• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

PS3 vs 360 hardware comparison - Clarifications

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

diaz

Member
Joined
Aug 27, 2007
Location
Canada
Here is a side-by-side

Comparison of GPU specs
Clockspeed in MegaHertz: 550 (PS3), 500 (360)
GigaFlops (Billions of Floating Point Operations Per Second): 1,800 (PS3), 240 (360)
Billions of Dot Products per Second: 1,800 (PS3), 240 (360)
Millions of Polygons per Second: 900, 1,800 with CPU handling culling (PS3), 500 (360)
Shader operations per clock cycle: 136 (PS3), 96 (360)
Billions of Shader operations per second: 74.8, 100 with CPU (PS3), 48 (360)
Texture lookups per clock cycle: 24 (PS3), 16 (360)
Texture lookups per second: 13,200 (PS3), 8,000 (360)
Vertex/Pixel Shader pathways: 24 pixel, 8 vertex (PS3), 48 shared (360)
Shader ops per pathway: 5.7 (PS3), 2 (360)

Video RAM
Amount in Megabytes: 256 Dedicated, 256 Shared (PS3), 10 Dedicated, 512 Shared (360)
Clockspeed in MegaHertz: 700 Dedicated, 3,200 Shared (PS3) Unknown Dedicated, 3,200 Shared (360)

So as you can see, 360’s GPU is not superior to RSX in any single way. Using official specs.

As for Bluray vs DVD

Disc Transfer rate in Megabytes per Second
-Single Layered discs
PS3: 9 (average: 9)
360: 6.65 to 16 (average: 13.3)

-Dual Layered discs
PS3: 9 (average: 9)
360: 4.389 to 10.64 (average: 7.5145)

Seek times are also faster on Bluray since the data is packed far closer together. And 99% of 360 games come on dual layered discs


The PS3 has the upper hand in all areas. Theoretically, the GPU is about 2-3x faster on the PS3. Although the 360's unified 48 pipe architecture is very efficient, it only performs 2 operations per pipeline. The PS3 has dedicated 24pixel and 8 vertex, but they are both capable of 5.7 operations per pipeline.

The use of 7 cores on the PS3 might seem overkill, but in reality it allows for physics headroom. More objects onscreen with more interactivity with the environment, such as destroying buildings + trees.

The entire performance is dependant on programming. If games would make use of the PS3's entire power, games would appear somewhat better and more interactive. But not by a whole lot.

The Xbox 360 has the right formula with the unified architecture, but needs a boost when it comes to shader speed and ops per cycle. The 360 is obviously easier to program since it is much more straightforward.

What everything boils down to, if a game would be perfectly programmed for both consoles; in terms of visual quality, would be slight to medium noticeable visuals such as additional viewing distance and maybe better AF/AA for PS3. In terms of performance, identical visual quality would yeild about 1.5x to 2x the framerate on the PS3.

To me that is not a reason whether to decide on one or the other, since the visual quality would only be noticed through side-by-side comparison and even then, once you really get into a game you pay less attention to graphics. Sometimes better graphics do help enhance the experience, also makes it easier to see vilans in the distance etc...

Real world; programmers seem to optimize games very well for the 360. They have a very good grasp on its hardware and the united architecture is a programming paradise. Games will tend to look better on the 360 since the optimisation is easier. On the PS3, programming specific instructions for vertex and pixel shaders is difficult and time consuming. Several programmers will instead decrease visuals until a suitable level of performance/visuals compromise is achieved. If programmers would balance the pixel/vertex shaders and instructions per cycle, they would be able to crank the visuals way up. The Cell processor is majorly untapped, and won't be for another while. Parallel processing is just at its programming birth and very little programmers are SME's in the matter.

When it comes to real world, the 360 has had the upper hand for a while now, and as PS3 titles are comming out, the PS3 is looking better and better as programmers get the hand of the code. The PS3 will continue to look better as games take advantage of its full potential.

As for selecting a console, any of the two are perfectly suitable for gaming. I would say they are even. Tough choice ;)

-D
 
Yeahhhhh... That's all well and good.

But there's a reason why mathematicians can't get girlfriends (j/k before you ask :) ):

Numbers lie.

On your driver's license application it asks for your height and weight: You can put any number you want. So it goes for product specs. You have to be pretty clear about what hardware is present and whatnot... but as far as what that hardware can actually DO... you can be as approximate as you want.

If your average is somewhere between 7.1 and 8.5... you can just say your average is 9. Or you can just write down what your speeds are on a GOOD day. Notice how you have a lot of averages for the 360 next to solid numbers for the PS3. I also noticed you're not listing your sources... so YOU could theoretically be lying yourself. (Not saying you are...)

So yeah. Mathematicians probably have more money, drive nicer cars, and even get to travel a bit more... but what are they doing friday nights?
 
PS3 hardware is definitely impressive, and way superior over the Xbox360. However, it's their software that suck. That's why most games today, the PS3 is no more or just slightly better looking compare to the Xbox360. Also, MS developed the Xbox360 in a way that allows PC developers to join the console war. That definitely helps the Xbox360.
 
The numbers, wherever you copied them from, look like they are straight from Sony propaganda. No mention of what is included in the 360 to give free AA and scaling. Anyone who's been following video cards for the last 10+ years also knows those sort of number comparisons are meaningless for predicting real world performance. For all the imbalance in numbers shown, most developers actually point to the 360 as having a minor GPU performance advantage.

What's really needed is exhaustive and objective benchmarks. That's not practical, so all we can do is look at the games and the context surrounding the games.
 
A performance advantage does not mean its better. It just means its got something going for it. Unified architecture. Its a great technology, and the one to have. The PS3's RSX simply has 2-3x more raw power. You can hack it down however you want, but when it comes to numbers, the features that DO matter are the shader pipelines and operations per cycle, as well as the texture filtering units. They are the bottleneck in visual quality/performance.

-D
 
A performance advantage does not mean its better. It just means its got something going for it. Unified architecture. Its a great technology, and the one to have. The PS3's RSX simply has 2-3x more raw power. You can hack it down however you want, but when it comes to numbers, the features that DO matter are the shader pipelines and operations per cycle, as well as the texture filtering units. They are the bottleneck in visual quality/performance.

-D

It's all blah-blah-blah because even counting pipelines, units, and clockspeeds are all gross over-simplifications of what is actually happening in the hardware and software to accomplish tasks. Unless the architectures are the same or extremely similar except for higher clocks/more parallelism, it's near useless for comparison. I can't count the number of times I've compared such statistics for both professional and gaming applications and discovered that the numbers told a vastly different story than real world benchmarks and applications did.

In other words, when nVidia comes out with a new version of their own chip, kept the architecture nearly the same, and beefed it up for more performance, then start counting pipes, clocks, and units. Then, when ATI/AMD comes out with some new architecture and you want to compare it to nVidia, ignore all those numbers that don't have any real point of reference and look at benchmarks. Look at the results.

A performance advantage does mean it is better if it is with the application in question.
 
Last edited:
Where did you get those numbers?

I can tell you from different developer sites and things I have read from devs they say the 360 has the more powerful GPU.

The 360s Xenos has the 48 parallel floating points, that can be used for vertex or pixel shading. The PS3 has 24 pixel and 8 vertex, I think this is where the big limitation is on the PS3 GPU and why the 360s GPU is actually more powerfull.

I will agree the PS3 is harder to develope for, this has been true since the PS2. So they can definitely improve on the quality of the graphics and the FPS. But how much who knows, I am sure the PS3 has a powerfull CPU but how much can it does when its bottlenecked by that GPU?

Again I am not sure all of those numbers are apples to apples... at least in terms of true real world performance we know they are not...
 
The 360s Xenos has the 48 parallel floating points, that can be used for vertex or pixel shading. The PS3 has 24 pixel and 8 vertex, I think this is where the big limitation is on the PS3 GPU and why the 360s GPU is actually more powerfull.

This is the key factor that most people misconceive. The 48 pipes only resolve 2 operations per cycle. The 24 pixel solves 5, and 8 solves 2. The big difference here is that the 48 from the 360 are dynamically programmable in two ways. The first is the most evident; pixel and vertex can share the pipes on the fly. The second; is the type of ALU's used are also programmable on the fly. The ALU can be any of the following: Vector, scalar, dual/co-issue, fog or texture.

So for the 2 ALU's (operations per pipeline on the 360) they can be any combination of Vector, Scalar etc.. on any pipe at any given time. That is the whole beauty of the 360, its not fussy.

On the PS3 side, the RSX uses 24 pixel and 8 vectors. Each of these have a certain specific way and order of handling shading;
- 24 pixel pipe handles 2 x Vector, 2 X Scalar/dual/co-issue/fog, and 1 Texture.
- 8 Pixel pipe handles 1 x Vector, 1 x Scalar/Dual-issue

Potentially, the PS3 has the most power. If the game code is given in the "2vector, 2scalar, 1texture" format, then the PS3 is in heaven. But programmers dont always program games that way. If the code is programmed 1vector,1texture then 1scalar,1texture ect... then the RSX's efficiency is going to be decrease considerably since it is wasting cycles to accomplish the work. You can see how this quickly becomes a nightmare for programmers. For the same job (1vector,1texture) the xbox 360 would be signing along @ full potential.

On the other hand, if the game code is written 2vector, 2scalar the PS3 has the advantage. It is wasting its texture spot, but at least the pipe is populated with 4 ALU's. In the same scenario, the 360 would have to split the job into 2 separate pipes, therefore decreasing its efficiency in half. Either it would separate the code into 1vector, 1scalar + 1vector, 1scalar -or- 1vector, 1vector + 1scalar, 1scalar.

The mathematical equasion is actually quite simple:
(# of pipes) x (ALU's per pipe) x (core frequency) = ALU's per cycle / Shader operations per second.

So ideally, the 360 would look like this:
(48pipes) x (2 ALU) x (500Mhz) = 48 Billion

RSX (PS3):
[(24 Pixel x 5 ALU) + (8 Vertex x 2 ALU)] x (550Mhz) = 74.8 Billion
(Specs say it reaches 100 billion when using the Cell processor in synergy)

This is only in IDEAL conditions for both machines. The most likely is the 360 where any code can be hashed with higher efficiency. Using un-optimised code on the PS3 can quickly reduce its efficiency in half. At this point the 360 will obviously have the upper hand.

P.S. Sources are posted in this thread / wiki. Im not taking sides, I'm just giving out the facts.

-D
 
One of the things to realize is that, even if a design has very high theoretical maximum output, it's not simply a matter of developers spending more time writing more complex code to "take advantage" of it.

The reason is because we're playing games. The load and task requirements presented to a GPU is constantly shifting and changing even within the same game from moment to moment. (As evidenced by spikes and drops in FPS during gameplay.) The load is dynamic. Scenes that fit nicely into and fill a rigid architecture one moment might not make for efficient utilization in the next moment. Reality is that some architectures are inherently very good at filling capacity during regular gameplay, and some architectures are inherently poor at it. In other words, one architecture might be able to reasonably expect an average of 80% utilization while another one might only be able to expect 55%.

Developers can tweak their code to squeak a little more image quality or performance out of one architecture or another, but they can't sprinkle fairy dust on a GPU and suddenly get max theoretical output all the time.
 
This thread is confusing, simply because I am not sure if the poster is asking for a carification in which system is better or if he is trying to carify to other's which system is better in either case when it comes down to what system is stronger, it doesnt really matter because as night said:

Who cares about specs? The games are the real specs...

And that is what will make most people decide whether they want to buy it or not. Not to mention that you cant do much more with a PS3 or Xbox360 that will utilize its hardware...

this is like the :

Ati vs Nvidia
AMD vs Intel
Apple vs Microsoft

Knowing more doesnt make your e-penis bigger.

Sorry for putting sticks in your wheels.
 
Back