Xbox 360 "Xenon" CPU/PS3 "Cell" CPU - Any Info?

[SickBeast]

I remember reading a quote by a game developer stating that the Xenon is essentially only twice as fast as the 733mhz P3 in the original Xbox. I don't see how this can be true! A triple core 3.2ghz should be 6 times faster, even if it has deep pipelines and performs like a dog no?

And what about the Cell? It's very similar no? What's the deal on these chips. Do they really suck or what?

 

[SickBeast]

See, nobody's stoked. :thumbsdown:

 

[Zenoth]

My opinion on this is: Who cares, as long as the games are good.

Bah ... that's just me I know. But I don't care about such technical details when it comes to video game Consoles. I see them as pure gaming machines, that's it. Don't get mixed about upgrading, compatibility problems, Power problems, reboots, and such things.

You know, it's just a matter of "insert the game media and play". I'm glad it's still remains such a way for that next generation.

 

[SickBeast]

Originally posted by: Zenoth
My opinion on this is: Who cares, as long as the games are good.

Bah ... that's just me I know. But I don't care about such technical details when it comes to video game Consoles. I see them as pure gaming machines, that's it. Don't get mixed about upgrading, compatibility problems, Power problems, reboots, and such things.

You know, it's just a matter of "insert the game media and play". I'm glad it's still remains such a way for that next generation.

No, it's not just you it's EVERYONE. Noone is stoked about this stuff! I don't think I've ever seen a thread in here on this topic.

Speaking of which, does IBM intend to port this stuff to the PC and make CPUs again? Or are these things not x86 and totally Windows incompatible? I'm guessing yes.

 

[stratman]

I'm not too stoked because multiple cores have not traditionally been beneficial to games, and because IBM's processors have not traditionally been good at running games (maybe it's the fault of the rest of Apple's platform, but I can't help but see some correlation).

Developers are not too stoked because programming games to be multithreaded is apparently pretty tough; you've probably seen Gabe Newell's reflections on the matter.

Edit: yeah, these chips are def. not x86, they're a derivative of IBM's POWER line, if I remember correctly. IBM's POWER4 chips are already used in today's Macs, and their POWER5 chips are used for servers/enterprise stuff IIRC. I don't believe either of those families of chips are better than our current AMD/Intel offerings, they just have different strengths.

 

[milomnderbnder21]

Multithreaded games are the future...Because you can only do so much with a single chip. Developers better get used to it...If multiple cores have not been helpful before, it's because developers have been too lazy to utilize them right. Have you seen the 20 minutes of Elder Scrolls Oblivion footage? If not, go to fileplanet or something and do so...Multiple Cores (with a nice GPU) can make some nice stuff...

 

[klah]

Until they release the SPECfp, SPECint and BLAS results I am going to assume they are the equivalent of x86 cpus from 5 years ago.

 

[SickBeast]

Originally posted by: klah
Until they release the SPECfp, SPECint and BLAS results I am going to assume they are the equivalent of x86 cpus from 5 years ago.

:Q

 

[bobsmith1492]

The article at Anandtech (that got strangely pulled or something?) said they lack basic features of modern CPUs such as branch predictors, and being apparently deeply pipelined, are affected hugely by stalls. Also, having that many cores doesn't help much in the majority of gaming code situations since it is impossible to code sequential operations to be split among processors (i.e. you need the answer to question "1" before you can start question "2"...)

That's the gist of what I remember from that article. Does no one else remember that?

 

[Topweasel]

Thats the thing Is computers are at the Point where they beat them out of the gate even before they are released, It came close last generation and this ones not even going to be close. So technologically niether one is worth a darn. In the end it will all be about implimentation and game Titles, whomever does this wins the war. Sony has had a better history getting backers and has many hands in side games that are released solely for their hardware. This might change though since the Xbox is using a form of DX and the cores will easier to program for. Sony devleopers are stuck in a situation where they are using a 3rd new arch, whith a proreitary API that only Sony has worked with, and 7 simple and feature less cores with limited usability that they have to squeeze as much proformance out of.

Non of this matters as both will look good and you will have to make a decision on what titles each will carry before making your decision.

 

[Wingznut]

Originally posted by: stratman
I'm not too stoked because multiple cores have not traditionally been beneficial to games, and because IBM's processors have not traditionally been good at running games (maybe it's the fault of the rest of Apple's platform, but I can't help but see some correlation).

The ONLY reason for the above is that developers have not created games that take advantage of multiple processors. That has changed with the Xbox 360.

You can have one core handling the AI, another core handling the physics, and the other handling the actual gameplay. How can that not be beneficial to games?

There are going to be things going on with the Xbox 360 that we've never seen before. Kameo is an excellent example. Hundreds of enemies onscreen, each with their own individual AI. Amazing.

 

[RichUK]

im buying one, i have already pre-ordered it, and it will arrive on december the 2nd

i think im about the only one who is going to buy one on this forum , but hey i have a sweet 42" plasma TV in which to use it on .. so xmas is going to be sweet

 

[Continuity28]

Originally posted by: milomnderbnder21
Multithreaded games are the future...Because you can only do so much with a single chip. Developers better get used to it...If multiple cores have not been helpful before, it's because developers have been too lazy to utilize them right. Have you seen the 20 minutes of Elder Scrolls Oblivion footage? If not, go to fileplanet or something and do so...Multiple Cores (with a nice GPU) can make some nice stuff...

Yeah but think of it this way:

You are a processor now, single core, here are your instructions:

A = 5
B = 4
Add A and B
Z = 37
Divide B by Z and add that to AB.

Notice, this code is utterly simplistic and not useful, its just an example. Also notice that you cant make this multiple threaded. Otherwise you get this:

Processor A:

A = 5
Add A and B
Divide B by Z and add that to AB

Sorry but Processor A can't do that, its missing too many variables, which processor B has... once they communicate with each other on it, the delay has far exceeded that which would have existed if one processor did it all.

The bottom line: Not EVERY piece of code CAN be made to work on multiple threads. Certain programming is also random to an extent, especially in games. You have freedom of movement and control, the processor needs to be ready for your input, and with a million combinations, this can also have an impact on IN-ORDER processors like those in these consoles, and also multi-threaded capability. It's not as easy as you make it sound, and in many cases it would be impossible, like my simple math example.

 

[biostud]

wlecome to half a year ago:
http://www.anandtech.com/video/showdoc.aspx?i=2453
http://arstechnica.com/articles/paedia/cpu/xbox360-2.ars/1
http://arstechnica.com/articles/paedia/cpu/cell-1.ars
http://arstechnica.com/articles/paedia/cpu/cell-2.ars

 

[Wingznut]

Originally posted by: Continuity28

Originally posted by: milomnderbnder21
Multithreaded games are the future...Because you can only do so much with a single chip. Developers better get used to it...If multiple cores have not been helpful before, it's because developers have been too lazy to utilize them right. Have you seen the 20 minutes of Elder Scrolls Oblivion footage? If not, go to fileplanet or something and do so...Multiple Cores (with a nice GPU) can make some nice stuff...

Yeah but think of it this way:

You are a processor now, single core, here are your instructions:

A = 5
B = 4
Add A and B
Z = 37
Divide B by Z and add that to AB.

Notice, this code is utterly simplistic and not useful, its just an example. Also notice that you cant make this multiple threaded. Otherwise you get this:

Processor A:

A = 5
Add A and B
Divide B by Z and add that to AB

Sorry but Processor A can't do that, its missing too many variables, which processor B has... once they communicate with each other on it, the delay has far exceeded that which would have existed if one processor did it all.

The bottom line: Not EVERY piece of code CAN be made to work on multiple threads. Certain programming is also random to an extent, especially in games. You have freedom of movement and control, the processor needs to be ready for your input, and with a million combinations, this can also have an impact on IN-ORDER processors like those in these consoles, and also multi-threaded capability. It's not as easy as you make it sound, and in many cases it would be impossible, like my simple math example.

In that instance, multiple cores wouldn't be processing the single instruction. While Processor A is figuring out that calculation, Processor B can be doing another, and Processor C can be doing a third. In your example, the application would have to wait for the single processor to become available for the third calculation.

Yes, this is all very much an oversimplification. And yes, not EVERY instance will benefit from mutiple cores. But if the software is designed purely for the multiple core environment, most of the work can be done in parallel.

 

[White Widow]

From what I have read, there are a couple basic reasons why no one is stoked about these chips.

First, these CPU's only really shine when all of their specialized cores are used. Very little CPU power is dedicated to "general" processing as opposed to stuff like SIMD. If a game developer doesn't code to use these specialized cores, then they don't benefit at all from them even being there.

Second, it is WAY more difficut and time-consuming to write code to be multi-threaded in a useful way. Most game programmers have little or no experience writing this kind of code, and even if they do, it is quite simply far more difficult. So, developers are faced with a choice of whether to extend their development time SUBSTANTIALLY to try and leverage the extra performance - which may or may not even be possible (a lot of code simply can't be multi-threaded).

Even those development houses that are trying to write code for these speciailized cores are only using them in marginal ways - for example to do some physics.

Every interview I have read with developers has them feeling rather confused and exasperated at the situation. They still get more overall single-threaed CPU power with these consoles (like 2x over the P3-733), but if Sony and/or M$ has just used a higher-freq. single-core CPU they would have gotten a LOT more. Especially Sony seems to have been slammed with criticism for devoting too much CPU power to specialized registeres and not enough to general processing.

Note - one reason for the enhanced performance of these consoles over the previous generation is the greatly improved GPUs. It kind of goes without saying, but these things simply ROCK their older counterparts, and the guys at the game development houses ARE stoked about these.

 

[akugami]

Originally posted by: stratman
I'm not too stoked because multiple cores have not traditionally been beneficial to games, and because IBM's processors have not traditionally been good at running games (maybe it's the fault of the rest of Apple's platform, but I can't help but see some correlation).

Developers are not too stoked because programming games to be multithreaded is apparently pretty tough; you've probably seen Gabe Newell's reflections on the matter.

Edit: yeah, these chips are def. not x86, they're a derivative of IBM's POWER line, if I remember correctly. IBM's POWER4 chips are already used in today's Macs, and their POWER5 chips are used for servers/enterprise stuff IIRC. I don't believe either of those families of chips are better than our current AMD/Intel offerings, they just have different strengths.

I agree that multi cores haven't been really helpful in games but you have to remember that games haven't really been built on multi-core processors. It's only been in the last year or so that dual cores (or dual procs) have been available to the general mainstream market.

IBM also makes the Gecko core based PPC CPU's used in Nintendo's Gamecube which is actually very competitive for games with the PS2 and Xbox. I'm sure that even though the new Xbox 360 and likely the Nintendo Revolution CPU's will be based on Power cores, they will be implementing things they learned making the Gamecube CPU.

Game coders have been sloppy and a bit lazy in their coding with the ever faster processors Intel and AMD have made. Whereas now they're hitting a wall with processor speeds and going dual core to help gain more power out of these chips. This will require new coding and will stress efficient code and anyone who's still writing bloated unoptimized code will soon get left behind.

I think console game makers will also have to adjust to working with making their games better at multi-threading but at the same time they are more used to optimizing code than their PC brethrens since game consoles, in general, do not get more powerful as time goes by. They have to squeeze more power out of hardware that is static to try to differentiate themselves from their competitors.

In the end, it's all about the games. Game developers will either get used to multi-threaded designs or they won't. If they don't, they will likely be left on the side of the road. Adapt or die.

 

[SickBeast]

Interesting. It sounds like these consoles have tons of future potential when people start figuring out how to effectively code multithreaded games. I wonder if they could just use a multithreaded graphics driver like the new nVidia drivers.

 

[RichUK]

 

[Hacp]

Originally posted by: RichUK

Exactly

 

[bobsmith1492]

In reality, how much of the graphics actually depend on the CPU? Not much, really. I mean, in the overclocking roundup thing lately, the slowest A64 (Venice 3000) was still pushing something like 80+ frames in the most intensive games out there. The only limiting factor was the GPU when it came to upping the graphics - it's not really worth it to spend a lot on the CPU unless you are already at the lower edge of tolerable frame rate.

The only thing these console CPUs need to do is handle the data coming through fast enough to keep up 30 FPS or whatever it is TVs handle these days. The GPUs will limit the resolution and amount/detail of items onscreen. I'm thinking that's why there's not too much fuss about the CPUs in these things.

 

[Velk]

Originally posted by: akugami

Originally posted by: stratman
I'm not too stoked because multiple cores have not traditionally been beneficial to games, and because IBM's processors have not traditionally been good at running games (maybe it's the fault of the rest of Apple's platform, but I can't help but see some correlation).

Developers are not too stoked because programming games to be multithreaded is apparently pretty tough; you've probably seen Gabe Newell's reflections on the matter.

Edit: yeah, these chips are def. not x86, they're a derivative of IBM's POWER line, if I remember correctly. IBM's POWER4 chips are already used in today's Macs, and their POWER5 chips are used for servers/enterprise stuff IIRC. I don't believe either of those families of chips are better than our current AMD/Intel offerings, they just have different strengths.

<snip>

Game coders have been sloppy and a bit lazy in their coding with the ever faster processors Intel and AMD have made. Whereas now they're hitting a wall with processor speeds and going dual core to help gain more power out of these chips. This will require new coding and will stress efficient code and anyone who's still writing bloated unoptimized code will soon get left behind.

I'd suggest that bloated and unoptimized code is actually the easiest to convert to multithreaded, with extremely efficient and highly optimized code the most difficult.

Edit:typo.

 

[Wingznut]

Originally posted by: bobsmith1492
In reality, how much of the graphics actually depend on the CPU? Not much, really. I mean, in the overclocking roundup thing lately, the slowest A64 (Venice 3000) was still pushing something like 80+ frames in the most intensive games out there. The only limiting factor was the GPU when it came to upping the graphics - it's not really worth it to spend a lot on the CPU unless you are already at the lower edge of tolerable frame rate.

The only thing these console CPUs need to do is handle the data coming through fast enough to keep up 30 FPS or whatever it is TVs handle these days. The GPUs will limit the resolution and amount/detail of items onscreen. I'm thinking that's why there's not too much fuss about the CPUs in these things.

It's an even consideration. Things like advanced AI, more objects (and more of them acting independantly) on screen at a time, better physics, better lighting... All of those things can be improved by the cpu.

 

[stratman]

For sake of interest to this people in this thread, here are Gabe Newell's (the Valve boss) comments regarding the developing of multithreaded game engines and the next generation of consoles, which I made reference to in my first post in this thread:

"Technologically, I think every game developer should be terrified of the next generation of processors. Your existing code, you can just throw it away. It's not going to be helpful in creating next generation game titles," said Newell.

"Most of the problems of getting these systems running on these multicore processors are not solved. They are doctoral theses, not known implementation problems. So it's not even clear that over the lifespan of these next generation systems that they will be solved problems. The amount of time it takes to get a good multicore engine running, the Xbox 360 might not even be on the market any longer. That should scare the crap out of everybody."

Newell cautions "Really good engineers are going to be much more valuable and engineers who used to be valuable writing game code in the previous generation may end up becoming thorns in the side of key programmers who can write multi-core game code."

But if the XBOX360 platform running an IBM CPU is good enough to get an Intel engineer (IIRC) excited about, it's good enough for me

 

[akugami]

Originally posted by: Wingznut

Originally posted by: bobsmith1492
In reality, how much of the graphics actually depend on the CPU? Not much, really. I mean, in the overclocking roundup thing lately, the slowest A64 (Venice 3000) was still pushing something like 80+ frames in the most intensive games out there. The only limiting factor was the GPU when it came to upping the graphics - it's not really worth it to spend a lot on the CPU unless you are already at the lower edge of tolerable frame rate.

The only thing these console CPUs need to do is handle the data coming through fast enough to keep up 30 FPS or whatever it is TVs handle these days. The GPUs will limit the resolution and amount/detail of items onscreen. I'm thinking that's why there's not too much fuss about the CPUs in these things.

It's an even consideration. Things like advanced AI, more objects (and more of them acting independantly) on screen at a time, better physics, better lighting... All of those things can be improved by the cpu.

Was gonna reply in a similar manner, then saw your post. Let's face it, the world's greatest GPU will do squat for you if you pair it with a crap CPU that can't properly handle physics (collision detection, gravity, etc) and AI without slowing to a crawl.