djroberts said: Entroper said: Oh, the ignorance. The Cell has almost nothing to do with whether or not a game can run at 1080P at 60 frames per second. That's completely up to the RSX. ROFL, wow that's some ignorance for you... Please see the following post to help educate yourself: http://www.neogaf.com/forum/showpost.php?p=7598043&postcount=1 A good example is Super Star Dust HD which runs 1080p@60fps with tons of effects.

SPUs don't fill pixels.

We're talking about running at 1080p instead of 720p.  If you want to run at higher resolutions and fill more pixels, you need more fillrate.  That's all RSX.

SPUs can speed up a number of tasks, especially physics and vertex-related tasks.  They do not run pixel shaders.

@ Entroper

Here's an interesting read for you:

"This paper studies a deferred pixel shading algorithm implemented on a Cell-based computer entertainment system. The pixel shader runs on the Synergistic Processing Units (SPUs) of the Cell and works concurrently with the GPU to render images. The system's unified memory architecture allows the Cell and GPU to exchange data through shared textures. The SPUs use the Cell DMA list capability to gather irregular fine-grained fragments of texture data generated by the GPU. They return resultant shadow textures the same way. The shading computation ran at up to 85 Hz at HDTV 720p resolution on 5 SPUs and generated 30.72 gigaops of performance. This is comparable to the performance of the algorithm running on a state of the art high end GPU. These results indicate that a hybrid solution in which the Cell and GPU work together can produce higher performance than either device working alone."

"We have explored moving pixel shaders from the GPU to
the Cell/B.E. processor of the PLAYSTATION®3
computer entertainment system. Our initial results are
encouraging as they show it is feasible to attain scalable
speedup and high performance even for shaders with
irregular fine-grained data access patterns. Removing the
computation from the GPU effectively increases the frame
rate, or more likely, the geometric complexity of the models
that can be rendered in real time.

We can also conclude that the performance of the Cell/B.E.
is superior to a current state of the art high end GPU in that
we achieved comparable performance despite performance
limitations and despite using only part of the available
processing power. Our current implementation loses
substantial performance due to DMA waiting. This results
from the fine-grained irregular access to memory and is
specific to the type of shaders we have chosen to
implement. We have explored shaders based on shadow
mapping [15] which require evaluating GPU fragments
generated from multiple viewpoints. These multiple
viewpoints are related to each other by a linear viewing
transformation. Gathering the data from these multiple
viewpoints requires fine-grained irregular memory access.

This represents worst-case behavior for any memory
system. "

http://research.scea.com/ps3_deferred_shading.pdf

Problem is, if you allocate 5 SPUs in a PS3 to run pixel shaders, you won't have much left for anything else. That's a nice article coming from Sony, basically bosting the Cell (hey, all we researchers have to boost our stuff, even if we don't have a money stance in it, so why not Sony?)

It still makes a hell of a lot more sense to run logic (and most AI) on the PPE, geometry and physics on the Cell, and fragment stuff on the RSX than anything else. This is despite the fact that all of these (except maybe the GPU) universal machines capable of computing anything we throw at them.

@ KruzeS

I think you misunderstand the significance of this early research. The SPEs are able to do additional work in addition to the main GPU regarding pixel shading. You do not need to allocate 5 SPEs to handle pixel shading, as the PS3 has the RSX as well! Do you understand?

Anyhow the point was, the statement was wrong.

Who do you prefer to do the research, Microsoft or Nintendo?

Of course it's parties like Sony, IBM, universities and PS3 game engine developers who lead the way!

Looks like there will be a demo on the PSN:


"According to Kojima Products, the development team is working on an English version of the Metal Gear Solid 4 demo for American PS3 consoles.

The news is very exciting as Metal Gear Solid 4 was playable on the PS3 for the first time and seen by the public in this form for the first time at TGS 2007 not very long ago.

An announcement should be coming soon in regards to the demo coming to American PS3 consoles.

Kojima Productions Ryan Payton stated "As of tomorrow, we’re going to start working on the English demo."

Maybe this demo will come out in the same time as the Crisis Core one ... never :P

I think you misunderstand the significance of this early research. The SPEs are able to do additional work in addition to the main GPU regarding pixel shading. You do not need to allocate 5 SPEs to handle pixel shading, as the PS3 has the RSX as well! Do you understand?

Yes, I think I do. I obviously know that this is in addition to the work performed by the RSX. But the numbers the article you mentioned talked about (the 30Gops) seem to be attained reserving 5 SPUs for shader work. That's a little unrealistic in an application that does something else besides shader work.

Also I am unconvinced that moving a ton of data between the GPU, the CPU, and the SPEs local caches (which is quinda required for "cooperation") is going to be that fast for pixel work.

Obviously, And I didn't mean to imply otherwise. I only said that it's natural to bolster one's research results as if they mean something in real life conditions. I know I don't like it one bit, and am often forced to do it myself.

MikeB said: @ KruzeS I think you misunderstand the significance of this early research. The SPEs are able to do additional work in addition to the main GPU regarding pixel shading. You do not need to allocate 5 SPEs to handle pixel shading, as the PS3 has the RSX as well! Do you understand? Anyhow the point was, the statement was wrong. hey, all we researchers have to boost our stuff, even if we don't have a money stance in it, so why not Sony?) Who do you prefer to do the research, Microsoft or Nintendo? Of course it's parties like Sony, IBM, universities and PS3 game engine developers who lead the way!

I knew as soon as I posted that, that there would be some research paper where someone has experimented with running pixel shaders on the Cell.  So let me revise my statement.  The Cell can run pixel shaders, but it would probably not be wise to ask it to do so.

I read the white paper, and what's striking about it to me is that instead of comparing the full pixel shader on RSX to their hybrid implementation on PS3, they compared it to a GPU implementation on a PC.  It seems strange to me that when the purpose of your paper (and indeed, your conclusion) is to show that the Cell can speed up shader work on a PS3, that you would fail to show that the Cell can speed up shader work on a PS3!

Furthermore, the test has the CPU and GPU running nothing but the specific shader they have in mind.  The CPU isn't running a game engine!  It's not doing AI for hundreds of entities, it's not running physics, it's not doing anything except running this shader.  The GPU isn't being asked to render a game world with lots of other shaders involved, plus a few shadows, it's rendering one dandelion or one tree with monochromatic color.  Kruse pretty much already said this, in any real application, you're going to be allocating resources for more than one task, and consuming 5 SPEs and tons of memory bus bandwidth to pass all this data around is not going to be the best use of your resources.

@ Entroper

That's better, the first part being factual and the latter part being a personal opinion clarified well enough by using "probably".

However this was a rather early research test. The article and game developers don't seem to agree. Guerrilla Games already uses similar approaches with regard to their first PS3 game, Killzone 2.

Here's an interesting PDF detailing some benefits and approaches regarding using the SPEs for deffered rendering:

Deferred Rendering in Killzone 2

MikeB wrote:

Another interesting read, this time from Insomniac regarding their new igPhysics system.

Introducing SPU shaders:
http://www.insomniacgames.com/tech/articles/0907/files/spu_shaders_introduction.pdf

Benefits of SPU Shaders to igPhysics
● Pipeline well defined and completely SPU-driven
● SPU processing completely asynchronous
● Data well-organized and well-defined.
● No (or minimal) PPU intervention