Interview w/ Matt Lee, MS ATG

[Rockster]

Thought some might find this interesting:

http://arstechnica.com/articles/culture/mattlee.ars

 

[sonyps35]

Lets just get right to the juicy part shall we?

What do you think of the PS3's Cell architecture? How much harder is it for developers to write asymmetric multiprocessing code that will run on the PPE and the SPE's? How difficult will it be to port code that is tuned for the Cell over to the 360 while maintaining good performance? Do you think Sony made the architecture complex deliberately to make porting that way harder, or were they just trying to get maximum performance out of their system?

I don't think the Cell is as well designed for game development as Sony would have you believe. Some aspects of the SPEs, such as the lack of branch prediction, make them particularly unsuited to running most game code, which contains a lot of branches. They appear to be designed more for serialized streaming math code, more common in video codecs and audio processing, the traditional domain of digital signal processing chips. The memory architecture of the SPEs, specifically their lack of automatic cache coherency in favor of DMA transactions, seems like a lot of overhead is needed to feed work units to the SPEs and copy the results back to system memory.

The PPE appears to be essentially identical to one of the Xbox 360 cores, except without the VMX128 enhancements and with half the cache. However, a much greater assortment of work has to be crammed into this single core—all of the game loop, all of the rendering commands, and the system allegedly takes over some time as well. Only the second and third cores on Xbox 360 use a small timeslice to provide cool stuff like the Guide, music playback, Dolby Digital encoding, and more things that we can add in the future to all games, past and present.

I think porting from Xbox 360 to PS3 will be reasonably difficult, since the Xbox 360 has a lot more general purpose processing power that can be flexibly reallocated, and all of the Xbox 360 CPU cores have equal access to all memory. The asymmetric nature of the Cell could easily lead to situations where the game has too little of one type of processing power and too much of another. And the content might suffer as well, since you'll never see a PS3 title with more than 256MB of textures at any given time, due to the split graphics and system memory banks. When we announced 512MB of unified memory on Xbox 360, I think all of our game developers (and the artists too) did a little happy dance. It's easier to use and gives developers much more flexibility in how they allocate memory for various resources.

In terms of performance, I think that the PS3 and the Xbox 360 will essentially be a wash. We ran the numbers a while back and the two systems come up surprisingly close in theoretical peak performance, despite the one year difference in release dates. However, I know for a fact that we have a great advantage in software and services—our development environment and tools are years ahead of the competition, and this will ensure that Xbox 360 game developers can easily realize all of this performance and make superior games. Xbox 360 is a great system to develop on, a real pleasure—and I believe our developers agree.

 

[rabidrabbit]

What I find funny in these MS "PS3 slashing" PR, is that while at first they prove the xbox360 CPU (and the Xenos) are that much better, in the end they always fall back to "it's a wash" and "the systems are equal in performance".

If the CPU and GPU are that much better, shouldn't they say "the xbox360 games will look and perform better than PS3 games".

Confusing.

 

[Tap In]

Matt Lee:

I have worked at Microsoft full-time for almost five years now. Before that, I was an intern at Microsoft for two summers, while I was at MIT. After I finished my master's degree in June 2001, I started in Microsoft Game Studios...

...I'm an engineer in the Game Technology Group. My job is to help game developers make better Xbox 360 games. I'm one of about a dozen people who share a similar role here, and each of us have certain specialties. Mine are graphics programming and content pipelines, but we have CPU experts, networking experts, audio experts, and more—everyone overlaps to some degree. We also have a content team that reviews artwork and gameplay in the same way I review code and performance.

this guy is no PR person

it is not confusing. He's not saying Cell is 'bad', Xenon good, as much as perhaps just saying that Cell is not the end all, be all processor for games that Sony may want people to believe, therefore making it more on par with the Xenon (in his eyes) rather than way ahead of it.

 

[Acert93]

this guy is no PR person

But he is clearly doing PR work. It may be of the semi-technical sort for a technical audiance, but it is still PR. Which of course being a MS employee and being interviewed due to his position in the company, one would expect. They are not talking to Matt, the free independant geek who can give his free opinions on anything and everything, but with Matt, the MS employee who is paid to present the companies products, vision, and in sticky situations is expected to tow the party line.

 

[Tap In]

But he is clearly doing PR work. It may be of the semi-technical sort for a technical audiance, but it is still PR. Which of course being a MS employee and being interviewed due to his position in the company, one would expect. They are not talking to Matt, the free independant geek who can give his free opinions on anything and everything, but with Matt, the MS employee who is paid to present the companies products, vision, and in sticky situations is expected to tow the party line.

point taken


but ... at least he knows what he's talking about and works with devs on how to get the best from the hardware and to do what we talk about here everyday. I was pointing that out for those who may choose not to read the entire article.

 

[Tap In]

Multithreading sounds like it is a real work in progress and that they are really just feeling their way through it right now to find what works best.

We collect as much multithreading knowledge as possible from all areas of computer science in practice and research, and strive to apply that knowledge to games. The best way to multithread a game has yet to be conclusively decided, and I doubt it ever will be, due to the wide variety of threading needs across different game genres. We continuously work on a set of best practices and provide them to our developers. Tools are equally important. Our primary performance tool, PIX, is the best performance tool in the industry. Part of this tool is a CPU timing capture, which graphically shows various blocks of code on timelines for the six hardware threads. This helps game developers visualize the threaded execution of their code and things like synchronization become much easier to understand. We have other tools that measure stalls and usage of synchronization primitives, which alert developers to problems in their threading architectures. However, the work is not finished here; we have an extensive list of performance tool improvements planned that will make multithreaded development and debugging much, much easier.

Also regarding the tessellator in Xenos and procedural synthesis...

The tessellator in the Xbox 360 GPU is indeed a very powerful piece of hardware, and you're right—most games have yet to take advantage of this. I think you'll see more titles use it in the future. As for procedurally generated worlds, I believe the biggest obstacle to overcome is how to design and build the content for such a system—it can be quite a departure from today's art pipelines. Game studios will figure it out though—it's crucial to generating and delivering ever larger worlds without having to exponentially grow the size of the art team.
Right now, the best example of using procedurally-generated worlds is a third-party bit of middleware "SpeedTree," which was used to generate all the trees in Oblivion, although this was not done on the fly. This cuts down a lot of work for the artists, who then don't have to spend days making dozens and dozens of slightly different tree models. The idea of using on-the-fly procedural synthesis is one that even Microsoft hasn't explored yet.

 

[scooby_dooby]

lol, this guy ported Dope Wars over to Palm I'd forgotten all about this game, used to play it all the time on the old graph calculator
http://mytreo.net/downloads/details-152.html?DopeWars

 

[Bad_Boy]

So basically in a few years we will be seeing games on the ps3 a lot more impressive than the launch titles due to the learning curve. Which in a way, happened to the ps2 as well. Not unexpected, but I really dont see better looking games in the future as a huge negative as long as the games in the beginning still look great.

*note, not saying the 360 games will not get gradually better.

 

[PiNkY]

What I find funny in these MS "PS3 slashing" PR, is that while at first they prove the xbox360 CPU (and the Xenos) are that much better, in the end they always fall back to "it's a wash" and "the systems are equal in performance".

If the CPU and GPU are that much better, shouldn't they say "the xbox360 games will look and perform better than PS3 games".

Confusing.

Hmm, where did he say this. He points out that the SPE's are unsuited to branchy code and goes on about issues of asymetric designs in general and porting between the two plattforms (, which IMHO from what I've seen are pretty equal in performance anyways). Actually I thought that was a pretty good interview. However talking to their graphics expert and not asking a single GPU-related question made me kinda sad..

 

[Tap In]

So basically in a few years we will be seeing games on the ps3 a lot more impressive than the launch titles due to the learning curve. Which in a way, happened to the ps2 as well. Not unexpected, but I really dont see better looking games in the future as a huge negative as long as the games in the beginning still look great.

*note, not saying the 360 will not get gradually better.

what article did you read?

there were those two paragraphs when he answered the question of cell and the other 4 pages were on X360 tools and hardware. :smile:

 

[Nesh]

Why in the world are they asking a competitor to comment on a competitors product than a developer or someone who is independent from them?

 

[Acert93]

So basically in a few years we will be seeing games on the ps3 a lot more impressive than the launch titles due to the learning curve. Which in a way, happened to the ps2 as well. Not unexpected, but I really dont see better looking games in the future as a huge negative as long as the games in the beginning still look great.

*note, not saying the 360 games will not get gradually better.

I have a post that has been sitting in Word for months about, "Why games will begin to look substantially better 16-24 months from now". If I were not so busy writing my other epics I would probably get around to posting it.

But you are right, both consoles have a lot of upside this time around. e.g. SM3.0 has barely been touched on the PC side of things. Besides being faster, it opens up some new techniques to make things look better. Both consoles also have fast connections from the CPU to the GPU through a number of methods and both have a lot of memory as the "baseline". They also have a lot of power that needs to be tamed, but onced domesticated could be substantial workhorses. It would seem that good tools is what is needed to get us there though... at least 95% of the dev world. The Squares and Epics of the world have a lot of technical savvy and "just get it done" but as a gamer I want to see the smaller guys do well too.

 

[Tap In]

I have a post that has been sitting in Word for months about, "Why games will begin to look substantially better 16-24 months from now". If I were not so busy writing my other epics I would probably get around to posting it.

But you are right, both consoles have a lot of upside this time around. e.g. SM3.0 has barely been touched on the PC side of things. Besides being faster, it opens up some new techniques to make things look better. Both consoles also have fast connections from the CPU to the GPU through a number of methods and both have a lot of memory as the "baseline". They also have a lot of power that needs to be tamed, but onced domesticated could be substantial workhorses. It would seem that good tools is what is needed to get us there though... at least 95% of the dev world. The Squares and Epics of the world have a lot of technical savvy and "just get it done" but as a gamer I want to see the smaller guys do well too.

Man when you say you are busy, you are really busy.

oh and you need to post that thread.

 

[Acert93]

Actually I thought that was a pretty good interview. However talking to their graphics expert and not asking a single GPU-related question made me kinda sad..

Hey, he asked about tesselation!

But yeah, after Xenon I was like, "Finally, some dirt on Xenos!" Meh!! I am kind of surprised they only have like 12 people on this stuff. I could be wrong headed about this, but having a 50+ member team working on code, both samples for the SDK, helping developers get the most out of the console, highlighting how features can be used together, improving tools even more and APIs, etc... would seem like a great investment.

Calling Dave Baumman: Get a followup with Matt! Email him with the angle, "I noticed in the Ars interview you got few graphic questions. As a 3D enthusiest site we were all dissappointed. Would you mind doing a graphics-oriented interview with B3D?"

 

[kimg]

I didnt anticipate xbox360 where a branching monster from seeing pre launch specs and papers, is i my understanding way out of whack? I assume 360 is terrible at branch prediction, compared to my AMD athlon?

Im agreeing in PS3 seem to need more devlopment to get harder at start, ive never seen anything so weird as the assembly language to the SPU's to be honest, im sort of beliving K Kutaragi got right in PS3 needs a new generation of thought to get it really going.
I dont belive its wasted, as this article sugest, that Cell's multicore capacity cannot make it better than xbox360's 3 PC like cores.

 

[assen]

I didnt anticipate xbox360 where a branching monster from seeing pre launch specs and papers, is i my understanding way out of whack? I assume 360 is terrible at branch prediction, compared to my AMD athlon?

Don't bring Athlons in the discussion. EVERYTHING is terrible at branch predition compared to an Athlon.

 

[rabidrabbit]

I have a post that has been sitting in Word for months about, "Why games will begin to look substantially better 16-24 months from now". If I were not so busy writing my other epics I would probably get around to posting it.

So that's the secret of your long, elaborate posts!!
I had thought they were realtime, but now it's revealed they are all prerendered!!!!

 

[ROG27]

Don't bring Athlons in the discussion. EVERYTHING is terrible at branch predition compared to an Athlon.

Both CPUs are terrible at branch prediction...it just happens CELL happens to be a little worse off than Xenon.

 

[Tap In]

So that's the secret of your long, elaborate posts!!
I had thought they were realtime, but now it's revealed they are all prerendered!!!!

perfect