Starbreeze take on the Ps3 vs Xbox 360 (the Darkness Int)

[Phil]

I've been dreaming of proper physics since I've first played GT3 or MGS2 on PS2 which was already a very large step forward from previous generation. IMO, I'd love to see a developer try to make a game with perhaps only double the polygon budget but increase physics and add weather-like effects to the game.

Think of Gran Turismo 3's wet-stage and now add MGS2-rain like effects to it and better lighting. IMO, despite the cars not having more polygons, I think the results would be much more outstanding than many other next-generation efforts that simply quadruple polygons, textures and what not...

Just my 2 cents...

 

[expletive]

Indeed. In every argument against Cell's peak values versus XeCPU, I've yet once to hear the same arguments applied to the XeCPU. What if Cell runs at 50% efficiency, and XeCPU runs at 50% efficiency? Cell has the advantage still (in numbers, which can't be directly correlated to useful work performed).

The only way XeCPU won't be behind on these numbers is if the efficiency of the two chips runs from

XeCPU = 100 % : Cell = 65% efficiency
XeCPU = 80% : Cell = 50% efficiency
XeCPU = 65% : Cell = 40% efficiency

Basically, Cell has to be pretty lousy in the efficiency department compared with XeCPU to match it in Flops terms. Anyone wanting to make that argument ought to present some evidence, notably real-world efficiency of XeCPU, especially when they make the point of asking evidence from others to substantiate their opinions.

You also have to factor in the developers ability to utilize 6 SPEs continuously/effectively which appears to be a more arduous task than the 3 cores on Xenon. So its a mixture of the efficiency of the deign and the ease in which that design can be utilized. But yes i agree that the Cell will always win the FLOPS debate, its just a matter of by how much.

 

[Alpha_Spartan]

This is the generation where individual consoles won't stand out, but rather individual technologies.

 

[dukmahsik]

You also have to factor in the developers ability to utilize 6 SPEs continuously/effectively which appears to be a more arduous task than the 3 cores on Xenon. So its a mixture of the efficiency of the deign and the ease in which that design can be utilized. But yes i agree that the Cell will always win the FLOPS debate, its just a matter of by how much.

we keep on talking about how cell is more powerful, okay it has a flop advantage, what about everything else? is a cpu only doing flop calculations? how hard is it for devs to get perf out of each cpu? is xenon easier to get closer to its theoretical limit? it would certainly seem so. which system has better dev tools?

as much as each side likes to do damage control, in the end, ps3 and 360 games will be nearly indistinguishable and that's the cold hard facts.

 

[Lycan]

You can't state it as fact, sorry. We can argue all day long, but only a valid comparison (based on games in motion) would make sense. E3 is 4 days off (or is it 5 ?)...

PS : Looking forward to the evolution of Heavely Sword...

 

[Acert93]

and doesnt the Xecpu get downgraded in real world games and aplications? or does that just count for Cell?

Heh, depends, what side of the fence do you sit on? Even restricted to game code that is float intensive neither is going to hit anywhere their peak in game situations.

XeCPU = 100 % : Cell = 65% efficiency
XeCPU = 80% : Cell = 50% efficiency
XeCPU = 65% : Cell = 40% efficiency

Basically, Cell has to be pretty lousy in the efficiency department compared with XeCPU to match it in Flops terms.

Tasking/threading aside (on both chips), CELL should perform much better in certain tasks due to size advantage and architecture. Of course the annoying thing, in general, is that discussion tends to rotate around one metric of CPU performance. There is a lot more to both game code & the two chips designs that sheer float performance and I think we castrate both chips when we frame the entire discussion in those terms.

My CPU maths and background is not as strong as some of the others here, and maybe it is because this is a 3D site, but it seems a lot of the CPU discussions are very watered down in regards to performance and architecture. I know some dislike Hannable, but as far as I can tell he is one of the few to take a real stab at the architectures (which he has a starting perspective like anyone which doesn't always jive with individual developers or fans or all market realities).

Dissappointed that over 1 year since we got some concrete information on CELL and for the most part discussions still revolve around flops. I know benchmarking is out of the question in regards to real game code for the most part, but G70/71 and Xenos are picked to death architecturally. Of course GPUs are easier in that they do more constrained tasks so there are "easier" answers and CPUs have to deal with a lot of various code -- and done in a variety of ways (not all AI code is created equal for example). I guess my dissappointment is that we will never have any really firm data, much like EE, outside vieled dev comments and a cherry picked tech test/demo of some sort.

Anyhow, every can continue with their discussion about why StarBreeze is right/wrong (IMO the guy was open and candid and his opinion is his experience for their design goals and their assessment of the hardware... the answers may change depending on dev team size and skill, budget, and game design goals).

 

[PSman]

I got a question, when he comparing both of these console, is he comparing "specs on paper" or "real world performance?". I'm asking because, I have been told by some tech expert i know that "on paper Xenos definitely look more powerful, but in real world performance RSX has the advantage". I guess every developers have different opinion, and that their opinion will change as time goes when they're more experience with both of these console

Still he's a developer and i completely respect him:smile:

 

[patsu]

as much as each side likes to do damage control, in the end, ps3 and 360 games will be nearly indistinguishable and that's the cold hard facts.

Still educated guess at this moment. Statement like the above makes me just want to buy 1 console and 1 game in each genre, but the thing is I look to the exclusive franchises and AAA teams to set the benchmark. Art direction will also make a difference.

I'm dying to see Halo 3 as much as Heavenly Sword or any next-gen Eye-Toy game like Eyedentity. Fundamentally, I believe developers will exploit the strength of individual consoles to deliver their visions. It's a question of talent, culture and business. The hardware "merely" enables or simplifies/complicates the process (e.g., hard disk, camera, hi-power GPU, super-CPU, ...).

 

[Shifty Geezer]

My CPU maths and background is not as strong as some of the others here, and maybe it is because this is a 3D site, but it seems a lot of the CPU discussions are very watered down in regards to performance and architecture.

Dissappointed that over 1 year since we got some concrete information on CELL and for the most part discussions still revolve around flops.

I don't think any better conversation can be had out of the info we have. We have substantial documentation on Cell. We have some documentation on XeCPU. Actually working out from the paper docs how these things will work with real code, especially when there's little real code examples to work from, is nigh on impossible. Even with paper specs you can't account for how different approaches to solving problems will work. It may be CPU A has a distinct advantage over CPU B in code type X using then well known algorithm M, but when a console dev gets their hands onto CPU B and finds he can use it a certain way, inventing algorithm N, suddenly it's much better in code type X.

Without that hands on experience and experimenting, the only concrete info is the paper specs. Cell has provided some real world examples, I think all Flops based. I doubt we'll get such examples from XeCPU. At the end of the day, both machines will be trying to run the same type of software, games, and if there is a difference in overall capacity measured in real-world performance, that's where it'll be noticed. This comparison will reflect both power of the hardware and accessibility to the developers, the entire system package as it were. But it'll take a couple of years after PS3's launch to get enough games of suitable quality to start to get that overall picture. The debate is really reserved for next-gen, after this one has been and gone and everyone can look back at the consoles' histories. At which point no-one will care as they'll be bickering about whether XB20000 is better than PS4 or not

 

[ERP]

On paper it would be tempting to pick the chip with eDram for that, but I've also seen arguments that suggest GPUs using external memory can reach near theoretical peaks with zixel fill just the same - which could tip the balance in favour of chip with more ROPs/higher clock.

Unless NVidia has changed their Z compression writing stencil values disables Z compression on the quad (haven't tested this lately).

 

[scooby_dooby]

I'm confident of Cell's ultimate real world efficiency, particularly as time wears on.

Based on what?

 

[rounin]

Based on what?

I would assume based on the fact that there are numerous benchmarks released and that it is actually gaining acceptance. And because there are individuals out there testing it and showing us what they've been able to achieve with it. There have been real world benchmarks where the CELL has indeed reached very close to theoretical max and there have been also theoretical speculations made as to why this is possible. To me, it all adds up to be something that gives a lot more assurance than simple promises from PR on efficiency (ie. Xenon and Xenos) as well as speculations...

 

[scooby_dooby]

I would assume based on the fact that there are numerous benchmarks released and that it is actually gaining acceptance. And because there are individuals out there testing it and showing us what they've been able to achieve with it. There have been real world benchmarks where the CELL has indeed reached very close to theoretical max and there have been also theoretical speculations made as to why this is possible. To me, it all adds up to be something that gives a lot more assurance than simple promises from PR on efficiency (ie. Xenon and Xenos) as well as speculations...

To me it adds up to nothing more than custom made benchmarks and are essentially meaningless.

PS3 does not run benchmarkls, it runs games, IMO, for anyone to be reasonably sure of CELL's efficiency, we need to get some real numbers as to how it performs in a GAME. Otherwise it's not much more than extreme optimism.

 

[Edge]

You also have to factor in the developers ability to utilize 6 SPEs continuously/effectively which appears to be a more arduous task than the 3 cores on Xenon.

Programming for any multi-core architecture when it comes to games is difficult, as look at the PC world were we have had dual-cores for a long time now, and do next to nothing for the frame-rate of games being released today. Today PC dual cores are more akin to Xenon, than CELL, so think about that.

Nice thead by the way, it's flogging a dead horse with all the same things being said before. I'm beginning to wonder if anything new can be said here about the different architectures.

 

[Edge]

PS3 does not run benchmarkls, it runs games, IMO, for anyone to be reasonably sure of CELL's efficiency, we need to get some real numbers as to how it performs in a GAME. Otherwise it's not much more than extreme optimism.

The benchmarks that Rounin is referring to, is the SPE's operating on large matrices, and no question, it will run circles around Xenon in that regard, and can be very useful in games, but the end result I don't think will be console busting defining.

In the end people wanting Gran Turismo on the PS3 are not going to care if it's all that graphically less or more than a 360 game, but that it is the next Gran Turismo.

 

[rounin]

To me it adds up to nothing more than custom made benchmarks and are essentially meaningless.

PS3 does not run benchmarkls, it runs games, IMO, for anyone to be reasonably sure of CELL's efficiency, we need to get some real numbers as to how it performs in a GAME.

For the record, I was just suggesting why I think CELL would be able to achieve close to theoretical max, maybe he has other reasons. In any case, it still is something when there are individual sources and actual benchmarks doing these and providing us with numbers. It seems to me some people have no problem assuming in accepting USA's efficiency and are self-assured that the two platforms are similar suddenly are very specific in demanding when talking about CELL's efficiency even though there are supporting documents and benchmarks for the latter and not the former.

In other words, if these benchmarks and industry acceptance (and further speculation based on these) are essentially meaningless, what of the speculation made on Xenon/Xenos based on PR and cloudy speculation? That is where I think people banking on CELL's efficiency is drawing their relative confidence from.

 

[SugarCoat]

Programming for any multi-core architecture when it comes to games is difficult, as look at the PC world were we have had dual-cores for a long time now, and do next to nothing for the frame-rate of games being released today. Today PC dual cores are more akin to Xenon, than CELL, so think about that.

Nice thead by the way, it's flogging a dead horse with all the same things being said before. I'm beginning to wonder if anything new can be said here about the different architectures.

PC world has had dual core for a long time? It literally just had its 1 year birthday since the introduction of the first economic retail models. Thats no comparison time wise to single core. The problem that keeps happening, why threads keep turning into this, is because the same people keep quibbling how awesome and what a super advantage cell is, when in fact its going to provide no 'large' advantage in games, especially visually.

 

[Titanio]

Based on what?

The reception the chip has had, my knowledge of it, the comments of various developers and the fact that it's going to be in a closed box that will have a lot of developers working on it for 5+ years.

To turn the tables, what makes you think otherwise?

My CPU maths and background is not as strong as some of the others here, and maybe it is because this is a 3D site, but it seems a lot of the CPU discussions are very watered down in regards to performance and architecture.

...

Dissappointed that over 1 year since we got some concrete information on CELL and for the most part discussions still revolve around flops.

I tend to agree on one hand, but on the other, a lot of the architectural ins and outs and design choices of these CPUs has actually been discussed to death here. It's probably for the best that those discussions are not ressurected any time somebody expresses their thoughts on how they compare

To be honest I think the relative flop ratings is a rather conservative indicator of the potential performance difference between the two CPUs, and in many ways it's not doing Cell justice to 'reduce' it to simply that number. But people gravitate to them because they're nice and easy, I'm guilty of it myself.

Well this thread was about Xenos Edge. But it seems that everytime someone talks about Xenos, and the fact that its a sligthly better chip, sony defenders start jumping out from the trees Cell this and Cell that, Cell will Cell is, Omg Cell, Cell, Cell, and because of Cell Xbox 360 will be like to the PS3 what dreamcast is to the Xbox.

Jesus.

Get a grip. This thread was about a comment made by a developer that referenced everything from CPUs to Blu-ray to GPUs.

 

[Edge]

PC world has had dual core for a long time? It literally just had its 1 year birthday since the introduction of the first economic retail models.

That's more than long enough, especially since some 360 games at launch was bragged about, like Activision's Call of Duty 2 to be using three cores. Well, what about their PC game!!!

 

[Kryton]

That's more than long enough, especially since some 360 games at launch was bragged about, like Activision's Call of Duty 2 to be using three cores. Well, what about their PC game!!!

3 words: closed box system.

The PC may have dual core chips but without these being present in *every* PC (and with things like SMT in the mix too) it isn't worth the effort to write decent engines that can expose the full power and which also scale back on single CPU systems. If you can/do invest a lot into it then yes, you can do it but, as this is a new technology only a few people will have a platform that illustrates your work so you get very little 'overall' utility and you'll get moans from those who don't have the latest and are suffering (think how the OS would schedule two 100% CPU reliant threads on a single CPU and the problems your likely to encounter with continual context switches).

With X360/PS3 you have to sink the investment for a new engine anyway so you might as well 'do it properly' because everyone benefits without problems as you aren't considering the lowest common denominator.