AnandTech article up (X360 vs PS3)

There's very few universal constants in the world. Anand being a tool is one of thoses.

I mean...
The most important thing to keep in mind is that the revolution in physics engines and collision detection isn’t going to happen over night. The first games for both consoles will, for all intents and purposes, be single threaded titles.More adventurous developers may even split up execution into two concurrent threads

I'm being harsh, though, the article is not that bad...
 
Anand said:
The bottom line is that Sony would not foolishly spend over 75% of their CPU die budget on SPEs to use them for nothing more than fancy DSPs. Architecting a game engine around Cell and optimizing for SPE acceleration will take more effort than developing for the Xbox 360 or PC, but it can be done. The question then becomes, will developers do it?
PS2 developers who will develop for PS3 will.
 
Bit early to compare PS3 and x360, considering we really don't know much about RSX yet. Also, Anand himself doesn't have as firm a grasp over technical specifics as one would like to think considering he runs a hardware tech site, particulary when it's console-related. His comment about singlethreaded games is nothing short of laughable, assume instead this generation of consoles was the first with hardware accelerated 3D graphics, he might just as well have said that the first gen of games would use software rendering.

Multithreaded programming is the ONLY way to wring any kind of power out of these boxes. Of course programmers will take advantage of that, unless they're writing a tetris clone or such...
 
Tetris clone with 1080p resolution. :)

He also didn't tack on the fact that Sony didn't give any restrictions on resolutions either. We could very well be seeing 480i/p games.
 
I found the article to be one of the clearest articles i've read in a long time. Technical without being "over my head" complicated.

Very nice.

If they're wrong - and there are still a lot of unanswered questions about RSX - it's cool, but it was a good article.
 
hasanahmad said:
Xenos is not a 24 pipe card like the RSX. its 32. Mistake in the conclusion of article

Its neither.

But he would be comparing it to a 24 pipe card because of vertex shaders (i.e. 24 + 8 ).

But it is a fairly pointless comparison without knowing how effecient the design truly is in the real world. Either it is a design win and revolutionary or it is just a new, and not necessarily better, take on an age old problem. We do not know yet. But what we do know is that we do not have the info to compare it to a traditional # of pipes. Take a gander at Dave's article.
 
It seems silly that these companies are spending billions for these new cpus yet they can't outperform a p4 at alot of important processor tasks. There is something wrong there.

One question that I have is if the RSX replaces their vertex units with pixel ones how much redesigning of the gpu would that require being that pixel shader results are coming from a different source. Would developers have control on how many spes are used for geometry shading?
 
I didnt apreciated the fact that he treated the Xenos as a 24 pipe R420.....

There's alot of personal assumptions and opinions in the article, its cool to read, but it really is just an opinion based on "Ifs".
 
therealskywolf said:
There's alot of personal assumptions and opinions in the article, its cool to read, but it really is just an opinion based on "Ifs".

Thats what I thought, not much stubstance.
 
ralexand said:
It seems silly that these companies are spending billions for these new cpus yet they can't outperform a p4 at alot of important processor tasks. There is something wrong there.

What's wrong is that people think consoles should "beat P4's" in Word and Excel.

I'd LOVE to see a P4 powered PC "beating" the next gen consoles in what the consoles are supposed to do: GAMES.
 
london-boy said:
ralexand said:
It seems silly that these companies are spending billions for these new cpus yet they can't outperform a p4 at alot of important processor tasks. There is something wrong there.

What's wrong is that people think consoles should "beat P4's" in Word and Excel.

I'd LOVE to see a P4 powered PC "beating" the next gen consoles in what the consoles are supposed to do: GAMES.
But that's the thing. I'm hearing developers are complaining about game performance.
 
ralexand said:
london-boy said:
ralexand said:
It seems silly that these companies are spending billions for these new cpus yet they can't outperform a p4 at alot of important processor tasks. There is something wrong there.

What's wrong is that people think consoles should "beat P4's" in Word and Excel.

I'd LOVE to see a P4 powered PC "beating" the next gen consoles in what the consoles are supposed to do: GAMES.
But that's the thing. I'm hearing developers are complaining about game performance.

Look, games on next gen consoles will run VERY well. I don't think you should be worried.
Developers just need to start taking advantage of the multithreaded nature of consoles. They're gonna have to, since PCs are following suite too.
 
I find it kinda funny that this gen there's probably going to be a very distinct split in perspectives for programming for these consoles.

The devs that learned to wring performance from the PS2 are going to switch and talk about how happy they are that next gen is so easy to program for while the PC devs and Xbox devs are going to scream bloody murder.

I wonder if being PS2 centered this gen might have some big gains next gen around? Yeah, I made a joke about easy, but there really should be a sort of seperation, no?
 
ralexand said:
But that's the thing. I'm hearing developers are complaining about game performance.
I imagine they need to learn to write for the new hardware then. Next-gen needs new approaches and mindsets. Anyone coming at it with a conventional approach will hit aggravating bottlenecks. A huge amount of coding is writing to work with the hardware, not just to implement a routine that makes sense to the coder. I've written stuff that runs at a crawl until I've ditched my humanistic approach and deal with the hardware. Likewise next-gen running current gen software designs is gonna come a cropper.

They may all moaning now, but in two years times when they've adjusted all the devs will be cheering merrily about how great the next-gen systems are!
 
The article had holes. I love Anandtech and read all their featured write-ups, but just like their write-up on Cell fell short of what David Wang did at RealWorld, so to did this article just not bring it all together.

I think they give Cell short-thrift, but appreciate that they clearly try to remain as unbiased as possible. They're not going out of their way though to understand ways in which developers might program on the SPE's, however, and seem to have been satisfied with words from Sweeney - in fact, surprised that he was even using them as much as he was.

Also with the RSX transistor count, they didn't even entertain the possibility of PureVideo being stripped out but rather, focused entirely on the added logic tweaked TurboCache would require.

And... What does a blu-ray player launching at $500 have to do with anything? He can't be assuming that the blu-ray ROM drive will cost $400 or more to include - I mean c'mon.
 
one said:
Anand said:
The bottom line is that Sony would not foolishly spend over 75% of their CPU die budget on SPEs to use them for nothing more than fancy DSPs. Architecting a game engine around Cell and optimizing for SPE acceleration will take more effort than developing for the Xbox 360 or PC, but it can be done. The question then becomes, will developers do it?
PS2 developers who will develop for PS3 will.
LOL. Yes, as far as they are concerned, the SPE's are just 5 extra "vector units" to ignore. Anand seems to forget that. I don't think there's one PS2 game that uses both VU's to their full potential.
 
The article brings up something I've been wondering about myself for a long time - impact of the shared L2 in the XCPU.

Every SMP box's CPU has their own L2. The tradeoff is that it always requires sophisticated designs to keep them in sync. I look forward to seeing the effect of a shared L2 when devs had spent more time with it. At best there is insignificant effect on a gaming application. At worst it prefers fine-grained parallelism and shuns other implementations.
 
Alpha_Spartan said:
LOL. Yes, as far as they are concerned, the SPE's are just 5 extra "vector units" to ignore. Anand seems to forget that. I don't think there's one PS2 game that uses both VU's to their full potential.


Depends what you mean for "full potential".

I can definately tell you that they are used, and some games use them very very well. The VU0 could be used more, but the VU1 has been used very well since the beginning, or else you'd see no polygons on PS2 games.

In the end the SPEs are not VUs.
 
Back
Top