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

Optimisation being a relative term when it comes to middleware.....

 

Both Cell and Xenon are already outclassed in terms of raw performance potential in games by AMD and Intel Processors. I'd bet huge bucks the up and coming conroe would make a mockery of Cell or Xenon if it recieved the same TLC such as time and devotion, by the game developers. You have to also remember these designs are made to scale very well in terms of price of production as well. They will cut performance to save money, little things like disabling a unit or keeping it below its rated speeds are perfect examples of this.

When it comes to a gameplay perspective, the thing people fail to completely grasp is that there is very little effects/performance that you can experiance on an Xbox360 that you cant experiance on a PS3 or vice versa. Basically i'm saying you're not going to sit there drooling at all cell is doing knowing consciously its something unique to the Cell processor. Especially if you're use to high end PC hardware/performance. I am shocked, and i mean shocked at the PR, the myth, the legend, Cell has made. They couldnt of done better.

 

Of course, but that applies for the other system too When I look at the presentations on the AGEIA port from GDC, the compromises are clear, but you still have the likes of Epic raving about it (and yes, even explicitly indicating that it offers more than on other systems). It only makes it more exciting to think of future improvements to that middleware and/or "roll-your-own" engines from the best devs.

*It is disappointing though, if middleware companies don't knuckle down and do a good job on each platform. The idea should be that they do the heavy lifting for the benefit of everyone (in exchange for a nice fee, of course ). I guess though, like everyone else, they have a learning curve too.

 

I see. If PS3 has more than adequate resources to deliver good/comparable gaming experiences once optimized, it may indeed make sense for Sony to exploit some of the "extra" resources in other useful home media functions.

My premise is any (successful) company will try to deliver the best consumer experiences and value by leveraging on their strength one way or another.

 

That's not really the issue, a middle ware providers primary goal is satifying as many customers as possible and I don't care how fast it is if it doesn't solve my problem. As a result middle ware is designed with flexibility as its primary requirement, rather than performance, where as purpose built solutions tend to go the other way.

Here's an example - I can use random piece of physics middle ware to provide a generic vehicle model or I can write one that just solves my problem and runs 10x (or more) faster. The question isn't which one is faster, or even which one is the better solution it's can I accept the middleware solution and what doesn't make it into the game if I spend my time writing a vehicle simulation.

 

That of course is huge part of the concept, but I still think middleware providers should strive to tap the systems as best as is possible also. You'd think there might be some incentive also in the presence of competition, even if other factors like support etc. are deemed more important.

 

Yes, I can believe it's a mixture of flexibility and performance in general. In other software industry, companies typically have their own in-house tools/libraries that are more suitable for their own needs; some are actually highly customized 3rd party tools. I'm more refering to these software assets in my original post. The other raw form is of course the talent and experiences gained.

 

Maybe outclassed on certain problems, but on streaming single-precision FLOPs TLP oriented code, current AMD and Intel processes are outclassed and Conroe won't change it. There's no way a Conroe computing a streaming algorithm will outclass 6-7 SPEs on this workload. Conroe only has two SSE units, so unless it runs at 4 times issue rate of the SPEs, it's not gonna touch em.

I'm shocked how much high-end PCs still stuck, because no one is going to tune a game for a high-end PC. As a result, you can get stuff like the MGS4 demo in realtime on a GeForce 7800 devkit, but you can't anything like it on a PC. It's actually a damn shame how underutilized high end CPUs and GPUs are ont the PC.

MintMaster said he best when he brought up literally how many bytes and math ops can be executed *per pixel* by an R580 on a 1600x1200 framebuffer. It's surprising, nah staggering, amount of bandwidth and computation that can be used, but nothing comes close to exercising the power.

 

Yeah I agree, all those processors, all that floating point power, all that integer power, all that huge bandwidth internally and externally is just one big con by IBM/Sony/Toshiba that they spent $400 million on. Who would have thought we could be so stupid to fall for it.

If that's all you got to try and discredit CELL, then you're failing.

 

I really wish B3D would get back to the way it was, without heavily biased users and posts...

 

This is unfortunetly something we will only know towards the end of its life cycle. How much of its potential was actually used. Where did the rest go and what was wasted. Cell has alot of high and impressive peak outputs but ease of implimentation and real world application can easily sing to a different tune so to speak.

Vista will hopefully help here since it should employ minimum running specs high enough to close the huge gap that is used for current game development and optimization which causes the very waste of hardware you speak of. I dont deny consoles put out some very impressive things, but its mainly because devs know exactly what they're coding for, so for the most part the hardware is exploited quite well. When i've been ranting its only on the basis of pure hardware potential making the transision to the real world. And thats where i think the Cell is glorified far above its reality.

 

Its innovation but that doesnt always mean it smells like roses. New technologies are almost always a pain in the ass to impliment and work with. Why should Cell be any different? Cell is fancy, almost too much so, and thats where many see problems including myself. Would a current iteration of AMD or Intel processors have outperformed either Xenon or Cell had they been used? Sure. But it also would of cost alot more over time and it wouldnt of been as sexy to market. Cell has been, for lack of a better word, "pimped" more so then the PS3 itself. You know what i think when i read about Cell? That cool silicone colour tint it has. I dont need to read thousands of lines of propaganda and peak output performance numbers, that may very well never see the light of day in games, to know that no matter the colour of my console box, im going to have an equally great experiance.


side note: word limit per post is too small

 

For video\multimedia and other predictive content no P4 or AMD is even going to come close to Cell. For every day REAL-WORLD general performance including gaming neither the Xenon or Cell can keep up with current high end\upcoming Intel or AMD multicore offerings.

Unfortunetly these CPU's cost more than the entire PS3 w/blue ray together.

 

People are excited over this new chip, and just because it does not meet you approval mean that these people are wrong. It's your arrogant attitude of being "shocked", that makes your post non-constructive, and annoying. It would be equivalent to me saying,

It's not something I would say, or even believe.

 

For you to answer the question "Would a current iteration of AMD or Intel processors have outperformed either Xenon or Cell had they been used?" with "sure", I do not feel you are representing the situation in a complete manner. In the sense of "streaming single-precision FLOPs TLP oriented code" (pointed out by Democoder), it is not "sure". In the same way that you like to generalise with statements such as "new technologies are almost always a pain", i would also say that "fancy" is not necessarily a "problem". Cell is sexy to market, but if you don't like it for that reason, it is certainly your call. Perhaps you do not want to appreciate the fact that Cell is engineered for cost-effective floating point performance, with the bells and whistles of a conventional general processor deliberately left out of the equation (I believe). Whether it works or not we can judge when the hardware arrives. But it seems you already show a kind of indifference that not all of us share.

 

excuse me edge but i'm not the one praising the entire console as being superior with all it will do based on a single component. i have never insinuated that the PS3 as a whole will be inferior or superior to anything.

I stated other reasons why im seriously skeptical, i never said i disliked it because it was fancy. And as much as the quoted FLOP numbers are used, you and i and many others here know that the entire processor will never be dedicated to one application. I already said that the answer to how its exactly used and how well, will come towards the consoles end which seems to be the exact same thing you just said..

 

Well noone says we need to use actual hw stencil writes though.
Not that I can say for sure alternative would be faster - but if you're right, it's possible.

Wouldn't be the first platform where hw stencil was performing worse then "software" alternative (on that note, do hw designers sometimes have no idea why they are putting checkbox features in there, is it really just to cross them off from list supplied by marketting dep., usability be damned?).

 

Haha...design as an afterthought! Who would have imagined?

 

That is a true statement, though it is also arguable that the one application that Cell is dedicated to (if there is only one) is probably the predominant one in gaming. There shouldn't be anything wrong with that.

 

CELL is superior on:

- total amount of internal memory (2304 KB versus 360's 1024 KB: 2.25 times)
- number of processors (8 versus 360's 3: 2.7 times )
- number of *concurrent* threads (8 versus 360's 3: 2.66 times)
- GFLOPS rating (204 GFLOPS versus 360's 77 GFLOPS: 2.65 times)
- integer rating (~51.2 billion instructions per second versus 19.2 BIPS: 2.66 times)
- internal bandwidth (204.8 GB/s versus shared 102 GB/sec L2 cache bandwidth: ~2 times)
- external bandwidth (60 GB/sec versus 360's 22 GB/sec: ~3 times)

On the order of 2 to 3 times for everyone of those specs/features over the Xbox 360 CPU. The only counter argument over all these months is that it will be harder to program, and that is agreed apon. Some of those numbers might be off by a bit, but certainly not by a lot, and no amount of explaination can make them go away.

It's hard to argue with facts, and that's not "myth", "legend" or "PR". But keep it up, everyone loves to hear again how superior CELL is over Xenon.