a) it's a Sony, and those that have bought PSP's or other Sony products the last couple of years do have the right IMO, to feel a bit cheated. I just put that custom FW by Dark_Alex on my PSP and am never going back to another Sony FW. Making people buy a PS3 in order to play PS games on the PS...screw that.
b) it's stupidly expensive
c) no unique games (although I say the same about the 360). Console and PC ports...yawn
Sony comes out with really slick products, but they don't care about their customers at all.
Bad handling is not offset by a more powerful engine. Unless you believe Detroit.
Quote:
Originally Posted by vinea
And PS3 devs say different and the little I've read of IBM's docs (a while ago admittedly but its not like it would have gotten worse in terms of compiler support) imply otherwise.
There's what they say publicly and then there's what they say in private.
Quote:
Originally Posted by vinea
Your analogy is broken as folks have certainly found ways to use a 3.2Ghz SPU to do stuff beyond multiplying numbers together. We can argue about the lack of OOO and rename registers in addition to the lack of branch prediction if you like but the SPUs are reasonably capable little guys.
The SPUs are certainly more capable that the horrendous VU0 and VU1 they supercede. You could even write some general purpose code on them. But the type of code I was describing is nastily multi-modal. I might be wrong, but the idea of offloading game entity logic onto the SPU is not what they were designed to do. The SPUs would have to trawl through data-structures and pointers to answer queries like "What's are the first surface tag along this ray". Getting a vector processor do do this seems bizarre. It might be possible, but three general-purpose processors, to me, seems like a more elegant solution.
Quote:
Originally Posted by vinea
The fact is both Xenon and Cell are nice hardware in comparison to the last gen.
Very true.
And with that... I would like to wrap this up.
I think that I have made a reasonable case to support my assertion. Which is....
The PS3 is simply not a better architecture for creating and running video game content. It is OK. It can run OK games. It might even be able to run amazing games. But its unconventional architecture is a mix of advantages and disadvantages- and in my opinion (and the opionion of many game developers) the benefit of humoungous floating point performance does not overcome the double whammy of a bandwidth choked GPU and a single over-worked CPU.
Does this matter? Not in the slightest. Personally I have stopped playing games and taken up reading books. The graphics are awesome!
As we have seen before, an architecturally weaker machine does not necessarily mean that it will collapse at market. Because consoles, just like many business, are held aloft by the power of confidence.
But confidence has a low boiling point, and can evaporate faster than you can say "Enron."
So I suggest we stop, wait and see. And reconvene in six months or so.
By then we should be able to see...
Whether the PS3's inability to shift games has turned into a problem for Sony.
or whether some badly needed new titles will restore some confidence.
Bad handling is not offset by a more powerful engine. Unless you believe Detroit.
Given that you criticize the RSX as an afterthought because the Cell was expected to handle graphics then stating that the Cell actually CAN do some graphics to offset the fact that the RSX is a traditional GPU and not the u-a GPU in Xenos is a valid rejoinder.
Quote:
There's what they say publicly and then there's what they say in private.
They say the same thing in anonymous industry forums too. You can guess which dev team they are on but that's not the same as knowing who they are. Enough venting seems to occur that if something truly sucked it wouldn't be hidden very long.
Not like Sony is going send ninjas to hunt them down for blabbing...
Quote:
The SPUs are certainly more capable that the horrendous VU0 and VU1 they supercede. You could even write some general purpose code on them. But the type of code I was describing is nastily multi-modal. I might be wrong, but the idea of offloading game entity logic onto the SPU is not what they were designed to do.
The point is that a SPU isn't a VU. The difference between a SPU, vector unit and a CPU is like the difference between C, assembly and Java although that's not a perfect analogy either but conveys the rough usability for general tasks. C isn't a high order language but its not a low level language either. It straddles both worlds. A SPU is similar in that sense. The rest of the analogy would fall apart if you try to take it any further.
Game entity modelling may not be optimal or the intent but there's more to games than individual object modelling and you can still do game entity modelling within certain constraints.
Quote:
The SPUs would have to trawl through data-structures and pointers to answer queries like "What's are the first surface tag along this ray". Getting a vector processor do do this seems bizarre. It might be possible, but three general-purpose processors, to me, seems like a more elegant solution.
The SPUs can do a reasonably decent job IF the objects are in the local store. Trawling through data structures is something I mentioned it does poorly because these live in main memory. However you can break some of these problems down into small subsets...again as I mentioned. The PPE would have to do this decomposition but that's not the same as saying the SPEs can't help do the task.
Whether three PPEs are more elegant or not depends on mindset. In many ways the Xenon three PPE architecture is much like the RSX...good but more conventional. The Cell is much more like Xenos with a new approach. It paid off pretty well for Xenos. I believe that it should pay off for the Cell but it will take longer.
Given that Havok has better SPE support in 4.5 I think that will start to happen as the middleware companies get to know Cell better. A physics company makes sense to be the first to optimize for the SPE but I believe I remember hearing noises that Emergent has added some thing to Gamebryo for the PS3. Floodgate is supposed to be able to handle punting stuff off to the idle SPUs without much programmer interaction.
That implies that Emergent thinks there are a good number of gaming tasks that can be punted over to SPUs...
Quote:
Very true.
And with that... I would like to wrap this up.
I think that I have made a reasonable case to support my assertion. Which is....
And I think I've done a credible job of addressing your technical arguments supporting your rather bald assertion. You haven't really responded on a technical level except to run back to generalities and asserting "insider" knowledge of what PS3 devs "say in private".
Quote:
The PS3 is simply not a better architecture for creating and running video game content. It is OK. It can run OK games. It might even be able to run amazing games. But its unconventional architecture is a mix of advantages and disadvantages- and in my opinion (and the opionion of many game developers) the benefit of humoungous floating point performance does not overcome the double whammy of a bandwidth choked GPU and a single over-worked CPU.
Whether a game is "OK" vs "amazing" has much more to do with the game designer and coders than underlying hardware.
What can be objectively discussed (to some degree...elegance is not one) is the advantages and disadvantages of the architecture. Your repeated assertion that SPEs are nothing but FP vector units doesn't make that a true statement...likewise your other assertions.
Quote:
As we have seen before, an architecturally weaker machine does not necessarily mean that it will collapse at market. Because consoles, just like many business, are held aloft by the power of confidence.
But confidence has a low boiling point, and can evaporate faster than you can say "Enron."
So I suggest we stop, wait and see. And reconvene in six months or so.
By then we should be able to see...
Whether the PS3's inability to shift games has turned into a problem for Sony.
or whether some badly needed new titles will restore some confidence.
C.
Please, making inflammatory remarks and then saying "we should stop here" is just bogus. You haven't addressed any of the technical arguments thus far so it boils down to "Carniphage thinks the PS3 sucks and I'm a game dev so bow down to my opinion".
And if you hate MS and aren't a PS3 dev one does wonder which market segement you address. Because frankly a good number of my former co-workers who moved from traditional programming jobs into game development were doing really lame assed games like Mahjong or other technically meaningless titles a far cry from AAA titles and cutting edge tech (hey hey...but twice the hours at the same pay...).
Yeah, I'm close to Hunt Valley so we do have game jobs here. There's also a good number of former game devs back in the general programming work force given some of the studio closures.
And game devs are not the only devs looking at 3D tech, game middleware and other similar technologies...granted we don't look all that closely at consoles for work...
Ok... You don't want to stop. So Let's plod through these points one at a time...
Quote:
Originally Posted by vinea
Given that you criticize the RSX as an afterthought because the Cell was expected to handle graphics then stating that the Cell actually CAN do some graphics to offset the fact that the RSX is a traditional GPU and not the u-a GPU in Xenos is a valid rejoinder.
No. it. can't. Do some, but it doesn't help. Here's why.
The last phase of the pipe is rasterization. It can't be avoided or bypassed. Rasterization is not the by-product of lazy programming, or an unwillingness to adopt to new thinking (Splinemodel).
At the end of the day your games console is all about drawing triangles into framebuffers. The very act of writing pixels eats into the bandwidth. Zbuffer comparisons do. The rasterizer is sucking in vast texture sets. All eating away at that same bandwidth. And there we have the PS3s achilles heel.
You accept that the bandwidth issue is real? The PS3 has less bandwidth to do this job. So however many zillion polygons you can transform and light, you simply can't draw them.
The rasterization process cannot be assisted in any way whatsoever by the Cell.
Cell can do transformation. Cell could do lighting. Cell could do animation. All lifting some stuff off the GPU. Sure, it makes coding harder and more expensive. It makes game slip. But it is possible.
The cell horse is a runnin' but at the very last hurdle, the horse falls down. Because the Cell can not help with rasterization. It just can't. The 360 can jump over the last fence. The PS3 has to have a special, lower fence.
As the folks in Germany realised some time ago, more horsepower in a car can't improve traction. In fact it can make things worse. The PS3 my friend, is a Corvette.
Next up....game entity logic.
Quote:
Originally Posted by vinea
The SPUs can do a reasonably decent job IF the objects are in the local store. Trawling through data structures is something I mentioned it does poorly because these live in main memory. However you can break some of these problems down into small subsets...again as I mentioned. The PPE would have to do this decomposition but that's not the same as saying the SPEs can't help do the task.
I think you are getting it. That's what I was hoping you'd see.
To help with game logic, the SPUs would have to be spoon-fed subsections of the game database. So that they could operate on local store. OMFG!
In order to farm-out game logic in a way that SPUs could contribute, the CPU would have to do the farming, Imagine a simple shooter where any character can fire a gun anywhere. Each and every game entity needs to make raytrace style queries of the entire game database. It's not something you'd want the SPUs to be doing. Just the act of coding a farmed-out interaction mechanism could add months onto a game's development. These co-processors might be able to run C - but they were not designed to perform this class of task. And if asked to do so would perform it with the elegance of bears on motorcycles.
Could it be done? Well possibly. But I could possibly get to New York in a rowing boat. It just not necessary or desirable or optimal. It would suck and I would get blisters.
Physics on the other hand. Like Havok - Where interactions are mostly local and granular. Yes. Absolutely. The Cell could rock. But physics is not game entity logic. The type of code I am talking about is a lot more like an OS running lots of threads each doing different stuff. But each possibly needing access to every other part.
Quote:
Originally Posted by vinea
Whether a game is "OK" vs "amazing" has much more to do with the game designer and coders than underlying hardware.
Well quite. Which is probably why I paid for a Wii and own neither a 360 or a PS3.
But isn't the reason Sony is charging so much for the PS3 has to do with its claims about the architecture? Sony is telling the world that to have the better gaming orgasms, we must invest in the worlds ultimate gaming hardware?
That's the boldest statment of all, and in the absence of any visible proof, I happen to think it is worth challenging. I dig the whole skeptical bit.
Quote:
Originally Posted by vinea
And if you hate MS and aren't a PS3 dev one does wonder which market segement you address.
I have moved into a that part of the games market segment which is sexy, has lots of growth and much less risk than the economically insane programming bit.
a) it's a Sony, and those that have bought PSP's or other Sony products the last couple of years do have the right IMO, to feel a bit cheated. I just put that custom FW by Dark_Alex on my PSP and am never going back to another Sony FW. Making people buy a PS3 in order to play PS games on the PS...screw that.
b) it's stupidly expensive
c) no unique games (although I say the same about the 360). Console and PC ports...yawn
Sony comes out with really slick products, but they don't care about their customers at all.
No. it. can't. Do some, but it doesn't help. Here's why.
The last phase of the pipe is rasterization. It can't be avoided or bypassed. Rasterization is not the by-product of lazy programming, or an unwillingness to adopt to new thinking (Splinemodel).
At the end of the day your games console is all about drawing triangles into framebuffers. The very act of writing pixels eats into the bandwidth. Zbuffer comparisons do. The rasterizer is sucking in vast texture sets. All eating away at that same bandwidth. And there we have the PS3s achilles heel.
Hmmm...but if you do your backface or occlusion culling on the SPE you're pushing less across to the RSX (which can also do culling) to save bw.
There are enough examples of where the SPE's help the RSX on the various dev boards and the specific b/w gobbled by the rasterizer depends on different factors. MDX is what? 4 levels of abstractions from the GPU? So my personal knowledge of the rasterization stage and optimizing b/w at that level is minimal (from a "we did this" standpoint).
But the general assertion that the SPE's can reduce the b/w load to the RSX appears false even if it doesn't help in the rasterization stage. Especially since Eggerbrecht (Lair) has said that the RSX is fine for 1080 as long as the Cell does geometry and culling.
Quote:
As the folks in Germany realised some time ago, more horsepower in a car can't improve traction. In fact it can make things worse. The PS3 my friend, is a Corvette.
Your analogy but in the motortrend tests the 2006 Z06 beat the 2005 911 on the Mazda Raceway at Laguna Seca by 2.7 seconds and tied in the slalom. It was of course no contest in the drag race...
Quote:
Next up....game entity logic.
I think you are getting it. That's what I was hoping you'd see.
To help with game logic, the SPUs would have to be spoon-fed subsections of the game database. So that they could operate on local store. OMFG!
In order to farm-out game logic in a way that SPUs could contribute, the CPU would have to do the farming, Imagine a simple shooter where any character can fire a gun anywhere. Each and every game entity needs to make raytrace style queries of the entire game database. It's not something you'd want the SPUs to be doing. Just the act of coding a farmed-out interaction mechanism could add months onto a game's development. These co-processors might be able to run C - but they were not designed to perform this class of task. And if asked to do so would perform it with the elegance of bears on motorcycles.
Could it be done? Well possibly. But I could possibly get to New York in a rowing boat. It just not necessary or desirable or optimal. It would suck and I would get blisters.
There's setup but its something you likely want to do for entity AI for better parallelism whether programming for SPEs or 6 hw threads on the 360. Doing LOS to cull the objects of interest for the AI as part of setup is a job you need to do at some point anyway. Once done you can pass that info in as part of the setup to the SPU.
In any case, Eggebrecht said they were even doing troop AI on the SPUs in Lair. Of course that's probably not much more than "if (being_eaten) scream_a_lot();" but its being done. Its not as stupid as you make it out to be even if sub-optimal for the SPU design. Worst case is that you have the equivalent of a 286 running at 3.2Ghz with 128*4 registers and 256KB local store doing your AI work with a 50% hit rate on branch prediction with an average 9 cycle penalty.
And that's without playing games with branches as I outlined above which you haven't addressed. And while the local store is small if you know what you need to fetch out of main memory the SPUs can do that reasonably quickly with a gather scatter list DMA.
And the insomniac folks have used 2 SPE's for collision detection which might involve ray-casting but probably not. Still branchy code IIRC.
Quote:
I have moved into a that part of the games market segment which is sexy, has lots of growth and much less risk than the economically insane programming bit.
Meaning we have about the same level of current coding expertise at the insane programming bit....aka minimal.
Going back to the title of this thread, what has happened to the PS3? In five trips to area BBs and CCs over a period of four weeks, I've only managed to find one PS3 on the shelves. It appears, at least in my locality, that Sony is still unable to keep up with demand. Since some self-styled experts in this thread have pontified that the PS3 is a poor game playing machine, the explanation must be that all the home theatre people are buying them up for use as a Blu-Ray player.
Going back to the title of this thread, what has happened to the PS3? In five trips to area BBs and CCs over a period of four weeks, I've only managed to find one PS3 on the shelves. It appears, at least in my locality, that Sony is still unable to keep up with demand. Since some self-styled experts in this thread have pontified that the PS3 is a poor game playing machine, the explanation must be that all the home theatre people are buying them up for use as a Blu-Ray player.
Look around or order online. They're around and Sony seems to be able to keep up with demand. They're shipping a lot more than folks expected them to given how constrained the launch numbers were.
So, is it too early to say that the PS3 == the new Betamax?
Slightly superior hardware, with too high a price point and no overpowering reason (ie, no must play games) to prefer it over the competition?
Yes. The 360 didn't get its first must-play game until a year after it came out, yet you expect the PS3 to launch with one? Have you paid attention to any system launch, ever? The first wave of titles is never good. Developers are still making more money putting out games for last generation, their projects for this generation just started ramping up maybe a year or two ago.
Another important factor is that the PS3 has been outselling XBOX 360 by a fair margin since its launch, despite the success of Gears of War.
So, is it too early to say that the PS3 == the new Betamax?
People get this mixed up. Betamax came out first, and was later overcome by VHS, which appeared on the scene later. Today, the Xbox came out first, and is now being outsold by the PS3. Which one is today's Betamax?
Sony was over confident and short sighted with Betamax, but learned from the experience. Sony isn't about to repeat those mistakes, and hasn't from what I can tell.
Though you say the PS3 priced too high and has no compelling reason to prefer it over the Xbox, it outsells the competition. Folks must think it's worth it.
Sony hasn't learned much. It still is trying to impose new formats and impose restrictions on its customers in order to dominate the market.
Well, Sony learned at least three things. Develop the best technology, even if it means your competitor introduces a product first. Next, let other companies get in on the action, making and selling players and discs. And last but not least, garner the most support possible.
Personally, I believe that the PS3 is to the Playstation brand what The Phantom Menace was to Star Wars.
Sony's three lessons are.
1) Introduce at least 3 proprietary formats (owned by Sony) each year. Half of them will be picked up by consumers and then we can lock 'em in and ream them. Hai!
2) Hardware Engineers know more about software than programmers. In fact they know more about products than consumers. In fact, that's a very handsome shirt you are wearing, engineer san. Let me rest my head in your lap.
3) If things look bad - lie, lie and lie again. Journalists are happy to repeat the lies if we pay them enough. Loyal customers will believe us. And if enough people belive us, it does not matter! Hey Journalist-san, come to party with 50-centu. Nice girls.
Not that I was planning to buy a PS3 but it is certainly easier to add a new type of controller than improve the graphics in a console. Unless Nintendo has a patent lock on every way to add motion controllers so no one can replicate the remote.
That seems unlikely given some of the the other VR prior art and InterLink's patent. Wii's IR sensor thingy may be one of the cheapest methods but others should be available.
I'm buying a 360...not because of hardware but because of Mass Effect...a 360 exclusive. If I were a fan of some PS3 exclusive I'd likely buy that instead but I like BioWare RPGs...tho' my platform of choice is the PC for RPGs.
Heck...which PS3 exclusives are still left? Either way...Wii isn't even in the running for me. There are no titles I care that much about on the Wii platform except Zelda and only because I played Zelda waaaaay back when...and Wii wont scale with the 360 and PS3 when the other two platforms get better in 2007/2008 as devs more fully exploit the hardware.
Wii is not getting anything close to Gears of War, Resistance:Fall of Man, etc and these are early titles that will be eclipsed in the future. And that's because its a souped up GameCube.
Vinea
I got to go to the PS3 launch party in Toronto, Canada... I played Resistance: Fall of Man. I wasn't all that impressed. I didn't like it that much. It definitely doesn't make me want to buy a PS3. I didn't see any games at the launch that made me want to go out and buy a PS3. I have a feeling that Gears of War is probably a better game (i haven't played it though).
As far as exclusives go, I think that the PS3 is losing a lot of previously Playstation-exclusives. I guess that games like 'Ratchet and Clank' and 'Jak & Daxter' will still be PS-only, but games like the Grand Theft Auto-series are going multi-platform.
The only series that I see as being able to carry Sony would be maybe Final Fantasy, Metal Gear Solid, or Gran Turismo. Even then, I'm not sure if any of those are going XBox360 as well.
The only series that I see as being able to carry Sony would be maybe Final Fantasy, Metal Gear Solid, or Gran Turismo. Even then, I'm not sure if any of those are going XBox360 as well.
In case you were curious, those are all Japanese games.
PlayStation3 was/is waaaaaay overpriced. Only an "idiot in a hurry" would shell out $600 for a gaming system (even with Blu-Ray, which it looks like is sputtering in terms of sales). Wii is a more reasonable sell to families and the average consumer. Most people would probably spend the $$$ on an iPod anyway.
Comments
a) it's a Sony, and those that have bought PSP's or other Sony products the last couple of years do have the right IMO, to feel a bit cheated. I just put that custom FW by Dark_Alex on my PSP and am never going back to another Sony FW. Making people buy a PS3 in order to play PS games on the PS...screw that.
b) it's stupidly expensive
c) no unique games (although I say the same about the 360). Console and PC ports...yawn
Sony comes out with really slick products, but they don't care about their customers at all.
Sure. Offset by the Cell.
Bad handling is not offset by a more powerful engine. Unless you believe Detroit.
And PS3 devs say different and the little I've read of IBM's docs (a while ago admittedly but its not like it would have gotten worse in terms of compiler support) imply otherwise.
There's what they say publicly and then there's what they say in private.
Your analogy is broken as folks have certainly found ways to use a 3.2Ghz SPU to do stuff beyond multiplying numbers together. We can argue about the lack of OOO and rename registers in addition to the lack of branch prediction if you like but the SPUs are reasonably capable little guys.
The SPUs are certainly more capable that the horrendous VU0 and VU1 they supercede. You could even write some general purpose code on them. But the type of code I was describing is nastily multi-modal. I might be wrong, but the idea of offloading game entity logic onto the SPU is not what they were designed to do. The SPUs would have to trawl through data-structures and pointers to answer queries like "What's are the first surface tag along this ray". Getting a vector processor do do this seems bizarre. It might be possible, but three general-purpose processors, to me, seems like a more elegant solution.
The fact is both Xenon and Cell are nice hardware in comparison to the last gen.
Very true.
And with that... I would like to wrap this up.
I think that I have made a reasonable case to support my assertion. Which is....
The PS3 is simply not a better architecture for creating and running video game content. It is OK. It can run OK games. It might even be able to run amazing games. But its unconventional architecture is a mix of advantages and disadvantages- and in my opinion (and the opionion of many game developers) the benefit of humoungous floating point performance does not overcome the double whammy of a bandwidth choked GPU and a single over-worked CPU.
Does this matter? Not in the slightest. Personally I have stopped playing games and taken up reading books. The graphics are awesome!
As we have seen before, an architecturally weaker machine does not necessarily mean that it will collapse at market. Because consoles, just like many business, are held aloft by the power of confidence.
But confidence has a low boiling point, and can evaporate faster than you can say "Enron."
So I suggest we stop, wait and see. And reconvene in six months or so.
By then we should be able to see...
Whether the PS3's inability to shift games has turned into a problem for Sony.
or whether some badly needed new titles will restore some confidence.
C.
Bad handling is not offset by a more powerful engine. Unless you believe Detroit.
Given that you criticize the RSX as an afterthought because the Cell was expected to handle graphics then stating that the Cell actually CAN do some graphics to offset the fact that the RSX is a traditional GPU and not the u-a GPU in Xenos is a valid rejoinder.
There's what they say publicly and then there's what they say in private.
They say the same thing in anonymous industry forums too. You can guess which dev team they are on but that's not the same as knowing who they are. Enough venting seems to occur that if something truly sucked it wouldn't be hidden very long.
Not like Sony is going send ninjas to hunt them down for blabbing...
The SPUs are certainly more capable that the horrendous VU0 and VU1 they supercede. You could even write some general purpose code on them. But the type of code I was describing is nastily multi-modal. I might be wrong, but the idea of offloading game entity logic onto the SPU is not what they were designed to do.
The point is that a SPU isn't a VU. The difference between a SPU, vector unit and a CPU is like the difference between C, assembly and Java although that's not a perfect analogy either but conveys the rough usability for general tasks. C isn't a high order language but its not a low level language either. It straddles both worlds. A SPU is similar in that sense. The rest of the analogy would fall apart if you try to take it any further.
Game entity modelling may not be optimal or the intent but there's more to games than individual object modelling and you can still do game entity modelling within certain constraints.
The SPUs would have to trawl through data-structures and pointers to answer queries like "What's are the first surface tag along this ray". Getting a vector processor do do this seems bizarre. It might be possible, but three general-purpose processors, to me, seems like a more elegant solution.
The SPUs can do a reasonably decent job IF the objects are in the local store. Trawling through data structures is something I mentioned it does poorly because these live in main memory. However you can break some of these problems down into small subsets...again as I mentioned. The PPE would have to do this decomposition but that's not the same as saying the SPEs can't help do the task.
Whether three PPEs are more elegant or not depends on mindset. In many ways the Xenon three PPE architecture is much like the RSX...good but more conventional. The Cell is much more like Xenos with a new approach. It paid off pretty well for Xenos. I believe that it should pay off for the Cell but it will take longer.
Given that Havok has better SPE support in 4.5 I think that will start to happen as the middleware companies get to know Cell better. A physics company makes sense to be the first to optimize for the SPE but I believe I remember hearing noises that Emergent has added some thing to Gamebryo for the PS3. Floodgate is supposed to be able to handle punting stuff off to the idle SPUs without much programmer interaction.
That implies that Emergent thinks there are a good number of gaming tasks that can be punted over to SPUs...
Very true.
And with that... I would like to wrap this up.
I think that I have made a reasonable case to support my assertion. Which is....
And I think I've done a credible job of addressing your technical arguments supporting your rather bald assertion. You haven't really responded on a technical level except to run back to generalities and asserting "insider" knowledge of what PS3 devs "say in private".
The PS3 is simply not a better architecture for creating and running video game content. It is OK. It can run OK games. It might even be able to run amazing games. But its unconventional architecture is a mix of advantages and disadvantages- and in my opinion (and the opionion of many game developers) the benefit of humoungous floating point performance does not overcome the double whammy of a bandwidth choked GPU and a single over-worked CPU.
Whether a game is "OK" vs "amazing" has much more to do with the game designer and coders than underlying hardware.
What can be objectively discussed (to some degree...elegance is not one) is the advantages and disadvantages of the architecture. Your repeated assertion that SPEs are nothing but FP vector units doesn't make that a true statement...likewise your other assertions.
As we have seen before, an architecturally weaker machine does not necessarily mean that it will collapse at market. Because consoles, just like many business, are held aloft by the power of confidence.
But confidence has a low boiling point, and can evaporate faster than you can say "Enron."
So I suggest we stop, wait and see. And reconvene in six months or so.
By then we should be able to see...
Whether the PS3's inability to shift games has turned into a problem for Sony.
or whether some badly needed new titles will restore some confidence.
C.
Please, making inflammatory remarks and then saying "we should stop here" is just bogus. You haven't addressed any of the technical arguments thus far so it boils down to "Carniphage thinks the PS3 sucks and I'm a game dev so bow down to my opinion".
And if you hate MS and aren't a PS3 dev one does wonder which market segement you address. Because frankly a good number of my former co-workers who moved from traditional programming jobs into game development were doing really lame assed games like Mahjong or other technically meaningless titles a far cry from AAA titles and cutting edge tech (hey hey...but twice the hours at the same pay...).
Yeah, I'm close to Hunt Valley so we do have game jobs here. There's also a good number of former game devs back in the general programming work force given some of the studio closures.
And game devs are not the only devs looking at 3D tech, game middleware and other similar technologies...granted we don't look all that closely at consoles for work...
Vinea
Given that you criticize the RSX as an afterthought because the Cell was expected to handle graphics then stating that the Cell actually CAN do some graphics to offset the fact that the RSX is a traditional GPU and not the u-a GPU in Xenos is a valid rejoinder.
No. it. can't. Do some, but it doesn't help. Here's why.
The last phase of the pipe is rasterization. It can't be avoided or bypassed. Rasterization is not the by-product of lazy programming, or an unwillingness to adopt to new thinking (Splinemodel).
At the end of the day your games console is all about drawing triangles into framebuffers. The very act of writing pixels eats into the bandwidth. Zbuffer comparisons do. The rasterizer is sucking in vast texture sets. All eating away at that same bandwidth. And there we have the PS3s achilles heel.
You accept that the bandwidth issue is real? The PS3 has less bandwidth to do this job. So however many zillion polygons you can transform and light, you simply can't draw them.
The rasterization process cannot be assisted in any way whatsoever by the Cell.
Cell can do transformation. Cell could do lighting. Cell could do animation. All lifting some stuff off the GPU. Sure, it makes coding harder and more expensive. It makes game slip. But it is possible.
The cell horse is a runnin' but at the very last hurdle, the horse falls down. Because the Cell can not help with rasterization. It just can't. The 360 can jump over the last fence. The PS3 has to have a special, lower fence.
As the folks in Germany realised some time ago, more horsepower in a car can't improve traction. In fact it can make things worse. The PS3 my friend, is a Corvette.
Next up....game entity logic.
The SPUs can do a reasonably decent job IF the objects are in the local store. Trawling through data structures is something I mentioned it does poorly because these live in main memory. However you can break some of these problems down into small subsets...again as I mentioned. The PPE would have to do this decomposition but that's not the same as saying the SPEs can't help do the task.
I think you are getting it. That's what I was hoping you'd see.
To help with game logic, the SPUs would have to be spoon-fed subsections of the game database. So that they could operate on local store. OMFG!
In order to farm-out game logic in a way that SPUs could contribute, the CPU would have to do the farming, Imagine a simple shooter where any character can fire a gun anywhere. Each and every game entity needs to make raytrace style queries of the entire game database. It's not something you'd want the SPUs to be doing. Just the act of coding a farmed-out interaction mechanism could add months onto a game's development. These co-processors might be able to run C - but they were not designed to perform this class of task. And if asked to do so would perform it with the elegance of bears on motorcycles.
Could it be done? Well possibly. But I could possibly get to New York in a rowing boat. It just not necessary or desirable or optimal. It would suck and I would get blisters.
Physics on the other hand. Like Havok - Where interactions are mostly local and granular. Yes. Absolutely. The Cell could rock. But physics is not game entity logic. The type of code I am talking about is a lot more like an OS running lots of threads each doing different stuff. But each possibly needing access to every other part.
Whether a game is "OK" vs "amazing" has much more to do with the game designer and coders than underlying hardware.
Well quite. Which is probably why I paid for a Wii and own neither a 360 or a PS3.
But isn't the reason Sony is charging so much for the PS3 has to do with its claims about the architecture? Sony is telling the world that to have the better gaming orgasms, we must invest in the worlds ultimate gaming hardware?
That's the boldest statment of all, and in the absence of any visible proof, I happen to think it is worth challenging. I dig the whole skeptical bit.
And if you hate MS and aren't a PS3 dev one does wonder which market segement you address.
I have moved into a that part of the games market segment which is sexy, has lots of growth and much less risk than the economically insane programming bit.
C.
Why the PS3 is currently doing poorly:
a) it's a Sony, and those that have bought PSP's or other Sony products the last couple of years do have the right IMO, to feel a bit cheated. I just put that custom FW by Dark_Alex on my PSP and am never going back to another Sony FW. Making people buy a PS3 in order to play PS games on the PS...screw that.
b) it's stupidly expensive
c) no unique games (although I say the same about the 360). Console and PC ports...yawn
Sony comes out with really slick products, but they don't care about their customers at all.
Thanks, genius.
No. it. can't. Do some, but it doesn't help. Here's why.
The last phase of the pipe is rasterization. It can't be avoided or bypassed. Rasterization is not the by-product of lazy programming, or an unwillingness to adopt to new thinking (Splinemodel).
At the end of the day your games console is all about drawing triangles into framebuffers. The very act of writing pixels eats into the bandwidth. Zbuffer comparisons do. The rasterizer is sucking in vast texture sets. All eating away at that same bandwidth. And there we have the PS3s achilles heel.
Hmmm...but if you do your backface or occlusion culling on the SPE you're pushing less across to the RSX (which can also do culling) to save bw.
There are enough examples of where the SPE's help the RSX on the various dev boards and the specific b/w gobbled by the rasterizer depends on different factors. MDX is what? 4 levels of abstractions from the GPU? So my personal knowledge of the rasterization stage and optimizing b/w at that level is minimal (from a "we did this" standpoint).
But the general assertion that the SPE's can reduce the b/w load to the RSX appears false even if it doesn't help in the rasterization stage. Especially since Eggerbrecht (Lair) has said that the RSX is fine for 1080 as long as the Cell does geometry and culling.
As the folks in Germany realised some time ago, more horsepower in a car can't improve traction. In fact it can make things worse. The PS3 my friend, is a Corvette.
Your analogy but in the motortrend tests the 2006 Z06 beat the 2005 911 on the Mazda Raceway at Laguna Seca by 2.7 seconds and tied in the slalom. It was of course no contest in the drag race...
Next up....game entity logic.
I think you are getting it. That's what I was hoping you'd see.
To help with game logic, the SPUs would have to be spoon-fed subsections of the game database. So that they could operate on local store. OMFG!
In order to farm-out game logic in a way that SPUs could contribute, the CPU would have to do the farming, Imagine a simple shooter where any character can fire a gun anywhere. Each and every game entity needs to make raytrace style queries of the entire game database. It's not something you'd want the SPUs to be doing. Just the act of coding a farmed-out interaction mechanism could add months onto a game's development. These co-processors might be able to run C - but they were not designed to perform this class of task. And if asked to do so would perform it with the elegance of bears on motorcycles.
Could it be done? Well possibly. But I could possibly get to New York in a rowing boat. It just not necessary or desirable or optimal. It would suck and I would get blisters.
There's setup but its something you likely want to do for entity AI for better parallelism whether programming for SPEs or 6 hw threads on the 360. Doing LOS to cull the objects of interest for the AI as part of setup is a job you need to do at some point anyway. Once done you can pass that info in as part of the setup to the SPU.
In any case, Eggebrecht said they were even doing troop AI on the SPUs in Lair. Of course that's probably not much more than "if (being_eaten) scream_a_lot();" but its being done. Its not as stupid as you make it out to be even if sub-optimal for the SPU design. Worst case is that you have the equivalent of a 286 running at 3.2Ghz with 128*4 registers and 256KB local store doing your AI work with a 50% hit rate on branch prediction with an average 9 cycle penalty.
And that's without playing games with branches as I outlined above which you haven't addressed. And while the local store is small if you know what you need to fetch out of main memory the SPUs can do that reasonably quickly with a gather scatter list DMA.
And the insomniac folks have used 2 SPE's for collision detection which might involve ray-casting but probably not. Still branchy code IIRC.
I have moved into a that part of the games market segment which is sexy, has lots of growth and much less risk than the economically insane programming bit.
Meaning we have about the same level of current coding expertise at the insane programming bit....aka minimal.
Vinea
Going back to the title of this thread, what has happened to the PS3? In five trips to area BBs and CCs over a period of four weeks, I've only managed to find one PS3 on the shelves. It appears, at least in my locality, that Sony is still unable to keep up with demand. Since some self-styled experts in this thread have pontified that the PS3 is a poor game playing machine, the explanation must be that all the home theatre people are buying them up for use as a Blu-Ray player.
Look around or order online. They're around and Sony seems to be able to keep up with demand. They're shipping a lot more than folks expected them to given how constrained the launch numbers were.
Vinea
http://www.youtube.com/watch?v=R98qC0fd_1w
C.
Slightly superior hardware, with too high a price point and no overpowering reason (ie, no must play games) to prefer it over the competition?
So, is it too early to say that the PS3 == the new Betamax?
Slightly superior hardware, with too high a price point and no overpowering reason (ie, no must play games) to prefer it over the competition?
Yes. The 360 didn't get its first must-play game until a year after it came out, yet you expect the PS3 to launch with one? Have you paid attention to any system launch, ever? The first wave of titles is never good. Developers are still making more money putting out games for last generation, their projects for this generation just started ramping up maybe a year or two ago.
Another important factor is that the PS3 has been outselling XBOX 360 by a fair margin since its launch, despite the success of Gears of War.
So, is it too early to say that the PS3 == the new Betamax?
People get this mixed up. Betamax came out first, and was later overcome by VHS, which appeared on the scene later. Today, the Xbox came out first, and is now being outsold by the PS3. Which one is today's Betamax?
Sony was over confident and short sighted with Betamax, but learned from the experience. Sony isn't about to repeat those mistakes, and hasn't from what I can tell.
Though you say the PS3 priced too high and has no compelling reason to prefer it over the Xbox, it outsells the competition. Folks must think it's worth it.
CD, PS1, Walkman are all clear successes.
Betamax, Minidisk and PSP are all failures.
These failures are linked by a common theme.
When companies place their own agenda ahead of customers, the market can be cruel.
C.
Sony hasn't learned much. It still is trying to impose new formats and impose restrictions on its customers in order to dominate the market.
Well, Sony learned at least three things. Develop the best technology, even if it means your competitor introduces a product first. Next, let other companies get in on the action, making and selling players and discs. And last but not least, garner the most support possible.
Sony's three lessons are.
1) Introduce at least 3 proprietary formats (owned by Sony) each year. Half of them will be picked up by consumers and then we can lock 'em in and ream them. Hai!
2) Hardware Engineers know more about software than programmers. In fact they know more about products than consumers. In fact, that's a very handsome shirt you are wearing, engineer san. Let me rest my head in your lap.
3) If things look bad - lie, lie and lie again. Journalists are happy to repeat the lies if we pay them enough. Loyal customers will believe us. And if enough people belive us, it does not matter! Hey Journalist-san, come to party with 50-centu. Nice girls.
C.
Not that I was planning to buy a PS3 but it is certainly easier to add a new type of controller than improve the graphics in a console. Unless Nintendo has a patent lock on every way to add motion controllers so no one can replicate the remote.
That seems unlikely given some of the the other VR prior art and InterLink's patent. Wii's IR sensor thingy may be one of the cheapest methods but others should be available.
I'm buying a 360...not because of hardware but because of Mass Effect...a 360 exclusive. If I were a fan of some PS3 exclusive I'd likely buy that instead but I like BioWare RPGs...tho' my platform of choice is the PC for RPGs.
Heck...which PS3 exclusives are still left? Either way...Wii isn't even in the running for me. There are no titles I care that much about on the Wii platform except Zelda and only because I played Zelda waaaaay back when...and Wii wont scale with the 360 and PS3 when the other two platforms get better in 2007/2008 as devs more fully exploit the hardware.
Wii is not getting anything close to Gears of War, Resistance:Fall of Man, etc and these are early titles that will be eclipsed in the future. And that's because its a souped up GameCube.
Vinea
I got to go to the PS3 launch party in Toronto, Canada... I played Resistance: Fall of Man. I wasn't all that impressed. I didn't like it that much. It definitely doesn't make me want to buy a PS3. I didn't see any games at the launch that made me want to go out and buy a PS3. I have a feeling that Gears of War is probably a better game (i haven't played it though).
As far as exclusives go, I think that the PS3 is losing a lot of previously Playstation-exclusives. I guess that games like 'Ratchet and Clank' and 'Jak & Daxter' will still be PS-only, but games like the Grand Theft Auto-series are going multi-platform.
The only series that I see as being able to carry Sony would be maybe Final Fantasy, Metal Gear Solid, or Gran Turismo. Even then, I'm not sure if any of those are going XBox360 as well.
The only series that I see as being able to carry Sony would be maybe Final Fantasy, Metal Gear Solid, or Gran Turismo. Even then, I'm not sure if any of those are going XBox360 as well.
In case you were curious, those are all Japanese games.
None of them are ever going XBOX 360.
I am pretty certain that it was running at 15fps and using crossfading to hide the fact.
Gran Turismo is much faster, but feels like the PS2 game with up-rezzed graphics.
C.
Gran Turismo is much faster, but feels like the PS2 game with up-rezzed graphics.
C.
As can be expected for many titles this early in the platform lifecycle.
Vinea
PlayStation3 was/is waaaaaay overpriced. Only an "idiot in a hurry"
Yeah this was the first thing that went through my head. PLease see http://www.penny-arcade.com/comic/2007/02/10 for a good chuckle.