If you haven't already seen it, here's my coverage of Sony's Playstation 3 announcement today. I wrote the story while sitting in Sony's press conference, so it was a bit rushed but I wanted to post some of my additional thoughts that didn't make it into the first article.

Let me start first with the design; to me, the Xbox 360 is very Apple-like while the PS3 is very clearly a Sony product. Personally I prefer the looks of the Xbox 360, but the PS3 doesn't look bad at all in real life.

Although I've yet to use it, the PS3's controller scares me. I'm going to try my hands at it this week, but I really have no idea where that design came from.

The demos on the PS3 were absolutely *amazing*. I wouldn't call them "movie-quality" yet, but the things I saw came very close. Words really can't describe, the demos just looked amazing.

Virtually all of the games/demos on the PS3 had some degree of aliasing, some were unacceptably bad for a console with this sort of power. Don't get me wrong, about 95% of the games looked great, but those that had aliasing looked great...with jaggies. I'm not talking PS2 level of aliasing, but far too much aliasing for this level of hardware.

Without a doubt, ATI and NVIDIA are on very diverging paths with these two consoles. ATI went with a strictly unified memory architecture while NVIDIA used a combination of local graphics memory and GPU addressable system memory. ATI is backing their unified shader architecture, while NVIDIA doesn't appear to have embraced that on the hardware side. I will know more about ATI's GPU later this week, so stay tuned.

The dual HD output feature of the PS3 is very interesting; I'm not sure how many folks will take advantage of the 32:9 aspect ratio mode. I'm wondering whether this feature was put in to support sending different content to separate TVs (e.g. stream video to one display while gaming in another). Then again, I'm not sure how many people have that many HDTVs within close proximity of each other.

Sony clearly wants the PS3 to be much more of a media center style device. The demos weren't only about games, they were about decoding HD streams, navigating through video and picture content, they were about the entire picture. With built in blu-ray, I think the PS3 will have a huge advantage over the Xbox 360 as it should be able to act as a HD-DVD video player as well as a game console.

The 1080p output of the PS3 isn't that big of a deal for me. Given that basically the entire installed base of HDTVs right now only support 1080i, I seriously doubt we'll see a push to 1080p only all that quickly. That being said, I don't doubt that there will be an obvious difference between 1080p and 720p games. Given that it is essentially a resolution change, I see no reason for all developers to offer both 1080p and 720p options in PS3 games unless there are frame rate limitations. I did notice that some demos played much smoother than others, but I think it is far too early to make any calls on performance a full year before the console's release.

I'd say that Sony has the more powerful CPU on paper, but I'm curious to see how much of that gets taken advantage of in the real world. Difficulty of programming aside, the fact of the matter is that console development houses are very much of the write once, compile many mindset. Given the similarity of the Xbox 360's cores to the PS3's PPE, I'm afraid that the array of SPEs may go relatively untapped on the PS3.

From the very start I felt that Sony couldn't possibly bring the Cell to market in the PS3 as a 90nm chip. Disabling one SPE is a particularly interesting move, but one that makes a lot of sense. And the loss of a single SPE isn't a huge deal as I don't foresee the PS3 really being bound by the number of threads its SPE array can execute.

Overall, the PS3 looks to me to be the more complete package. The hardware is a bit more complete than Xbox 360, but at the same time given that it won't launch for another 6+ months after the 360 launches I'm not too surprised. Sony didn't really play up a competitor to Xbox Live, although it is very clear that the PS3 will be a net-enabled box. I have a feeling that Microsoft may bring to the table a much more complete on-line play package, while Sony brings a more powerful, more complete console.

Sony's strength with the PS2 has always been its game library, which I think will continue to be a strength with the PS3 (especially with full backwards compatibility all the way back to PS1). It's just that this time around, Microsoft appears to have a much stronger game library than with the original Xbox - and it's that key difference that will make the 360 and the PS3 worthy competitors.

I will be reporting from the show all week, but for now it's time to enjoy 24 a full 3 hours later than I normally would - how do you west coast folks do it? :)

Take care.
Comments Locked

125 Comments

View All Comments

  • Z - Tuesday, May 17, 2005 - link

    #27 problem is that with XBox360 you *have to* use multithreading in your code, which is complex. Time slicing isn't efficient performance-wise, with preemption each thread has to fight others for CPU time, I much prefer if they would cooperate.

    If you have enough CPUs, each thread can reside on its own CPU, so no preemption, and no complex multithreading either.

    The lack of branch prediction isn't really a minus, because we're comparing to an alternative that hasn't got branch prediction either, not to an Athlon or P4.
  • k - Tuesday, May 17, 2005 - link

    PS3 has only 256MB of RAM, while XBox has 512. Looking at PC games today I'd say both of these values are too small to handle a modern game. But XBox has an important edge here.
  • Jud - Tuesday, May 17, 2005 - link

    Another 24 fan. :)

    So (before the big 2-hour finale) what d'you think of this season so far?
  • David Smith - Tuesday, May 17, 2005 - link

    I'm not sure why everyone seems to think that the Xbox 360 has something like SPEs. It has 3 Altivec units, yes... and it looks like they're a variant of Altivec with more registers, but nowhere have I seen anything pointing to them being separate cores like the SPE. The PS3's PPE has Altivec as well, btw.

    As for which one will perform better... I think we're going to have to wait and see. Either it'll be possible to split games into lots of specialized subtasks effectively (and the PS3 will win) or it won't (and the Xbox will win), or it'll be somewhere in between.
  • Anonymous - Tuesday, May 17, 2005 - link

    #29 even.
  • OrphanBoy - Tuesday, May 17, 2005 - link

    #28, touche. What *did* happen to the Saturn anyway? :) Didn't they have funny game things in the controllers themselves?

    Software that's written for the Cell still has to be multithreaded, and the same synchronisation problems will still apply. Instead of having a bunch of homogeneous processors (like the Xbox360) there's going a bit more variety :) But I guess it does follow the traditional way...

    But either way: man, those poor programmers!
  • ecwx - Tuesday, May 17, 2005 - link

    #28 : Remember the Saturn. The saturn, given its parallel processor,is theoretically more powerful than PS1. But programming for efficiency in parallel processing is much more complex than on a single processing machine. MS, with 360, seems to choose the saturn way and Sony, with PS3, seems to choose the traditional way.

    Taking advantage of more than one multi-purpose processing units means that the software (game) must be designed to run multi-threaded, and as a real time program, the developer must take care of the execution sync between the threads to make sure that the processing pipeline is balanced.

    Taking advantage of multiple specialized processing units is already common practice today.
  • OrphanBoy - Tuesday, May 17, 2005 - link

    #27 true, you definately can't beat specialised silicon for these tasks - that's what DSPs are for!

    But then again, from a programming point of view I think it'd be much easier to give the tasks you've listed to a few general-purpose CPUs than more DSP-like processors - that's what scheduling's all about! Once a task running on a CPU has had it's time slice, the next task just gets pre-empted...

    It'll be harder to program the SPEs than regular CPUs as they lack certain important features, like branch prediction. And the main core on the Cell doesn't support out-of-order execution so it's going to take fancy programming/fancy compilers to get extract the full performance. Anyone who's ever programmed pixel shaders will know what I'm talking about here!

    As you say both MS and Sony will have to make the multi-threaded nature of their hardware easy to access within their dev tools!
  • Z - Tuesday, May 17, 2005 - link

    #21 Indeed!
    There is only so much that is only efficiently performed on a general purpose CPU, and as far as games or video are concerned, a lot of tasks, from AI to physics only require brute force, and brute force is what specialized hardware excels at, practically, only scripting and some overall game logic really require a general purpose CPU, none of these too being performance critical.

    An issue to have in mind is that keeping busy 3 general purpose CPUs won't be easy to do (programming-wise). One might think it would be worse with 7, but there is a difference: if you have a physics libraries that keeps busy 2 CPUs, an AI one that keeps busy 3 CPUs, a sound/voice one that take 1 CPU, etc. they will be much easier to reuse: they'll just take a "slot" (aka CPU), so you can assign each specialized library to a specialized CPU and be done with it.
    On XBox 360, funnily enough, there aren't enough CPUs to have them dedicated to a specific task for a particular library, so libraries will have to share CPUs power (which means trouble, complexity and potential conflicts), while on PS3, you could go the dedicated route.

    Real question becomes how MS and Sony will play that multi-CPU game in their development tools, MS may have more practice in that realm, but Sony has the best hardware potential.
  • OrphanBoy - Tuesday, May 17, 2005 - link

    #24, no he's right! What's the point in sending half-frames to a display that doesn't work in a naturally-interlaced format?

    And I can't resist this - I'm starting work at Sony UK in August and I'm pretty sure I'm gonna be able to play with one of these things months before you guys :)

Log in

Don't have an account? Sign up now