-
Go Noles!!! >>-----> -
Re: Arstechnica: Inside the Xbox 360, Part II
Yikes, this isn't good to hear, for either the Xbox 360 and PS3. Games will look great, but it's going to take alot work to make them play great.
Pretty games and dumb AI once again. Hey, that gamers manifesto was right.
Rumors and some game developer comments (on the record and off the record) have Xenon's performance on branch-intensive game control, AI, and physics code as ranging from mediocre to downright bad. Xenon will be a streaming media monster, but the parts of the game engine that have to do with making the game fun to play (and not just pretty to look at) are probably going to suffer. Even if the PPE's branch prediction is significantly better than I think it is, the relatively meager 1MB L2 cache that the game control, AI, and physics code will have to share with procedural synthesis and other graphics code will ensure that programmers have a hard time getting good performance out of non-graphics parts of the game.
Furthermore, the Xenon may be capable of running six threads at once, but the three types of branch-intensive code listed above are simply not amenable to thread-level parallelization. Sure, you can have one single thread for each of the three tasks, but you can't generally split those tasks up further into even more threads.
On the other hand, these types of code do benefit greatly from out-of-order execution, which Xenon lacks completely, a decent amount of execution core width, which Xenon also lacks; branch prediction hardware, which Xenon is probably short on; and large caches, which Xenon is definitely short on. The end result is a recipe for a console that provides developers with a wealth of graphics resources but that asks them to do more with less on the non-graphical side of gaming.
At any rate, Playstation 3 fanboys shouldn't get all flush over the idea that the Xenon will struggle on non-graphics code. However bad off Xenon will be in that department, the PS3's Cell will probably be worse. The Cell has only one PPE to the Xenon's three, which means that developers will have to cram all their game control, AI, and physics code into at most two threads that are sharing a very narrow execution core with no instruction window. (Don't bother suggesting that the PS3 can use its SPEs for branch-intensive code, because the SPEs lack branch prediction entirely.) Furthermore, the PS3's L2 is only 512K, which is half the size of the Xenon's L2. So the PS3 doesn't get much help with branches in the cache department. In short, the PS3 may fare a bit worse than the Xenon on non-graphics code, but on the upside it will probably fare a bit better on graphics code because of the seven SPEs.
In sum, the Xenon will certainly make the Xbox 360 a 3D graphics powerhouse. Though history suggests that the Xbox 360's games will probably never attain the level of graphical realism promised by Microsoft's pre-launch hype and portrayed in the pre-rendered "game demos" that were shown off at E3 (e.g. the infamous Killzone "demo"), gamers can nonetheless expect a significant advance in levels of graphical realism and visual immersiveness.
Shows you how important development tools are going to be next gen.
In contrast, with the Xenon, the hardware will be fixed, which means that programmers can go all-out in profiling and optimizing branchy game control, AI, and physics code using every trick in the book. Furthermore, console coders can also take heavy advantage of prefetching to overcome the Xenon's cache size limitations. So it's quite possible that as time goes on developers will find ways to get much better branch-intensive code performance out of the hardware. Just don't count on it in the first generation of games, though.Last edited by Flawless; 06-01-2005, 11:16 PM.Go Noles!!! >>-----> -
Re: Arstechnica: Inside the Xbox 360, Part II
This really sounds like it sucksComment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Yes, the XCPU and Cell are dumbOriginally posted by KDRE
Pretty much all the rumblings we've been hearing, it's going to take alot work and money to tap these consoles.
These problems will eventually be worked around. Also sounds like Microsoft is going to be right - next-gen isn't about the hardware, it's going to be about the software.Last edited by Flawless; 06-01-2005, 11:36 PM.Go Noles!!! >>----->Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Wow man, I cant imagine the looks on the faces of sports gamers with the AI not matching the visuals on a new gen console.Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
That's disappointing to hear, but as long as AI takes a step forward from what's on the current generation, I'll be satisfied.XBL: Mean Greene
PSN: OGMeanGreene
Twitter: @greenegtComment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Definitely going to wait awhile before I get a 360.
Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
really interesting to read article...datapaths, instruction cache, crazy techno detail can be informative.
I'll still be buying a 360...the games will be fun and look cool, to say i'm not getting one would be going into denialNintendo Switch Friend Code: SW-7009-7102-8818Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
After reading this article again, it makes some J Allards comments hit home.
"Both of these machines are so sophisticated that theoretical performance doesn’t matter – what matters is how much of that performance can be unlocked. The key to unlocking performance of the hardware is software."
Hannibal mentions very little about software.
I'll wait for Anandtech's article on the next-gen systems.
I thought this was a fantastic post by BigAl on Team Xbox
Someone who has been working on an Xbox 360 gameOriginally Posted by BigAl
Well, I'm going to take a stab at explaining my criticisms with this Arstechnica series of articles again. Personally, I find that this two part series overemphasizes procedural synthesis and misrepresents the Xenon CPU.
The first article makes it sound almost as if Microsoft single-handedly created the notion of procedural synthesis or that they have otherwise somehow "cornered the market" on procedural synthesis techniques. Both ideas are gross mischaracterizations in my view. The overemphasis on procedural synthesis leads many console gamers to have visions of Xbox 360 graphics that are crappily drawn, unartistic, unnatural, cookie-cutter w/ minor randomization effects.... or to otherwise get people to have misplaced expectations. The fact is that procedural synthesis, in one form or another, has been a staple in the bag of tricks for the computer graphics industry for probably some 20-25 years, at least. The associated algorithms for artfully creating realistic depictions of natural landforms and foliage have been research topics for some individuals for most of their professional careers. The impression the article mistakenly leaves is that Microsoft is creating, in a span of a couple of years, a device that solely specializes in realtime procedural synthesis and that they are virtually forcing such a technique upon the gaming industry whether wanted or not. The article clearly implies that procedural synthesis is purely a cost driven measure that will replace skilled artists with the power of the Xenon CPU. None of these characterizations are true, IMHO. Procedural synthesis is simply another tool, that may be increasingly used as the CPUs in consoles become increasingly powerful enough to handle the fairly complicated procedural synthesis algorithms that do result in realistic and natural looking representations. Game developers will not be forced to use the technique or otherwise be compelled to use it by cost concerns in such a manner that results in a bunch of repetitive looking, crappy games.
In the second part of the article, the author leaves the reader with the impression that the Xenon CPU is just a graphical powerhouse, essentially designed for and basically only good for streaming media applications and geometry calculations. Reading the article, it's almost as if Microsoft perhaps shouldn't have even bothered working with ATI so diligently on a powerful GPU for such tasks for the past 3 years or so. Furthermore, if you bother to read the Chris Hecker rant that the author refers to, one might even clearly be left with the impression that the 3 core 3.2 GHz Xenon CPU is less suitable for physics, AI, and gameplay code than the single core 733 MHz Xbox CPU! All of these impressions are gross mischaracterizations and rather shamefully implied by the author, IMHO. There is no excuse for leaving gamers shaking their heads wondering why Microsoft could have made such blunders.
At the end of the day, upon delivery of final hardware and final compilers, toolsets, and libraries, the Xenon CPU will clearly be a far superior processor to the Intel x86-based Xbox CPU. To imply otherwise is simply irresponsible journalism. Even a single Xenon core, albeit constrained by IOE, will be much more powerful and more useful in tasks such as physics, AI, and gameplay code. Contrary to what the author would like you to believe, and unlike the Cell processor the Xenon CPU was not designed or optimized primarily for streaming media applications; and was, in fact, design with game control and AI code in mind as well. That is the most significant difference between the architectures, in fact. And it is one difference that the author glaringly glosses over. The layout of the SPEs in the Cell are not simply representative of a slightly modified rearrangement of the Xenon's vector units. To compare the two arrangements that directly is verging on the absurd, IMHO. I would expect that the programming and use of the SIMD units on either architecture will be fairly radically different from the programmer's perspective.
Lastly, there is a very clear difference between having 3 generalized cores versus 1 generalized core, or that SPE code is very distinct and different than PowerPC core thread code.The author almost waves his hand at that as but a minor difference... leaving one with the impression that the PowerPC based cores aren't very generalized at all. True, they aren't as generalized as a traditional x86 core, by any means, but they are far more generalized than an SPE! Nor do they have to be as flexibly designed as an x86 architecture given the fixed nature of the console design and the opimization of toolsets that can occur. Overall, I just believe that this article leaves many misrepresentations in gamer's minds.
I get the feeling your maybe not really understanding what in-order vs out-order execution is (if you do, my apologies...)
They execute the same code, so can do the same things. The only difference is that potentially out-of-order runs 'bad' code faster, as the CPU looks to see if there are instructions further ahead that it can do now.
Potentially you have to be a bit more careful in how you write the code in an in-order processor. Good in-order code is as fast and can do everything out-of-order does, is just a bit harder to write.
One way of looking at it, is that PC processor (OOOE) makes you a bit lazy, whereas to get maximum performance out of XeCPU or Cell you have to be a bit more careful if you want absolute maximum performance.
To be honest though AI takes so little total CPU time, nobody is going to care. If it takes 5% of a P4 or 8% of a single XeCPU core, who's really going to care when have 2 extra cores and a good GPU? Sure its 3% slower (thats just an random figure BTW) but you still have 2 other cores. Thats why we got in-order, they are smaller, so we could get extra cores.
Most console devs, are happy to have in-order with more cores. Its only really a few PC devs who complain loudly about the fact. Except Gamecube and Xbox (which both basically used desktop CPU) have been in-order, so we are used to it.
Hope that helps explain why its not a big problem IMHOLast edited by Flawless; 06-02-2005, 03:41 PM.Go Noles!!! >>----->Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Just to let you guys know, Hannibal was a little 'off' with his article and thus has ammended some of it after being pointed in the right direction by some of his membership. In fact, AI and physics will be more than ok in a heavily parallized architecture environment.
So don't fret people, the next-gen games(especially by 2nd and 3rd generation) will be far more visually appealing, and the AI + physics will be better than what you play today.
Here's how some of his article reads now:
Furthermore, the Xenon may be capable of running six threads at once, but the three types of branch-intensive code listed above are not as amenable to high levels of thread-level parallelization as graphics code. On the other hand, these types of code do benefit greatly from out-of-order execution, which Xenon lacks completely, a decent amount of execution core width, which Xenon also lacks; branch prediction hardware, which Xenon is probably short on; and large caches, which Xenon is definitely short on. The end result is a recipe for a console that provides developers with a wealth of graphics resources but that asks them to do more with less on the non-graphical side of gaming.
Still, there is some hope on that front. In the PC market where there are multiple processors to support, developers can't fine-tune games for a specific CPU. This heterogeneity of hardware especially hurts with platform-sensitive optimizations like branch hints, which is one reason they don't get used much. In contrast, with the Xenon, the hardware will be fixed, which means that programmers can go all-out in profiling and optimizing branchy game control, AI, and physics code using every trick in the book. Furthermore, console coders can also take heavy advantage of prefetching to overcome the Xenon's cache size limitations. So it's quite possible that as time goes on developers will find ways to get much better branch-intensive code performance out of the hardware. Just don't count on it in the first generation of games, though.
At any rate, Playstation 3 fanboys shouldn't get all flush over the idea that the Xenon will struggle on non-graphics code. However bad off Xenon will be in that department, the PS3's Cell will probably be worse. The Cell has only one PPE to the Xenon's three, which means that developers will have to cram all their game control, AI, and physics code into at most two threads that are sharing a very narrow execution core with no instruction window. (Don't bother suggesting that the PS3 can use its SPEs for branch-intensive code, because the SPEs lack branch prediction entirely.) Furthermore, the PS3's L2 is only 512K, which is half the size of the Xenon's L2. So the PS3 doesn't get much help with branches in the cache department. In short, the PS3 may fare a bit worse than the Xenon on non-graphics code, but on the upside it will probably fare a bit better on graphics code because of the seven SPEs.
In sum, the Xenon will certainly make the Xbox 360 a 3D graphics powerhouse. Though history suggests that the Xbox 360's games will probably never attain the level of graphical realism promised by Microsoft's pre-launch hype and portrayed in the pre-rendered "game demos" that were shown off at E3, gamers can nonetheless expect a significant advance in levels of graphical realism and visual immersiveness.Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Keep in mind that the Xbox 360's architecture has been designed hand-in-hand with Microsoft's XNA middleware, which is specifically designed to streamline the intricacies and variances of the 360's hardware.
Every dev I spoke to at E3 was very impressed with the tools Microsoft is providing to them; specifically XNA.
It's importance couldn't be overstated - especially when looking at next-gen units purely from an architectural perspective - we're through the looking glass now, and systems can't be properly or fairly evaluated on hardware alone.
Software (and middleware specifically) plays a larger role then ever, and the Xbox 360 can only properly be considered as an amalgamation of the two.Last edited by sdrotar; 06-02-2005, 06:50 PM.Shawn Drotar
Former Managing Editor, OperationSports.com (2002-07)Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
what I see when I read that article.
blah blah blahComment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Yeah. It's very "techie"...Originally posted by fishepa
Better get used to it, though - because after Xbox 360 releases, you can expect Sony to push their specs - in detail- all over the web for about six months. LOLShawn Drotar
Former Managing Editor, OperationSports.com (2002-07)Comment
-
Re: Arstechnica: Inside the Xbox 360, Part II
Show me the software!!! Show me the software!!! LOLOriginally posted by sdrotarComment
-
Re: Arstechnica: Inside the Xbox 360, Part II
i second thatOriginally posted by fishepaComment
Comment