XBox 360 vs. PS3 thread

[Chris D]

I would be interested if somebody could summarize a reasonably impartial comparison between the Xbox 360 and the PS3. Based on the likely information that we know about the PS3 so far, which one is better in various categories? (i.e. graphics, processing power, video resolution, I/O technology like HDMI, audio, games, pricing, etc)

I've tried to stay away from game systems over the years, only because I already spend too much time and money on hobbies. I don't know if I can resist this next generation of gaming systems, though, and would like to buy the best. I don't want to buy an XBox 360 right now and then find out in 6 months that it's blown away by a PS3.

I can't find any good comparisons on the net, except for maybe diggin in WAY deep into gaming forums to extract a bit of info in all the talk from all the gamers out there.

Anybody able to make a good summary?

[Daryl Furkalo]

Impartial? Good luck. All that is concretely known about the PS3 is that it comes out in 2006. I don't think any of the demo footage shown from it currently is in game play so any comments on the abilities of the nvidia graphics or cell processor are speculation and I won't go there.

As an Xbox owner, on Live, and with an HDTV, the decision to buy a 360 was a no-brainer and so far mine is functioning flawlessly.

The decision on which system to get should be based on what games each has exclusively and whether you want to game online. If you don't have an HD set yet, wait. X360 has Live, I have no idea what PS3 will have.

[Lex]

About the only thing you could realistically do Chris is wait on the introduction of PS3, give it 6 months, then make a decision based on what is in the marketplace. PS2 IMO was a better platform from a software perspective. Sony I think had the upper hand over XBox, as so much was new for XBox. That's not to say Microsoft did not develop potential of the XBox, they did. But in some ways, it seemed like so much was territory the PSII could already handle, and did better. They already had the software contacts, and games developed from PS to build on.

I don't know, I played PSII and XBox, and I think I enjoyed some games from PSII better. I always enjoyed Wipe Out for one.

[Azeke]

Check out this thread from gamespot.

XBox 360 vs. Playstation 3

Regards,

Azeke

[audioqueso]

I'm gonna end up getting a PS3 solely because I want to play Final Fantasy. I don't really care which one is better or more powerful. However, I didn't know much about the Xbox360, so I started looking it up, and it's actually a really nice unit. I mean, for $400 you get a full gaming system, a computer, and HTPC, full wireless, and who knows what else. I think it's really nice for the price. But I won't bother getting one.

[Shane Martin]

I'm gonna end up getting a PS3 solely because I want to play Final Fantasy

You do know that Square and Microsoft have a deal don't you?

Tough to say right now which system is superior until both are out. Right now PS3's are vaporware until we see the real deal.

[Chris D]

I still don't know if I will end up buying either set.

Since I haven't owned a gaming system since the early 80's Atari whatever, (8200?) I have no knowledge of one manufacturer's games vs. the other's.

As an audio/video snob, though, I just can't understand how the XBox 360 doesn't include DVI/HDMI and resolutions above 1080i. So just in those specs, the PS3 seems to me to be the better option--add in high-def Blu-Ray and I'm leaning in that direction.

[PewterTA]

Xbox360 (according to Microsoft) will have an HDMI connection coming out later on, just didn't have it ready in time I guess. Supposedly they are working on another disc format for the Xbox360 HD or Blue-Ray, they didn't say...that they will work into the system come middle of next year. I don't know if this will be an internal thing in the next xbox360s and external for current, or whether it will just be an external device.

Basically you have to look at it this way, Xbox 360 will become the what PS2 is today, out first to capture the market and later down the line show it's weaknesses sooner. The PS3 will be like the current xbox where it's technically better than the PS2...

As far as games/software support, I think it will be pretty even, since the Xbox 360 will be out (for sake of arguement) almost a year before the PS3, the PS3 will have some ground to make up. However, software is Sony's stronge point. As long as the PS3 is decently easy to program for (which I hear the Xbox360 has a very short learning curve) and with it supporting Blue-Ray, I think that will end up pushing ahead.

The couple mistakes I think MS made....

One, not enough systems... I don't mean for there to be hundreds in every store, but at least 3-5x as many as there were at launch would've been perfect. Create a demand, but not so there's only 12 - 50 at a single store. The should have gotten more deveolpment kits to the third parties to help create more/better games. Not saying the games are bad, but not having to rush the stuff would've been better. More testing would have helped them find a lot of the "reported" problems with the systems that people are having. Also make more (out of the box) xbox games compatible with the xbox 360.

I think all their problems would've been resolved by delaying launch a month or so. Not a big rush considering PS3 was pushed back even further.

Probably when it all comes down to it, it'll be smart to buy the xbox 360 in about 4 or 5 months, then when PS3 comes out, give that 4 or 5 months and get one. Then you don't have to worry!

I do really like the Xbox360 controller, that thing is really nice! Can't wait to see (feel) what the PS3 controller will be like.

[Christian M]

I just cant help but think that MS may have shorted their game by not going with next-gen media. Just a thought but the sheer content they could have fit onto HD-DVD or BD media would have been rad (hi-def stuff aside).

With that, right now I own both and I like them for the games each offers. Gran Turismo for ps2 and halo for xBox... For me those are the games that "make" me buy each system. Since the past couple of GT's have been good, I just hope MS/Bungie dont screw up halo again. Halo2 was good but if i'm not playing on LIVE, I hardly play it... What map is better than Hang Em High, crazy king and rockets? Huh?!? Yeah... :-y

- Christian

[xixel]

2 words

gran turismo

thank you that is all carry on

[Lex]

also 2 words, "I'm broke". lol.

[wildfire99]

I ended up purchasing ALL the consoles from the last generation (GameCube, XBox, PS2) used, for the same price as one of them brand new. I also ended up buying games used. Pretty sweet deal, really.

That said, I eventually started to groan whenever a title I wanted wasn't on XBox. It was just the only console that had, to me, decent graphics. GameCube was nice too, but you'd never catch something like Halo on it. I play more PS1 games on the PS2 than vice-versa. Even newer PS2 games just don't look right, with blurriness and poor 16:9 support. It's too bad really, I agree with xixel... Gran Turismo on PS2 was hella fun in 16:9, as long as you didn't play it on a big screen (big screen = 10ft wide). XBox + big screen =

I'm not concerned about the cost of the consoles, I can pick them all up from someone burned out on the gaming scene for less than the cost of a good computer video card (I won't get started on how much better PC gaming is). The cost of games though is pretty out of control. Fortunately Wal-Mart has lots of fun $10 games from a couple years ago. Games that are more fun than tech demos *cough**doom3**cough*.

[Chris D]

Well, I decided to pass on the X-Box 360. Not set for sure that I'll buy a PS3, but more a decision not to buy the X360. Doing just a bit of research, the PS3 does look to have the upper hand--double the processing power, etc. (in theory)

We'll have to see what happens in 4-5 months.

[Kyle]

From what I gather, the PS3 is getting some crap from the media, rumor has it a lot of the hardware stuff is still in limbo, and the SDK is very buggy and hard to develop for.

I read that most of the first gen xbox 360 games aren't really all that multi-threaded because developers are having trouble taking advantage of the multiple cores, but like with the PS1 or even the Xbox, the first couple generations of games don't come close to eeking all of the performance out of the hardware.

Some PS3 rumors from Gizmodo yesterday: http://www.gizmodo.com/gadgets/home-...kes-145916.php

[cinema bob]

in my experience if you're not into first person shooters then get the sony. sure there are some good sports titles and racing games for the xbox but the breath and scope of non FPS games for the sony is mind blowing. and all of the games from the PS/1 and PS2 will be compatible with the PS3 for me that means that at launch I will own over 100 games that work with the PS3. Maybe when i can get rugby, cricket and gallop racer for the xbox then i will be interested till then I buy sony. (and gamecube used, gotta have zelda and metroid and pikman)

[derekbannatyne]

I'm getting the PS3 for free, so I think I'll stick with that :B

[Chris D]

Uhhhhh... okay. Well, obviously, post details, man.

[derekbannatyne]

Uhhhhh... okay. Well, obviously, post details, man.

It's one of those freebie things... they actually work but take a lot of work. The only reason I signed up for it is because I joined a line at www.freebieauthority.com . You can try it if you have 8 friends that will be willing to complete an offer...it's through a company called Freepay.

[JKalman]

The Xbox is definitely easier to program/code for than the PS2, and likewise with the 360 and the PS3 the Xbox360 is easier to program for. This is because it is basically the same as programming for the windows environment. It has a lot of the same abstractions, it has the same languages, etc. The Sony systems usually end up being highly proprietary meaning the programmers have to spend months just learning how the hell to start programming the hardware from the ground floor. The 360 builds on previous, familiar, paradigms that most programmers are already familiar with. Though, on a quick inspection of new updates to the wikipedia entry for the PS3 it looks like Sony learned their lesson from the PS2's initial programming fiasco and is working with open source and other technologies us programmers are very familiar with to make developing for the PS3 simpler from the start by providing essential development tools comparable in simplicity, hopefully, to the ones microsoft provided with the 360.

As far as hardware, the PS3 should be able to render in real time. That is something the xbox360 is on the verge of doing but won't be able to do the last time I was researching the topic (more than 6 months ago...). The Xbox360 has 3 PPE units and no SPE units while the PS3 has 8 SPE units and 1 PPE unit. The PPE units process sequentially while the SPEs are used for parallel executions, in the case of the PS3, they will enable highly detailed physics modelling... The Xbox 360 just uses 3 PPE units. The best thing to do is read the Anandtech articles if you have computer architecture background, they are all interesting technologies. It is worth reading anything you can find on Anandtech and Tom's Hardware, to name a couple of the more technical sites, to understand how these processors function to try and accomplish the same goals. If you don't have the background in computer architecture, then I apologize for not having enough time to hold your hand through it, it is a complex subject.:cry:

Ironically enough, the basic CPU technologies of both systems were developed by IBM and Toshiba, at the least the PPE which does the sequential processing is incorporated into both systems. I thought one other company was involved in developing the technologies but I can't find the info off-hand at the moment.

Here are links that will be helpful for doing your own research:

PS3 Wikipedia

Xbox 360 Wikipedia

Comparison of next gen consoles Wikipedia

Anandtech Cell Processor article (PS3)

Anandtech Xbox 360 article

[JKalman]

I already have the Xbox 360 and will buy the PS3 also. I'm not bothering with Nintendo this time around...

The Xbox 360 is worth it just for Elder Scrolls 4 IMO. Though you can get that for the computer I suppose, but it is a lot cheaper to buy the Xbox 360 so I don't have to upgrade my computer for another year. Then in a year I can put together a nice SLI system (maybe even a 4 way SLI if they really put one out by then!).

[JKalman]

In short, your topics (info can be gotten from the articles I supplied, but here are short answers):

graphics = PS3

processing power = Depends. Because the PS3 has only one sequential processing unit it will have to use the SPEs often which means that there will have to be a lot of loop unrolling in order to prevent branch prediction (since the SPE deosn't do branch prediction). This will make programming compilers for AI complex. Graphically the PS3 is a killer but with the 360 having 3 sequential units, it may be able to handle AI better, but as Carmack has said in the past about AI in games, (paraphrased very loosely) you can fake AI good enough without needing to actually have any AI.

It is hard to make a direct comparison and say one is better... In other words, they are very different architectural layouts. I personally think the PS3 appears to have much more potential though. It can massively parallel process amazing physics simulations, it can process graphics and physics simulations much better than the xbox 360 -- it literally blows it out of the water. Like I said, it can accomplish real time rendering with those SPEs and the graphics GPU (GPUs? I'm too lazy to go look it up).

video resolution = PS3. Xbox 360 will only have up to 1080i, PS3 will output 1080p.

I/O technology like HDMI = Both. Both will have HDMI adapters...

audio = Both, keep in mind, neither is audiophile gear and it isn't meant to be...

games = Both. I think both systems will have great games... The question is do you value graphics as the end all of what defines the best games. If so, then probably Sony.

pricing = There is no pricing model for the PS3 yet, but they will be including BluRay as part of the PS3, so possibly more expensive than the Xbox 360. For me it will be my first BluRay player (I'm not even going to buy HD-DVD or Blu-ray except for the one that comes with the PS3 -- those idiots on both sides who decided to start another format war can go screw themselves).

[james_dmi]

In short, your topics (info can be gotten from the articles I supplied, but here are short answers):

graphics = PS3

processing power = Depends. Because the PS3 has only one sequential processing unit it will have to use the SPEs often which means that there will have to be a lot of loop unrolling in order to prevent branch prediction (since the SPE deosn't do branch prediction). This will make programming compilers for AI complex. Graphically the PS3 is a killer but with the 360 having 3 sequential units, it may be able to handle AI better, but as Carmack has said in the past about AI in games, (paraphrased very loosely) you can fake AI good enough without needing to actually have any AI.

It is hard to make a direct comparison and say one is better... In other words, they are very different architectural layouts. I personally think the PS3 appears to have much more potential though. It can massively parallel process amazing physics simulations, it can process graphics and physics simulations much better than the xbox 360 -- it literally blows it out of the water. Like I said, it can accomplish real time rendering with those SPEs and the graphics GPU (GPUs? I'm too lazy to go look it up).

video resolution = PS3. Xbox 360 will only have up to 1080i, PS3 will output 1080p.

I/O technology like HDMI = Both. Both will have HDMI adapters...

audio = Both, keep in mind, neither is audiophile gear and it isn't meant to be...

games = Both. I think both systems will have great games... The question is do you value graphics as the end all of what defines the best games. If so, then probably Sony.

pricing = There is no pricing model for the PS3 yet, but they will be including BluRay as part of the PS3, so possibly more expensive than the Xbox 360. For me it will be my first BluRay player (I'm not even going to buy HD-DVD or Blu-ray except for the one that comes with the PS3 -- those idiots on both sides who decided to start another format war can go screw themselves).

The only in game graphics I have seen so far for the PS3 are the Fatal Inertia videos. (if you know of others then let us know) While it looks nice its nothing better then the current initial release XBOX 360 games. You also say that the PS3 will be able to render in real time but that doesn’t seem to mean much without saying rendering what? Pong rendered in real-time, it was just a few pixels but still real-time. I think it can be very easy to fall into the hype surrounding launches and expectations can get very high based on pre-rendered visualisations and non-game tech demos. From what Anand Tech were reporting neither the Cell or the Xbox 360 CPU are anything to get exited about with programmers saying if only they had used an Athlon 64 it would have been much faster. It looks like out –of-order processing, or the lack of, is the culprit. However he does concede that the GPUs are first rate but probably not very different in terms of performance. I think the reality will be a tie for Graphics and processing performance of the machines so it will just boil down to more HD content in games = PS3 or better online gaming and services = XBOX 360 and of course cost which is still an unknown.

Or get both like you and have all bases covered.


James

[james_dmi]

The story I have been referring to was on anandtech but has now been taken down. There is a little of the info posted here:

http://www.techenclave.com/forums/anandtech-on-ps3-xbox-360-a-9970.html

or

http://www.360insider.net/articles/06-29-2005/anandtech-bashes-the-xbox-360-and-ps3/

[Chris D]

Jeff, just FYI, Dell released a quad SLI unit 1-2 months ago. Had a specially painted case with big red flames.

[JKalman]

The only in game graphics I have seen so far for the PS3 are the Fatal Inertia videos. (if you know of others then let us know) While it looks nice its nothing better then the current initial release XBOX 360 games. You also say that the PS3 will be able to render in real time but that doesn’t seem to mean much without saying rendering what? Pong rendered in real-time, it was just a few pixels but still real-time. I think it can be very easy to fall into the hype surrounding launches and expectations can get very high based on pre-rendered visualisations and non-game tech demos. From what Anand Tech were reporting neither the Cell or the Xbox 360 CPU are anything to get exited about with programmers saying if only they had used an Athlon 64 it would have been much faster. It looks like out –of-order processing, or the lack of, is the culprit. However he does concede that the GPUs are first rate but probably not very different in terms of performance. I think the reality will be a tie for Graphics and processing performance of the machines so it will just boil down to more HD content in games = PS3 or better online gaming and services = XBOX 360 and of course cost which is still an unknown.

Or get both like you and have all bases covered.


James

Read up on what Real time rendering is, don't assume it is something that it isn't... :

In short (very short) it is a video game colloqualism having to do with the cut scene movies that occur inbetween sections of a computer game. Being able to render in real time means you will have that video quality during gameplay because the CPU and GPU can draw fast enough to render it all on the fly instead of it being pre-pathed which causes the ned for limited and less detailed pixel mappings.

[JKalman]

Jeff, just FYI, Dell released a quad SLI unit 1-2 months ago. Had a specially painted case with big red flames.

Ah, nice. I don't buy other people's systems though. I only build my own. Then I watercool the CPUs, GPUs and chipsets and overclock the hell out of them. :W

[JKalman]

The only in game graphics I have seen so far for the PS3 are the Fatal Inertia videos. (if you know of others then let us know) While it looks nice its nothing better then the current initial release XBOX 360 games. You also say that the PS3 will be able to render in real time but that doesn’t seem to mean much without saying rendering what? Pong rendered in real-time, it was just a few pixels but still real-time. I think it can be very easy to fall into the hype surrounding launches and expectations can get very high based on pre-rendered visualisations and non-game tech demos. From what Anand Tech were reporting neither the Cell or the Xbox 360 CPU are anything to get exited about with programmers saying if only they had used an Athlon 64 it would have been much faster. It looks like out –of-order processing, or the lack of, is the culprit. However he does concede that the GPUs are first rate but probably not very different in terms of performance. I think the reality will be a tie for Graphics and processing performance of the machines so it will just boil down to more HD content in games = PS3 or better online gaming and services = XBOX 360 and of course cost which is still an unknown.

Or get both like you and have all bases covered.


James

Actually the programmers you are referring to above, who said something about if only they had used Athlon 64s, was something Carmack said in the weeks leading up to the release of the Xbox 360. I read that article, it was good.

Keep in mind though, just because no one can do any high level programming for the system yet does not mean that one doesn't have more potential. The PS3, when you compare what it is capable of, and as they develop a great compiler for it (that takes care of branch predictions via unrolling, and other short cuts and tricks) looks to be capable of a lot more visually, e.g. real time rendering.

[james_dmi]

Actually the programmers you are referring to above, who said something about if only they had used Athlon 64s, was something Carmack said in the weeks leading up to the release of the Xbox 360. I read that article, it was good.

Keep in mind though, just because no one can do any high level programming for the system yet does not mean that one doesn't have more potential. The PS3, when you compare what it is capable of, and as they develop a great compiler for it (that takes care of branch predictions via unrolling, and other short cuts and tricks) looks to be capable of a lot more visually, e.g. real time rendering.

I work in 3D Graphics and what I meant was (although my example was a little tongue-in-cheek) is that rendering in real-time is only impressive if you state at what quality the rendering should be. i.e. ray-tracing, radiosity, HDR fp precision pipelines, post-processing (and to what level) anti-aliasing, normal mapping, sub-surface scattering, motion-blur, depth of field, polygon count, higher-order surfaces etc..... To do cinematic rendering then you need all of the above and that is something that takes a room full of computers weeks to calculate even short sequences. Neither system will get anywhere close to this. The cut scenes in video games are a step lower then this and vary greatly in quality from each-other. I think both systems will eventually get close to the average quality of the cut scenes we are used to seeing in the out going generation of games.

No where have I seen any figures or information that would make me believe that either system would be any better then a current top end PC or each other at generating these “real-time” graphics. There are lots of numbers that give peek theoretical fp performance and such like but these are meaningless in term of real world performance. Now if there were some SPEC FP numbers for the cpu's that would be different but they have not published these. So it goes back to the fact that the only game (i.e. real time rendered graphics) we have seen are about on par whit each other and about what you get with a High end PC.

When you say “high level programming” I think you are referring to optimised code which implies low-level programming and compiler technology. This is an interesting point. I concede that both designs are relying on a gradual improvement in compiler tech over their lifetime to see them through. This is a bit of a gamble especially with the cell as it’s the most different from the norm. Intel gambled on this for years with their long pipeline net-burst architecture (Pentium 4) hoping that they could develop compliers that would cause less branch-predation misses and thus computationally expensive pipeline stalls. That way they could have high clock speed and no penalties. This has not worked out for them and they are now switching back to short pipeline “wide” style processors because they were loosing ground. In this only time will tell who has the balance right in their designs and future predictions for compilers.

[JKalman]

I work in 3D Graphics and what I meant was (although my example was a little tongue-in-cheek) is that rendering in real-time is only impressive if you state at what quality the rendering should be. i.e. ray-tracing, radiosity, HDR fp precision pipelines, post-processing (and to what level) anti-aliasing, normal mapping, sub-surface scattering, motion-blur, depth of field, polygon count, higher-order surfaces etc..... To do cinematic rendering then you need all of the above and that is something that takes a room full of computers weeks to calculate even short sequences. Neither system will get anywhere close to this. The cut scenes in video games are a step lower then this and vary greatly in quality from each-other. I think both systems will eventually get close to the average quality of the cut scenes we are used to seeing in the out going generation of games.

No where have I seen any figures or information that would make me believe that either system would be any better then a current top end PC or each other at generating these “real-time” graphics. There are lots of numbers that give peek theoretical fp performance and such like but these are meaningless in term of real world performance. Now if there were some SPEC FP numbers for the cpu's that would be different but they have not published these. So it goes back to the fact that the only game (i.e. real time rendered graphics) we have seen are about on par whit each other and about what you get with a High end PC.

When you say “high level programming” I think you are referring to optimised code which implies low-level programming and compiler technology. This is an interesting point. I concede that both designs are relying on a gradual improvement in compiler tech over their lifetime to see them through. This is a bit of a gamble especially with the cell as it’s the most different from the norm. Intel gambled on this for years with their long pipeline net-burst architecture (Pentium 4) hoping that they could develop compliers that would cause less branch-predation misses and thus computationally expensive pipeline stalls. That way they could have high clock speed and no penalties. This has not worked out for them and they are now switching back to short pipeline “wide” style processors because they were loosing ground. In this only time will tell who has the balance right in their designs and future predictions for compilers.

I don't know why you are bringing up cinematic rendering, etc, I think it is fairly obvious we are discussing video games and video game technology, thus any terminology would be related to the field of computer/console gaming. Taking the conversation out of this context is not helpful, so I think it is better if we assume from now on that any terms being used (such as rendering) are being used in the context of their traditional meanings in the computer/console video gaming fields (rendering has been a term used a long time by computer/console gaming magazines). This will save me a lot of time and keep the topic on topic.

That being said, I don't recall if it was Tom's Hardware, Anadtech or another technical site which did the study, but one of the sites noted that the Xbox 360 would not be capable of real time rendering, while the PS3 had the technology to be capable of it, because of how its memory had direct pipes to the SPE units and the SPEs in turn had direct connections to the video cards. I think you should go back and take a good look at the SPE units inside the Cell processor. The way they can work their link with RAMBUS directly allows for extreme decompression of Sony's proprietary compression scheme (which I would imagine is on par with the compression scheme Microsoft is using in the Xbox 360) while continuously drawing from pre-allocated memory pools in a usefully parallel manner. Basically the SPEs could function as secondary video cards, or as very good physics simulators, whichever are needed at the time.

As far as being comparable to today's PCs. All you have to do is study the white papers for the PS3 or the Xbox 360 to see how it is potentially better than a regular PC for gaming. That is just hardware design, not including the added problems of a regular PC having to time share components, e.g. manage CPU time for programs (including the OS itself), memory space, I/O buses, HD space when it is used as extended memory, etc. If this were worthy enough of an issue, I would write a paper on what makes these systems better than a regular PC, but it just isn't worth it, and with a little bit of reading you can see for yourself since you have a Computer Architecture background of sorts. To start, each SPE has its own FPU! With the Xbox 360, which has no SPEs, it has three cores, which is one more than any normal PC CPU on the market, I believe. I admit I haven't been in the market for a processor recently, so there could be CPUs with three cores on the mass market for all I know. As you can see, the Cell processor with 8 SPEs can perform much more FP calculations in parallel than any normal PC or the Xbox 360.

No one has seen any games running on a PS3 yet, or has run tests on them, but what kind of tests can you even run on them? Seriously? Running tests meant for a PC on a PS3 would not tell you much. Benchmarks, as many people know, are not very useful, and between very different systems, are practically useless. You would have to write a new benchmark just for the PS3 because its architecture is so different from any PC on the market and any content it runs must be programmed specifically for it, not for x86 (Intel) or x86 compatible architectures (like the AMD chips).

If you take a look at the PS3 architecture and how it works overall, it is easy to see that it will do great things eventually when people learn how to program for it. The problem is, it does not fit into our normal programming paradigms, so it will take a lot of effort on the part of talented programmers to unleash its full potential. Programmers are going to have to learn how to think differently about their approach to programming in general, though compilers can help a lot, they can not do everything.

That games running on a 360 and a regular PC are on par with one another is no big surprise. No one has any idea how to push the Xbox 360 yet and won't for awhile. I don't think the Xbox 360 is any mindblowing breakthrough in technology. I think they rushed it out the door to try and gain a bigger foothold on Sony. No one has pushed the new dual core PCs very far either. Nothing is greatly optimized for dual cores on a regular PC, let alone the Xbox 360, which is what Carmack was complaining about in that article I mentioned in the previous post. Funny enough though, the Xbox 360s similarity to normal PC architecures might be its biggest advantage, since programmers will not have to relearn their whole sequence of programming, they might be able to reach the Xboxs full potential quicker, even if that full potential is not as great.

I disagree with you on your comments concerning Intel and their pipeline technologies. It did work, and it worked extremely well for quite some years. They merely reached the limits of that architecture and its technology. They can't increase the speed of the processor ad infinitum, so they have to move in other directions. Their branch predition units, using their branch prediction hardware and associated algorithms within the hardware, had a very high success rate, something of the order of 97% or higher, though quantifying this is difficult and makes any quantification questionable. Intel always found ways to increase their success rate with predicting the branch that will be taken, or offsetting branch misses, whenever changes in the architecture lowered their prediction success rate.

The change in pipeline lengths is more a result of switching to multiple cores, which they did because they had effectively reached the limits of increasing clock speeds and increasing pipeline lengths due to heat output. They can't use the same pipeline length and institute more cores since they had already reached the limit of heat output due to the clock speeds and pipeline length. Adding a second core and keeping the same pipeline length and CPU speed would mean meltdown. There is more to it than that, but this is an audio forum, and this is off-topic.

This has obviously gone pretty far off topic at this point. If you havn't read the white papers for the PS3, which I don't think you have from what you have been saying, you really need to read it to understand why it is capable of real time rendering when programmers begin to understand its architecture better. Reading an article on Anadtech won't give you that information, unfortunately. I recommend you read the white papers for both the Xbox 360 and the PS3.

[JKalman]

Jeff, just FYI, Dell released a quad SLI unit 1-2 months ago. Had a specially painted case with big red flames.

Chris,

Do you know if that motherboard is available anywhere, I don't seem to be able to find it? Is it custom made for Dell? :cry:]

Thanks.

[james_dmi]

Hi Jeff,

I'm not sure if this is getting a little personal and if not let me apologise for bringing it up but I'm not trying to upset you I'm just pointing out that in real world tests (what little we have) there seems to be no advantage in either platform. I admit that I may have at first mistaken what you were talking about when you referred to real-time rendering. I did not know that this was commonly used in games to refer to real-time game play graphics of the quality of pre scripted 3D graphics used in cut scenes. I do find is amusing though as nearly everything that Sony has shown so far has been pre scripted 3d visualisations. Apart from all the pre –renderd stuff.

You keep talking about theoretical performance but this does not equate to real world performance. Just to qualify this perhaps I should expand a little on what I do. I am a programmer specialising in real time 3D simulations. I also work with pre-renderd 3D visualisations so I do have a perspective from both the maximum quality side of CGI and what real-time graphics are capable. I also have a university background in compiler technology and CPU architecture so I have some knowledge there too.

Now I admit I have not read the white papers on either processor but, as I have contributed to technology white papers in the past, I don’t hold a whole lot of sway in them as far as measuring an actual products performance. Such papers tend to be as much engineers justifying their salaries (or hoped for pay rises) as anything else. Nor do I think that articles written on benchmarking sites will give me that knowledge either. However the article I was referring to was of interest because it contained interviews with developers who were actually working worth both these chips. Now I concede that they may well have been venting their frustration about the very different problems these chips and systems pose from what they are used to. I also believe that both will get better as new programming methods are developed and compliers are optimised. But as one of the few bits of real world information out there it’s all I have to go on and it wasn’t as rosy a picture as either Microsoft or Sony are portraying about there systems. The only other thing is current “in game” graphics and this at the moment is a tie. In the future if people can get more performance out of one system then the other then this may well change.

When you consider a 3D Game there are several stages that data goes through before it hits the screen.

1) The high level game logic, simulated AI, and human input UI
2) High level physics, such as solid body dynamics
3) Object Transform of vertex data to world co-ordinates
4) Low level physics, such as partials for explosions
5) Polygon generation
6) Pixel shading (normal mapping etc)
7) Image processing (for things like glow and effects on HDR data)

This is simplified, and I have missed out many things like shadowing, but it will do to illustrate the point I'm making. Now this is a pipeline and all data starts at 1 and ends at 7 there is no going backwards. In today’s games the CPU is responsible for 1 to 4 and the GPU 5 to 7 (although 4 could also be operated on by some modern GPUs) Games tend to be GPU bound with top systems because 1 to 3 take much less time then 5 to 7. You will see that I have missed out 4 and this is because not many of today’s games use complex low level physics simulations. Now this is the one place where I can see the Cells SPEs could help. Now in the situation where the Cells one normal CPU can keep up with the GPU and therefore is not a limiting factor then the SPEs will probably be able to crunch this data far faster then the x360’s two remaining CPU cores could. The thing is I struggle to imaging the type of game where this would be so. In other cases where the designers want to add more general intelligence to a game then the X360 will be faster. So you see it’s not so much a case of, which CPU can do more FP calculations per second, but more a case of which system has a more balanced design. Any graphics pipeline will only be as fast as its lowest performance stage! You could look at it another way and say because the GPU in the X 360 has so much more memory bandwidth then it will be 10 times faster at drawing 3d rendered frames then the PS3. Of course this is ridicules but it illustrates the point that you cannot look at a single aspect of the system in isolation.

Now I want to tackle the peak performance issue. This is the area where I disagree with your assessment most strongly. You seem to be basing your assertion that the Cell can do more calculations in parallel form the peak performance numbers. But in real world data processing this is a million miles away form reality. Peak FP performance is calculated by the number of FP capable units in a processor * the number of operations these units can do in a cycle * the clock rate. Even in the simplest possible situation which would be to have two sets of fp numbers arranged in order in two array in memory that the processor transverses multiplying the numbers together and storing the numbers in a third memory array you cannot even get close to this figure. There are other bottlenecks such as loading numbers form memory, cache checking, register organisation, and storing the numbers that will slow the whole thing down. And if you conceder something approaching useful computing you need to add branching logic, integer math, and many other things into the mix. Now until there are systems out there it is impossible to make any assessments as to how well the cell will really work. I have heard all sorts of things from companies wanting to push there new toys and more often then not it turns out to be bull. The PS2 launch comes to mind where they said it would have graphics as good as those seen in films. It turned out to just have graphics about as good as the PCs of the day.

[JKalman]

Hi Jeff,

I'm not sure if this is getting a little personal and if not let me apologise for bringing it up but I'm not trying to upset you I'm just pointing out that in real world tests (what little we have) there seems to be no advantage in either platform. I admit that I may have at first mistaken what you were talking about when you referred to real-time rendering. I did not know that this was commonly used in games to refer to real-time game play graphics of the quality of pre scripted 3D graphics used in cut scenes. I do find is amusing though as nearly everything that Sony has shown so far has been pre scripted 3d visualisations. Apart from all the pre –renderd stuff.

You keep talking about theoretical performance but this does not equate to real world performance. Just to qualify this perhaps I should expand a little on what I do. I am a programmer specialising in real time 3D simulations. I also work with pre-renderd 3D visualisations so I do have a perspective from both the maximum quality side of CGI and what real-time graphics are capable. I also have a university background in compiler technology and CPU architecture so I have some knowledge there too.

Now I admit I have not read the white papers on either processor but, as I have contributed to technology white papers in the past, I don’t hold a whole lot of sway in them as far as measuring an actual products performance. Such papers tend to be as much engineers justifying their salaries (or hoped for pay rises) as anything else. Nor do I think that articles written on benchmarking sites will give me that knowledge either. However the article I was referring to was of interest because it contained interviews with developers who were actually working worth both these chips. Now I concede that they may well have been venting their frustration about the very different problems these chips and systems pose from what they are used to. I also believe that both will get better as new programming methods are developed and compliers are optimised. But as one of the few bits of real world information out there it’s all I have to go on and it wasn’t as rosy a picture as either Microsoft or Sony are portraying about there systems. The only other thing is current “in game” graphics and this at the moment is a tie. In the future if people can get more performance out of one system then the other then this may well change.

When you consider a 3D Game there are several stages that data goes through before it hits the screen.

1) The high level game logic, simulated AI, and human input UI
2) High level physics, such as solid body dynamics
3) Object Transform of vertex data to world co-ordinates
4) Low level physics, such as partials for explosions
5) Polygon generation
6) Pixel shading (normal mapping etc)
7) Image processing (for things like glow and effects on HDR data)

This is simplified, and I have missed out many things like shadowing, but it will do to illustrate the point I'm making. Now this is a pipeline and all data starts at 1 and ends at 7 there is no going backwards. In today’s games the CPU is responsible for 1 to 4 and the GPU 5 to 7 (although 4 could also be operated on by some modern GPUs) Games tend to be GPU bound with top systems because 1 to 3 take much less time then 5 to 7. You will see that I have missed out 4 and this is because not many of today’s games use complex low level physics simulations. Now this is the one place where I can see the Cells SPEs could help. Now in the situation where the Cells one normal CPU can keep up with the GPU and therefore is not a limiting factor then the SPEs will probably be able to crunch this data far faster then the x360’s two remaining CPU cores could. The thing is I struggle to imaging the type of game where this would be so. In other cases where the designers want to add more general intelligence to a game then the X360 will be faster. So you see it’s not so much a case of, which CPU can do more FP calculations per second, but more a case of which system has a more balanced design. Any graphics pipeline will only be as fast as its lowest performance stage! You could look at it another way and say because the GPU in the X 360 has so much more memory bandwidth then it will be 10 times faster at drawing 3d rendered frames then the PS3. Of course this is ridicules but it illustrates the point that you cannot look at a single aspect of the system in isolation.

Now I want to tackle the peak performance issue. This is the area where I disagree with your assessment most strongly. You seem to be basing your assertion that the Cell can do more calculations in parallel form the peak performance numbers. But in real world data processing this is a million miles away form reality. Peak FP performance is calculated by the number of FP capable units in a processor * the number of operations these units can do in a cycle * the clock rate. Even in the simplest possible situation which would be to have two sets of fp numbers arranged in order in two array in memory that the processor transverses multiplying the numbers together and storing the numbers in a third memory array you cannot even get close to this figure. There are other bottlenecks such as loading numbers form memory, cache checking, register organisation, and storing the numbers that will slow the whole thing down. And if you conceder something approaching useful computing you need to add branching logic, integer math, and many other things into the mix. Now until there are systems out there it is impossible to make any assessments as to how well the cell will really work. I have heard all sorts of things from companies wanting to push there new toys and more often then not it turns out to be bull. The PS2 launch comes to mind where they said it would have graphics as good as those seen in films. It turned out to just have graphics about as good as the PCs of the day.

Nope, I'm not taking it personally at all, and sorry I have not been able to respond sooner, I was away for a couple of days. I don't know why the idea of theoretical vs. real performance has become an issue in this arguement. As you stated, and seem to forget later in your argument, there is no real world test of the Cell processor yet, so how can you compare the two and say it is a tie, according to your own statements? The point I have been making and which I still think is true is that the PS3 will display better graphics than the XBox 360. I think it will be a noticeable difference if you have the proper TV and if the game is programmed for 1920x1080. I think you are taking the discussion of video card technology a little overboard because the design of the GPUs used in both systems are proprietary, though they likely do follow similar pipeline stages as those you listed. Being designed specifically for a system allows them to integrate the card into the system in a much tighter, more useful, fashion, which is something normal video cards can not do as well because they must adhere to plugnplay type protocols allowing them to fit into a slot, and not to be directly fused into the best spot on the motherboard.

The reason I keep highlighting the SPEs importance is because they allow a lot of work that would normally be done in the GPU to be offloaded into the CPU cores. This and the ability of the SPEs to work in tandem on a single complex task, by continually drawing from its appointed memory space, give it a great advantage over the Xbox 360. Supposedly it will be capable of almost twice the number of dot-product operations as the Xbox 360 (PS3s 51 to Xbox 360's 33 billion, both approximate). This is a significant difference. I still don't know where you are getting your notion that their in game graphics are a "tie". This just isn't so, but I can't force you to believe that and I can't even get you to look into it more than you have, which I guess I don't blame you for, since it would take time probably better spent doing something else. Many of the stages you listed could benefit from offloading to the SPEs. Not that it matters if the PS3's NVidia GPU is much faster than the Xbox 360s GPU anyway, which I do not know. You also can't be certain what kind of implementations were put into the NVidia card to take care of potential stages that would cause bottlenecks. Luckily system designers tend to take these things into account when designing their systems and improve on them more as the design is fleshed out and stages of it are built and tested.

You don't need to explain how graphics cards or any of this stuff works, I'm well familiar with it, as I have a BS in Computer Science with a specialization in Networking (Upsilon Pi Epsilon, Delta Epsilon Sigma) and was working on an MS at Columbia University, which I am currently on a leave of absence from, because I am considering a career change. I am very familiar with Amdahls law and all the other basic, but fun stuff. I have been looking at these topics we are arguing within the gestalt though you seem to try and imply otherwise. I have also been computing for 21 years, since I started programming on a Tandy with using Basic at 10.

I have never even mentioned anything about peak performance numbers. I think you are going way overboard... You were the one who brought up FP in the first place not me. I only pointed out that the SPEs each have a FPU since you had posted something about the PS3 and a lack of FP operations or something to that effect... I'm too lazy to go back and look it up.

The curious thing I find is in one of your quotes, "Now until there are systems out there it is impossible to make any assessments as to how well the Cell will really work", you are doing exactly the opposite. You are assuming it will tie the Xbox 360 graphically even after saying you can't make such an assumption. I have no problems with making educated assumptions, that is what my whole argument is about, but if you do, it behooves you to be consistent.

As far as in game logic goes, I made this point previously in a past post. Carmack himself says this isn't needed at this point because AI can be faked much easier than it can be produced. The Cell would be at a disadvantage if someone were to create something with extreme amounts of logic, but face it, this is a video game machine, and video game producers don't have enough time allocated to any of their game budgets to give it the kind of AI that would create this problem, at least in this next generation of game consoles.

I don't see the point of arguing this further if you are not going to take the time to understand how the PPE and SPE units work together with memory, the GPU and the rest of the system. As far as I can tell you are still assuming the Anandtech author is correct and only arguing your side based on that without making any points or disproving anything I am saying.

It has been fun, I'm done with this side argument. I'm not taking this personally and I'm not angry, for some reason people tend to think I am often. Perhaps it is my writing style...

[Chris D]

Sorry, Jeff! Don't know anything about that Quad-SLI Dell. I just remember seeing it on an Intel electronic page as an ad, and when I clicked on it, it linked to the computer on the Dell website.

Perhaps a search on Intel or Dell's websites?

[james_dmi]

Hi Jeff,

As you have said you’re not upset by my counter arguments then I don’t even want to class this as an argument. You make many valid points in your posts which I find interesting. This is more of a conversation between people who are both obviously interested in the technology. It’s also a far more informed debate on this subject then the usual fan boy bashing you see elsewhere.

Let me explain why I have bought up several issues in my posts:

Theoretical vs. Real world: Well you sort of answered this yourself. Everything we have read about the Cell has been theoretical performance and it’s that I am questioning. I have read many a time about how great a product will be in theory only to be disappointed in reality.

Peak FP numbers: Much of the commentary about the effectiveness of the SPEs has been based on the theoretical performance figures Sony banded about at the PS3 launch. Although you have not bought this up before I wanted to put a counter to this as many of the articles you refer to do contain this information and it is usually the metric by which they base their arguments (understandably as there is nothing else at this time). You have now of course mentioned one of these peak performance numbers in your last post. (To those casually reading these posts if you’re interested and don’t think we are the biggest pair of techno nerds on the planet: FP performance is the most important aspect of a CPU for 3D CGI.)

Assuming the Xbox and PS3 will tie on rendering performance: Well I’m not assuming this but what I am saying is that if we only go on real world evidence of performance in real game situations then the only in game graphics we have seen would point to this being the case. Now of course those PS3 in game graphics could be from a very poor game and as I did say both architectures will undoubtedly grow and as they do one may become much more powerful then the other. But right now it still looks like a tie and there is no way of knowing which will grow more or sooner.

Anandtech Author: I’m not basing any of my opinions on anything written by a Anandtech author. The only thing I am referencing is interview with developers working on both architectures. This was very balanced and didn’t overly favour either system over each other. I made this point to caution against over inflated expectation people seem to have on how nex-gen either of these systems really are.

On SPEs being able to offload many of the stages in a 3D pipeline: It was exactly this assumption, made by many commentators, which I want to challenge. If, as you suggest could be the case, the PS3 RSX chip had been specifically designed to work with the Cell then you could be right. The area where the cell could help out would be in the pixel and vertex shading. If not then once the GPU takes over from the CPU then it is very difficult and will be computationally expensive to pass data back and forth between the two. Unfortunately for Sony it’s the Latter, it’s well known and has been admitted by Nvidia that the RSX is based on the G70 core. Also if you think about how long it takes to design chips the only way the two could work together in this way would be either for both to be designed by the same team or for one to exist before the other.

On Cell performance I am happy to admit I could be wrong I'm only making projections based on the evidence as I see it (just as you are). Time will tell who is closer to the truth.

I look forward to the next instalment in this conversation but right now I have bugs in my ray-tracing and deadlines looming.

Speak soon,

James.