Sony may lose grip in next game consoles war: Article

[mckmas8808]

Ummm, maybe because it's true. Why do you have a problem with people accepting this?

If I said the Xbox was more powerful than the PS2, people don't have a problem with this, so why is the current generation any different in the acceptance of one being more powerful than the other? Is it because people who are fans of the Xbox 360 are afraid that being seen as less powerful will limit it's chances at success?

Well Edge I agree that the PS3 looks to be more powerful, but to be honest at least the Xbox has proven that it's more powerful than the PS2. Scooby is saying we should at least what until games like KZ, Motorstorm, MGS4, etc prove that the PS3 is more powerful.

And I for you kinda agree with him. We can at the very very least wait until we get some true to life RSX information first.

 

[Alpha_Spartan]

What facts do you expect to be released that will provide the answer as to what system is more powerful? Or do you believe this can never be answered?

So you are saying specs are totally useless?

A final PS3 that actually runs. I have a problem with people comparing a PowerPoint presentation to an actual working console that hopefully I'll be playing in 6 hours.

 

[Titanio]

A final PS3 that actually runs. I have a problem with people comparing a PowerPoint presentation to an actual working console that hopefully I'll be playing in 6 hours.

It's funny how this argument is cropping up when we were all very happy to debate things previously when both were in-development consoles. Forgive me, but it smacks of stretching for any means to delay/avoid the comparison, now that you can differentiate one as being "out" and the other as not. Expecting/Suggesting PS3, or the PS3 described by Sony, will not become a "console that actually runs" or the like is a little bit much.

 

[scooby_dooby]

Ummm, maybe because it's true. Why do you have a problem with people accepting this?

That's the entire problem. It's NOT true. It's still to be decided.

If CELL is only ever able to reach 20% of it's theoretical potential, while XeCPU is able to reach 50% of it's theoretical potential, than it's the XeCPU that is more powerful.

Until we see this processor in teh real world, running real games, and we can see how EFFICIENT it really is, then and only then can you make a legitimate comparison.

Also, Software will play a huge role in determinging which CPU gets exploited to it's fullest potential.

So without knowing the realworld efficiency of CELL, and how much of that power dev's will truly be able to extract, and at the same time acknowledging that MS is likely to have a superior software tools(which will impact efficency), it's WAAAY to premature to conclude PS3 is giong to be more powerful.

It's not paper-specs that matter and peak #'s that matter, it's the the visuals on the screen.

 

[OtakingGX]

Read again, what I posted:

CELL has:
- More than twice the internal bandwidth
- More than twice the external bandwidth
- More than twice the number of processors
- More than twice the amount of on-chip memories
- More than twice the floating point rate
of the Xbox 360 CPU.

CELL is superior to the Xbox 360 in a number of CRITICAL areas that all help contribute to greater throughput.

I'm not just comparing GFLOP's.

Internal bandwidth doesn't really seem to matter, what with the old Athlon XP having 1/4 that of a P4 of the same speed and still being faster in many cases. The Athlon 64 has 1/2 that of the same speed P4 and is faster than P4's with a 50% higher clock.

Same goes for external bandwidth. The old XPs were rather deficient in that regard to the P4 and fared admirably in benchmarks.

I'm not so sure about "number of processors." The way I see it, the SPEs are little sub-processors. Each one is really good at vector math, and not much else.

Again, I'll point to the Athlon 64 trouncing the P4 with twice the L2 cache in many a benchmark. More cache doesn't necessarily equate to better performance.

There are also many ways of measuring peak FLOPS that I'm sure have been discussed before. Do you mean double-precision FLOPS, because the SPEs are geared for single precision. How about SIMD operations? I don't think FLOPS is an accurate way of judging a processor's performance either.

 

[Edge]

If that's the criteria around here, then fine. I don't mind waiting, and somewhat confused why this criteria is not more emphasized around here, as there are many posts comparing the two systems.

As matter of fact, Asher's comments in another thread was claimed to know what he was talking about in a comparison between the two systems, even by you Scooby Dooby, and I was told I did not know what I was talking about. Remember, you did not say the comparison was not valid, but said

Edge you're so far in over your head it's funny. This guy actually worked on the CELL processor, and you're sitting here arguing with him with numbers based on hyper-threaded P4-style multithreading(i'm assuming).

from here:
http://www.beyond3d.com/forum/showthread.php?t=25719&page=7&highlight=asher

Should it not have been brought to Asher attention that he cannot make a comparison at this point?

That entire thread was titled " Does Cell Have Any Other Advantages Over XCPU Other Than FLOPS?"

I don't see any strongly worded objections by you Scooby Dooby that the thread's discussion was invalid?

What am I missing here?

 

[Edge]

Internal bandwidth doesn't really seem to matter, what with the old Athlon XP having 1/4 that of a P4 of the same speed and still being faster in many cases. The Athlon 64 has 1/2 that of the same speed P4 and is faster than P4's with a 50% higher clock.

Same goes for external bandwidth. The old XPs were rather deficient in that regard to the P4 and fared admirably in benchmarks.

I'm not so sure about "number of processors." The way I see it, the SPEs are little sub-processors. Each one is really good at vector math, and not much else.

Again, I'll point to the Athlon 64 trouncing the P4 with twice the L2 cache in many a benchmark. More cache doesn't necessarily equate to better performance.

There are also many ways of measuring peak FLOPS that I'm sure have been discussed before. Do you mean double-precision FLOPS, because the SPEs are geared for single precision. How about SIMD operations? I don't think FLOPS is an accurate way of judging a processor's performance either.

Those are good points, but CELL has 256 KB of SRAM per SPE processor, and one can argue, because it's not cache, but localized memory, it will allow better predictability by programmers, as to where their data will be. Why? Because developers will be responsible for data flow in loading and unloading that memory. Give me a programmers brain anyday over a cache algorithm.

Yes I mean single precision FLOPS, and you don't need double precision for a game console.

Each SPE has SIMD instructions.

Each one is really good at vector math, and not much else.

I don't see any evidence of this. We are not talking about math coprocessors, but processors, that will have no problems running general purpose code. One could argue it's less efficient than the PPE in running integer code, just as the PPE is less efficient than an Athlon in running integer code. Now no one says because the PPE is less efficient in running integer code than an Athlon, that it's no good in running integer code. The SPE difference over the 360 CPU, is that there is more processors to make up the difference, all the while having massive floating point capability.

 

[Laa-Yosh]

Give me a programmers brain anyday over a cache algorithm.

You know, the problem with nextgen software development is that your supply of programmer brains is seriously limited...

 

[Lessard]

2) CELL is a lot more powerful than the 360 CPU, thus giving the PS3 a TWICE performance advantage in that area.

Yeah and the Jaguar was TWICE more powerful than the playstation you know ...
coz it was a 64bit console and the psx a 32 bit one ...

 

[Edge]

You know, the problem with nextgen software development is that your supply of programmer brains is seriously limited...

Seems to be lots of PS2 developers that are very clever in getting great performance out of the PS2. I assume those programmers will still be around for this generation.

 

[Laa-Yosh]

Seems to be lots of PS2 developers that are very clever in getting great performance out of the PS2. I assume those programmers will still be around for this generation.

Compared to the number of games released on the PS2, skilled developers are an absolute minority. And even if you have a lot of very good programmers, the amount of stuff that they can do in a reasonable timeframe still might not be enough. All that power from having low-level access might get left unused if there's no time to implement it properly, and what seemed to be an advantage might end up an obstacle.
And FYI, most of the stuff that could differentiate this next gen from the previous one are high-level things, which will be even more difficult to implement if the coders have to fight the low-level stuff...

Edit: for example, let's assume that we have a coder whose mad skillz can give us about 10 times as much processing power to spend on AI, so we can have 10 times as many enemies compared to the previous gen. Or, we have a coder who can get 5 times the processing power but he can use it to create the same number of AI enemies, but 5 times as clever. Which one would be better? Which would you consider "next-gen gameplay"?

 

[compres]

Give me a programmers brain anyday over a cache algorithm.

Damn edge, I am not sure how to reply to you, don't want to seem rude.

That little statement seems to suggest you have no clue about software engineering. When you write a program that 20 or more co-workers need to understand and interact with, what you need is it to be readable.

This hardware model will make code very specialized, which is not completly bad but will be harder to maintain and thus put a lot of strain in developers. Working on a tight schedule this can only be bad.

Most software engineers will tell you a CPU with a good cache mechanism which is transparent to the programmer, will run more efficiently most of the time. For example if your programmers ***** up with their optimization of memory access, basically your whole program will get to a crawl. I remember some big time developer comments about this. With a cache system the programmer can focus on optimizing other areas, now with the cell they would have to optimize for everything, and any minor change in code can lead to disartrous cosecuences performace wise.

So maybe you like a programmers brain, (and in an ideal world I would bet I too!) but in real life what we need is to make the right compromises, and in MHO they don't seem to have made the right one with regards to cell when it comes to development time, dev. cost and mantainability of code.

That said I am no expert in cell assembly code, and as this is beyond3d, any enlightment is gladly welcome.

PS: Have you ever wonder why java is so popular among programmers?

EDIT: Damn wasn't trying to go by the word filter, i really must stop cursing lol...

 

[scooby_dooby]

If that's the criteria around here, then fine. I don't mind waiting, and somewhat confused why this criteria is not more emphasized around here, as there are many posts comparing the two systems.

As matter of fact, Asher's comments in another thread was claimed to know what he was talking about in a comparison between the two systems, even by you Scooby Dooby, and I was told I did not know what I was talking about. Remember, you did not say the comparison was not valid, but said


from here:
http://www.beyond3d.com/forum/showthread.php?t=25719&page=7&highlight=asher

Should it not have been brought to Asher attention that he cannot make a comparison at this point?

That entire thread was titled " Does Cell Have Any Other Advantages Over XCPU Other Than FLOPS?"

I don't see any strongly worded objections by you Scooby Dooby that the thread's discussion was invalid?

What am I missing here?

Well simply put, I think CELL has recieved more than it fair share of "optimistic" discussion. I think it's much more interesting at this point to discuss the downsides, not for any stupid console war crap, but just for interests sake, and because the positive's of CELL have been discussed to death.

That's why I encouraged Asher's posts, finally another side of the debate from someone with solid technical knowledge on the matter. You can call it cheerleading or whatever, but I'm interested in cool technical discussions that aren't one-sided, I'll be buy a PS3 that's certain, so I have no vested interest other than a preference for xbox franchises.

And my only problem with your posting in that thread, was the 10-20% numbers you were throwing around were presumably based on P4 hyper-threading, and you were referring to these percentages as if they were set in stone without admitting that the numbers may be drastically different on an IO PowerPC with a different implementation of SMT. That's all.

Also, I'll point out that in the other thread nobody was claiming the X console was undeniably more powerful than Y console, it was simply a discussion about the upsides and downsides of a CPU.

- And for the record, Sony's in no position to really worry about "losing it's grip" x360 may take a good chunk of the market, but sony has nothing to worry about, it'll just be more even this time around.

 

[Edge]

Damn edge, I am not sure how to reply to you, don't want to seem rude.

That little statement seems to suggest you have no clue about software engineering. When you write a program that 20 or more co-workers need to understand and interact with, what you need is it to be readable.

This hardware model will make code very specialized, which is not completly bad but will be harder to maintain and thus put a lot of strain in developers. Working on a tight schedule this can only be bad.

Most software engineers will tell you a CPU with a good cache mechanism which is transparent to the programmer, will run more efficiently most of the time. For example if your programmers ***** up with their optimization of memory access, basically your whole program will get to a crawl. I remember some big time developer comments about this. With a cache system the programmer can focus on optimizing other areas, now with the cell they would have to optimize for everything, and any minor change in code can lead to disartrous cosecuences performace wise.

So maybe you like a programmers brain, (and in an ideal world I would bet I too!) but in real life what we need is to make the right compromises, and in MHO they don't seem to have made the right one with regards to cell when it comes to development time, dev. cost and mantainability of code.

That said I am no expert in cell assembly code, and as this is beyond3d, any enlightment is gladly welcome.

PS: Have you ever wonder why java is so popular among programmers?

EDIT: Damn wasn't trying to go by the word filter, i really must stop cursing lol...

This is getting absurd. I have you telling me I have "have no clue about software engineering", and then you follow through by saying "That said I am no expert in cell assembly code."

Yes, thank you for your contribution to this thread. :major rolleyes:

Next time, drop the rude tone, if you want to have a civil discussion. I won't reply to you again.

 

[Edge]

Well simply put, I think CELL has recieved more than it fair share of "optimistic" discussion. I think it's much more interesting at this point to discuss the downsides, not for any stupid console war crap, but just for interests sake, and because the positive's of CELL have been discussed to death.

That's why I encouraged Asher's posts, finally another side of the debate from someone with solid technical knowledge on the matter. You can call it cheerleading or whatever, but I'm interested in cool technical discussions that aren't one-sided, I'll be buy a PS3 that's certain, so I have no vested interest other than a preference for xbox franchises.

And my only problem with your posting in that thread, was the 10-20% numbers you were throwing around were presumably based on P4 hyper-threading, and you were referring to these percentages as if they were set in stone without admitting that the numbers may be drastically different on an IO PowerPC with a different implementation of SMT. That's all.

Also, I'll point out that in the other thread nobody was claiming the X console was undeniably more powerful than Y console, it was simply a discussion about the upsides and downsides of a CPU.

You're not being consistant, and your excusing your behaviour. It's either allowed or it's not, and going by the standard here, it's allowed. I don't think you should be trying to stop discussion, just because you don't like what you are hearing. I brought up some valid points about the performance difference that can be discussed for everyone's education.

 

[Edge]

Compared to the number of games released on the PS2, skilled developers are an absolute minority. And even if you have a lot of very good programmers, the amount of stuff that they can do in a reasonable timeframe still might not be enough. All that power from having low-level access might get left unused if there's no time to implement it properly, and what seemed to be an advantage might end up an obstacle.
And FYI, most of the stuff that could differentiate this next gen from the previous one are high-level things, which will be even more difficult to implement if the coders have to fight the low-level stuff...

Edit: for example, let's assume that we have a coder whose mad skillz can give us about 10 times as much processing power to spend on AI, so we can have 10 times as many enemies compared to the previous gen. Or, we have a coder who can get 5 times the processing power but he can use it to create the same number of AI enemies, but 5 times as clever. Which one would be better? Which would you consider "next-gen gameplay"?

There are lots of great games being released now on the PS2 that could not be done years ago. I should know, I'm buying those games. These games are coming from the developers that matter, the developers that help sell systems.

 

[Barbarian]

This hardware model will make code very specialized, which is not completly bad but will be harder to maintain and thus put a lot of strain in developers. Working on a tight schedule this can only be bad.

I used to think differently, but these days I have to agree. Skilled programmers are a minority and when you have a team of 50 programmers, stability issues dominate the time and resources of skilled programmers.
For whatever reason no publisher will believe you that you can make a next gen project with less than 30 programmers. Most new programmers come straight from college or with minimum experience. They get no ramp up and no training. So no surpise, the code base quickly becomes a mess.
The schedules don't get any shorter. Games are still expected to be done in 12 months or so, and most importantly - ready for Christmas.
With this kind of pressure, there is no time for hard-core optimizations. You need to get things done fast and stable.
And as much as I would hate to admit it Xbox360 dev environment is just light years ahead of PS3 (after all MS IS in the software business). Sony has been getting away with their arrogance in software support (or lack thereof) so far due to legions of dedicated and hardworking engineers, but now there are other options ...
And have in mind I enjoy coding for PS2 (and probably will for PS3).

 

[SubD]

And as much as I would hate to admit it Xbox360 dev environment is just light years ahead of PS3 (after all MS IS in the software business). Sony has been getting away with their arrogance in software support (or lack thereof) so far due to legions of dedicated and hardworking engineers, but now there are other options ...
And have in mind I enjoy coding for PS2 (and probably will for PS3).

What a joke.

So please go right ahead and list the projects you are referring to. Clearly someone with your experience should be able to demonstrate that:

1) PS2 projects have longer development schedules than similar Xbox projects

2) Or PS2 projects require larger development teams than similar Xbox projects

3) List specific examples of development issues that caused 1) or 2)

I have worked on and run many console projects over the years. I know vast numbers of people who work on and run console projects throughout the industry. I have personal knowledge of the team sizes, budgets, and development schedules for multiple multi-platform console products. And I have a rough estimates for a large number of other projects from a large number of companies that friends of mine work at.

PS2, GameCube, and Xbox projects have, for the vast, vast majority of case, virtually identical team sizes, schedules, and budgets for similar types of projects.

 

[ERP]

PS2, GameCube, and Xbox projects have, for the vast, vast majority of case, virtually identical team sizes, schedules, and budgets for similar types of projects.

Sure but that's more a function of similar budgets = similar number of people x similar amount of time. Quality and content get scaled to maintain the sum.

Most cross platform games use the primary SKU model where you really develop for one target and have small numbers of people developing and maintaining the other SKU's.

Outside of graphics, working on PS2 isn't radically harder than working on Xbox, but frankly the Xbox tools are a lot better and I'd rather debug gameplay code there than on PS2.

The current PS3 tools leave a lot to be desired, I don't think Sony will ever reach parity with MS on this front. I just don't believe they are willing to make the investment it would require, MS can only do it because they are leveraging years of work on PC dev environments. Honestly the toolside could hurt Sony in the medium term, but primary SKU's are decided by bean counters and projected installed bases, not by software engineers so I doubt it will make a real difference.

And FWIW if you have 50+ programmers on a project (my current gen project does), and your primary goal isn't readability and maintainability you are in for a world of hurt.

As team sizes increase you have to accept that the average quailty and experience level of the programmers on a team decreases, there simply aren't that many good experienced game developers out there.

 

[Ty]

And FWIW if you have 50+ programmers on a project (my current gen project does), and your primary goal isn't readability and maintainability you are in for a world of hurt.

50 programmers? I work on MMOs and we don't even throw that many on a project! You work for EA?