mentioned above. I'd just like to get a better idea of how much this architecture actually benefits the 360's graphics in comparison to the Cell + RSX processing combo. Although from your post count you've definitely been around here much longer and much more often than I have so maybe it has been brought up a few times before.

EDIT: I also just remembered the name of what I mentioned above. It's called tessellation and I just found a discussion that actually talks about what it actually brings (or doesn't bring) to the 360's processing capabilities. I'm going to read through it now. =)