360-to-PS3 Ports: Why the Lowered Texture Resolution?

[ShaidarHaran]

I'll say it: they're wrong, or disingenuous. What system you start or lead development on has nothing to do with the quality of the port. The idea that developers will ignore programming practices that are good on the 360 as well as the PS3 just because they lead on the 360 doesn't make sense unless that developer just doesn't have much experience, in which case leading on a system they can't figure out in the first place probably isn't a good thing.

The idea I've seen espoused is that by utilizing best-practices for PS3 programming (i.e. achieving high levels of data locality) transfers over well to any other modern multi-core MPU design.

I agree that something about this just seems "off" though, because it implies that the reverse is also true, but we see that is clearly not the case.

I think the poor ports we've seen thus far have been a result of poor understanding of PS3 programming, and lack of development time and other resources (like actual programmers).

 

[TimothyFarrar]

Are you saying that others claiming that developing for PS3 and porting to 360 allows the most benefits of each platform to be retained (thanks to the best-practices of Cell programming) are wrong? Seems a rather logical statement to me. If you have practical experience demonstrating otherwise though, I'd love to hear it. I personally have none, so I'm merely asking questions at this point.

Sure, develop for the PS3 first would be required to get a good cross platform solution. Can build SPU code with some kind of cross platform intrinsics interface so that DMAs get no-op'ed on 360/PC. Yet instruction latencies (and register numbers, ie PC) are different on the platforms, so you are at the mercy of the compiler doing proper loop unrolling and such for each platform (PC MSVC 2005, doing a horrid job with intrinsics, so have to use something else) to get good performance. Just scratching the surface of those details... but all of this requires a tremendous un-conventional engineer effort (see insomniac's R&D page to get an idea) something which I'm guessing most developers aren't able or willing to do. Hence crappy ports.

As for GPU stuff, there are things you can do on the PS3 which wouldn't work well on 360. Obvious stuff like issues with tiling come to mind. Non-obvious stuff like doing vertex work on SPUs, (or perhaps post processing on SPUs) isn't a good idea on 360. So you would have to design for the lowest common denominator (ie something which works with tiling, and isn't vertex bound on PS3, etc). Likewise there are things you can do on the 360 which simply wouldn't work well on the PS3 (PS3 is DX9 level, 360 has some DX10 level features). There are many details here which I am mostly sure I cannot publicly talk about.

 

[TimothyFarrar]

I don't believe anyone would budget the production of more textures just for the PS3 version of the game. Not today, not in 2009. Maybe after that, if everything goes the best possible way for Sony and the worst possible way for Microsoft.

This could easily happen with virtual texturing (idTech 5?), will certainly happen (ie make use of the blueray) on titles which are PS3 only.

 

[Shifty Geezer]

The idea I've seen espoused is that by utilizing best-practices for PS3 programming (i.e. achieving high levels of data locality) transfers over well to any other modern multi-core MPU design.

I agree that something about this just seems "off" though, because it implies that the reverse is also true, but we see that is clearly not the case.

The public front of developer consensus has been that the XB360 is very forgiving. Thus developers could approach that platform from a Good PC Practice front and get fair results. Then using these techniques on PS3, it goes ppbt! Now if you start on PS3 with a view to getting best performance, you'll use different techniques that are far more optimal, and when you port this thinking back to the other multicore architectures, you see benefits there too.

So leading on PS3 doesn't equate to better XB360 and PC games per se, which seems to be what's getting Chachi's back up. I think the idea is more that starting with a view to getting good performance from PS3 means starting with a new set of best-practices that has benefits, which are being ignored on XB360 and PC because developers haven't really needed to rethink things yet - they've had acceptable results without having to strain themselves. Which makes sense; MS have postiioned XB360 as an easy to develop for platform, and we've been hearing about it for long enough. Developing a game to make money isn't always (or often even?) about maximising the hardware, and if your way works on XB360, why bust a gut reinventing techniques for better results if it won't make much difference to sales? Whereas if you're gonna produce a PS3 version and it'll look sluggish, maybe you better had reconsider how you do things, in which case why not feed that back into XB360 and get better performance?

 

[ShaidarHaran]

Thanks for the info, Tim, and for expounding upon my earlier thoughts, Shifty.

 

[joker454]

As for texturing, there is no reason that you cannot get high quality textures on both platforms if you have a good texture streamer.

Streaming only helps with 'texture variety over location and time'. It doesn't help you when the player stops and just looks at what's currently on screen since that has to fit in available memory. Developers have two choices in that case. Leave the 360's extra memory unused, or cut down some PS3 textures. The easiest way is to just cut back some PS3 textures. The safest way is to just leave the 360's extra memory unused, then all the reviewers will marvel at how the developer 'amazingly' got both versions of the game to look the same.

I'll say it: they're wrong, or disingenuous. What system you start or lead development on has nothing to do with the quality of the port. The idea that developers will ignore programming practices that are good on the 360 as well as the PS3 just because they lead on the 360 doesn't make sense unless that developer just doesn't have much experience, in which case leading on a system they can't figure out in the first place probably isn't a good thing.

What they're really saying, if you read between the lines, is if you lead on the PS3 and port to the 360 then you (probably) won't see a disparity between the two versions that cause PS3 owners to get bent out of shape and start ranting about lazy developers.

You are my new best friend.

 

[MazingerDUDE]

The safest way is to just leave the 360's extra memory unused, then all the reviewers will marvel at how the developer 'amazingly' got both versions of the game to look the same.

I doubt those reviewers would marvel games like Iron Man no matter how same they are on both consoles

Actually, those games that look the same on both consoles, are also the ones with the best graphics in this gen, such as COD4, DMC4, Burnout etc

I think they deserve even more praise for creating best looking games with extra memory unused

 

[TimothyFarrar]

Streaming only helps with 'texture variety over location and time'. It doesn't help you when the player stops and just looks at what's currently on screen since that has to fit in available memory. Developers have two choices in that case. Leave the 360's extra memory unused, or cut down some PS3 textures. The easiest way is to just cut back some PS3 textures. The safest way is to just leave the 360's extra memory unused, then all the reviewers will marvel at how the developer 'amazingly' got both versions of the game to look the same.

With only 1MPix at 720P if you need something on the order of over PS3 GPU memory for texture streaming, you are doing something very very wrong. Effectively at say a low 150 MB you have a little under (alignment requirements) 150 texels per pixel on the screen (with compressed textures). Which is more than enough to cover what is on the screen, what is outside of view, and what you are moving towards.

 

[djskribbles]

I doubt those reviewers would marvel games like Iron Man no matter how same they are on both consoles

Actually, those games that look the same on both consoles, are also the ones with the best graphics in this gen, such as COD4, DMC4, Burnout etc

I think they deserve even more praise for creating best looking games with extra memory unused

yeah, its usually the not-so-good looking games that look noticeably worse on PS3, probably because they were ported to the PS3. whereas other games like CoD4, GTA4, and Burnout essentially had separate teams (not sure about DMC4, but i think the PS3 was the "lead") and all turned out very similar or virtually identical.

 

[assen]

Don't ever, for one second, believe the "separate teams" bullshit. No one is throwing a whole separate team on a platform SKU. The art tools, the gameplay code, the gameplay tools, the editor, the content, the missions are all done by the same people in a platform-neutral way. From there on, there are separate people for each, for the low level system stuff (streaming, threading) and for the renderer. At one extreme, you might have something like what Ubisoft said about Assassin's Creed, 13 people dedicated to platform-specific stuff for the PS3 version; at the other extreme, you might have someone boot the PS3 devkit once a week and beat up what's broken with a lead pipe. There is a continuum between these two extremes, and multiplatform studios are somewhere in-between. But the talk of "completely separate teams" and "parallel development" is PR.

 

[Fafalada]

It doesn't help you when the player stops and just looks at what's currently on screen since that has to fit in available memory

Depends on how well your texture storage data-structures can map to what's actually visible on screen at any given time. Eg. - ideal storage would only need a tiny fraction of 512MB for 1:1 coverage of 720P screen at any given frame.

Mind you I'm not arguing it's easy to exploit temporal coherency to the extremes, but we're at the point where it's more a SW engineering problem then HW resource one.

 

[_phil_]

Rage megatexturing and uncharted are good examples of that.

 

[grandmaster]

Assen is quite correct. The 'separate teams' spin holds no water. It simply makes no financial sense. Burnout Paradise for sure just had one - extremely talented - cross-platform development team that worked on 360 and PlayStation 3 simultaneously.

 

[rekator]

But on the marketing side the "separate teams" is very important.

 

[grandmaster]

At one extreme, you might have something like what Ubisoft said about Assassin's Creed, 13 people dedicated to platform-specific stuff for the PS3 version;

In this case then, the phrase 'unlucky 13' definitely springs to mind...

 

[RobertR1]

I'll say it: they're wrong, or disingenuous. What system you start or lead development on has nothing to do with the quality of the port. The idea that developers will ignore programming practices that are good on the 360 as well as the PS3 just because they lead on the 360 doesn't make sense unless that developer just doesn't have much experience, in which case leading on a system they can't figure out in the first place probably isn't a good thing.

What they're really saying, if you read between the lines, is if you lead on the PS3 and port to the 360 then you (probably) won't see a disparity between the two versions that cause PS3 owners to get bent out of shape and start ranting about lazy developers.

Oh well... sore point. I think it's better to have a high bar set and try to match that rather than aim for the lowest common denominator. I'd also like to see developers play to the strength of the particular consoles (e.g. PS3 is supposed to be much more capable at physics, etc.) but business realities apparently prevent this.

Reward the platforms the best you can. If the 360 due to it's easier programming and perhaps a better GPU allows for certain gains, use them. If the PS3 with the Cell, allows for better physics and other things, use that! Ofcourse, don't blow the budget doing so but don't intentionally cripple any verison for the sake of parity. The whole equality thing would make sense if the people would be buying a version for each console due to the "equality" but that's not the case.

Shame on devs for having this mindest or of the higher ups pushing this down their throats.

 

[AlphaWolf]

Reward the platforms the best you can. If the 360 due to it's easier programming and perhaps a better GPU allows for certain gains, use them. If the PS3 with the Cell, allows for better physics and other things, use that! Ofcourse, don't blow the budget doing so but don't intentionally cripple any verison for the sake of parity. The whole equality thing would make sense if the people would be buying a version for each console due to the "equality" but that's not the case.

Shame on devs for having this mindest or of the higher ups pushing this down their throats.

While I agree with you in principle I can certainly see the mindset. Any perceived deficiency from one version to another is magnified 100000x by websites looking for hits and forum yahoos with an axe to grind.

 

[obonicus]

While I agree with you in principle I can certainly see the mindset. Any perceived deficiency from one version to another is magnified 100000x by websites looking for hits and forum yahoos with an axe to grind.

You don't even need deficiencies. I've seen Kotaku trolls posting that because Burnout Paradise is the same on both consoles, the 360 version was gimped.

 

[chachi]

You don't even need deficiencies. I've seen Kotaku trolls posting that because Burnout Paradise is the same on both consoles, the 360 version was gimped.

That probably wouldn't have happened if Alex Ward wasn't such a huge Sony fanboy and went on and on how they were a Sony oriented studio, and then a patch came out later that supposedly fixed the 360 framerate issues and seemingly confirmed the lack of attention to that version.

Anyway I can see developers point in going for parity. I'd prefer if they played to the strength of each console but nobody likes bad press and when you're developing big budget games for console #2 and #3 I guess you don't have leeway to spend quite so freely and it'll probably get worse. Still I'd like to see some PS3 games where the Cell is used in really creative ways that make the gameplay more interesting, they have all that FP power and it seems like it's mostly just used to make up for the RSX and that's it because there isn't the budget for anything else outside of a few Sony funded studios.

 

[nAo]

Don't ever, for one second, believe the "separate teams" bullshit. No one is throwing a whole separate team on a platform SKU.

Oh..that's true, the rule usually is "put 30 engineers on 360" and throw another 2 or 3 at PS3. Separate teams indeed!