PhysX: PS3 can handle it; 360 features limited

[Carl B]

The whole 360 aspect aside, Ars was always wrong on the Cell's physics abilities - even Anand gave Cell that much.

Since there seems to be a lot of conflict over 360 vs PS3 in terms of physics with this news, I think we might as well drop the negativity towards the XeCPU, since indeed it is just one SDK at this point. Still, I think we can all agree that it's *good* news for Cell.

 

[dukmahsik]

The whole 360 aspect aside, Ars was always wrong on the Cell's physics abilities - even Anand gave Cell that much.

Since there seems to be a lot of conflict over 360 vs PS3 in terms of physics with this news, I think we might as well drop the negativity towards the XeCPU, since indeed it is just one SDK at this point. Still, I think we can all agree that it's *good* news for Cell.

yes but that doesn't mean physics will suck on 360. what if the difference is 2/10ths?

 

[Jawed]

The Leigh Davies PDF linked earlier is a real eye-opener. It's heartening to see that quite a lot of devs have already grasped the multi-threading nettle whole-heartedly.

Jawed

 

[Carl B]

yes but that doesn't mean physics will suck on 360. what if the difference is 2/10ths?

No, I don't think physics will suck on the 360. Though it's hard to say what we'll all define as 'sucking' five years from now. Do PS2's graphics 'suck'? That kind of thing...

Differences may be minor in absolute terms, but it won't take much for one console to become more strongly associated as the 'physics' console, or the 'graphics' console, or the 'online' console. Every minor difference in this field gets amplified a thousand times over since there's no real additional competetion or other points of reference once a new console generation has begun.

 

[Brimstone]

Ageia has acquired Meqon (Duke Nukem phyisics).

http://news.teamxbox.com/xbox/9146/AGEIA-Acquires-Meqon-Research-AB/

 

[Titanio]

yes but that doesn't mean physics will suck on 360. what if the difference is 2/10ths?

If Xenon is closer to the dual-core PC CPU end of the spectrum than the PhysX/PS3 end, I think the difference would be (a lot) more substantial than that. Say, with another core, X360 outperformed the dual-core CPU 2x, that's still a far cry from the other end in that benchmark.

Anyway, that's rather speculative. But that the difference was enough to compel them to cut features is rather suggestive. Hopefully we'll get actual data sooner rather than later.

 

[SlmDnk]

http://www.ageia.com/press/physx_carwash_video/physx_carwash_lavajets.exe
http://www.ageia.com/press/physx_carwash_video/physx_carwash_napalmlong.exe

very nice

I've collected all kinds of physX links (for some time now) on this page: http://personal.inet.fi/atk/kjh2348fs/ageia_physx.html

There's links to all the videos I know of. These are my favourites at the moment.

 

[Alpha_Spartan]

Essentially the 360lacks hardware physics acceleration capabilities. So realistic physics will have to be done through "software rendering". Which isn't nearly as efficient.

I don't believe that 360 lacks "hardware physics acceleration capabilities" or that Ageia proved this. I believe that the 360 architechure isn't suitable for Ageia's SDK. Saying that 360 lacks the ability to calculate robust physics is generalizing a bit.

I'd like to see some details on the PhysX chip, something tells me the architecture is probably similar to the Cell.

 

[Alpha_Spartan]

The whole 360 aspect aside, Ars was always wrong on the Cell's physics abilities - even Anand gave Cell that much.

Since there seems to be a lot of conflict over 360 vs PS3 in terms of physics with this news, I think we might as well drop the negativity towards the XeCPU, since indeed it is just one SDK at this point. Still, I think we can all agree that it's *good* news for Cell.

A voice of reason rings out in a sea of hype and speculation.

 

[Titanio]

Can anyone clarify from the 4Gamer.net article - was the demo running at 3-5fps with the dual-core intel and then 40fps with the PhysX chip? I'm a little confused as The Extremetech article talks about another demo where the dual-core apparently runs it at 40fps - a different demo?

 

[pc999]

Thanks for the info one.

 

[Acert93]

Neither would I. Their "programmer art" tech demos don't do them much favours

Yes, their water demos stunk.

e.g. the water fountain looked like a a fountain of mini-clear glass balls. They did not puddle, they did not flow. Justs lots of little balls. The soap and fire ones were equally meh.

From that perspective the Havok demos always gave a better impression. The art was not great, but they created "game like" scenarios to show off their stuff. While this does not mean Havok is better, it does show they are better at advertising their product.

If Xenon is closer to the dual-core PC CPU end of the spectrum than the PhysX/PS3 end, I think the difference would be (a lot) more substantial than that. Say, with another core, X360 outperformed the dual-core CPU 2x, that's still a far cry from the other end in that benchmark.

FP will be an area where Xenon outperforms a desktop PC. A dual Xeon server its about 10GFLOPs in the realworld. One of the reasons MS went with the PPC design was to strike a "better" balance between GP performance and FP performance.

Pulling out ficticious numbers like "X360 outperformed the dual X360 outperformed the dual-core CPU 2x, that's still a far cry from the other end in that benchmark" is more of a "what if" that got pulled from who knows where.

I am not sure what the hubbub is about. Devs here have not voiced too much concern, we already knew Physics would be an area where CELL could excell if certain hurdles were overcome, and this is one software house (which is trying to port a SDK to 5 platforms, 4 new--multithreaded/cored CPUs, PhysX, CELL, Xenon).

Also, it seems Xbox 360's architecture is not bad for physics, its that the Novodex engine does not perform as well with fluid dynamics in the Xbox 360 architecture. Whether this is a conflict between the SDK design or a problem with the 360 who knows. My guess is somewhere in between, relatively, as CELL should do well here (it is 50% larger and seems aimed at this very task!).

But I would not bet on the 360 being unable to do this. It is like MSAA + HDR, free AA, or having 3 better GP processors--there are always tradeoffs.

As Shifty said, CELL would be a FLOP if it did not excell in certain areas and this certainly is an example of where it excells. To steal his example from the other day--this could be the difference between PS3 having 400 objects and the 360 having 200.

Tradeoffs. And yet devs have been very good at overcoming limitations of the hardware to do what was thought initially impossible. Somethings are impossible, granted, but to say the 360 cannot do FD would not be one I would bet on. It obviously can to a degree, the question is can someone overcome the architecture to do so well? If it is important to gameplay then a dev will find a way. Just like Shiny figured out normal maps on the PS2. There is too much money involved not to.

 

[Titanio]

FP will be an area where Xenon outperforms a desktop PC. A dual Xeon server its about 10GFLOPs in the realworld. One of the reasons MS went with the PPC design was to strike a "better" balance between GP performance and FP performance.

Pulling out ficticious numbers like "X360 outperformed the dual X360 outperformed the dual-core CPU 2x, that's still a far cry from the other end in that benchmark" is more of a "what if" that got pulled from who knows where.

As I said, that's speculative. But you must admit that more than a small gap likely exists if they've been pushed to the point of cutting certain things.

I think it's likely that it's not a matter of X360 being as good as Cell or PhysX with "the rest" either..I'd say there's probably a performance difference there too.

Tradeoffs. And yet devs have been very good at overcoming limitations of the hardware to do what was thought initially impossible. Somethings are impossible, granted, but to say the 360 cannot do FD would not be one I would bet on. It obviously can to a degree, the question is can someone overcome the architecture to do so well?

I totally agree. But this applies as much to any system as it does X360 - there'll be more things PS3 can do too.

 

[Laa-Yosh]

This whole PhysX stuff is way overhyped IMHO...

 

[Acert93]

As I said, that's speculative. But you must admit that more than a small gap exists if they've been pushed to the point of cutting certain things.

1. Speculation on make-believe numbers is not speculations!

2. You are interpretting. From the above statement you seem to clearly indicate the issue is a gap (in performance) and thus resulting in the "cutting" of a feature.

This assumes it is a performance issue; instead the question seems to be *architecture*. The feature dropped, fluid dynamics, has specific needs to perform at a high level. The 360 architecture and the Novodex SDK seem to hit a wall here.

I think it's likely that it's not a matter of X360 being as good as Cell or PhysX with "the rest" either..I'd say there's probably a performance difference there too.

Your letting your Sony bias ass/u/me too much. This could very well be true--but what we are hearing specifically is about a tension between how Novdex does FD and the 360 architecture.

You are the one inserting comments about 1. make believe numbers and 2. expanding the issue beyond what has been officially stated.

I totally agree. But this applies as much to any system as it does X360 - there'll be more things PS3 can do too.

And that is the tendancy of your posts. What is the PS3 doing and the Xbox 360 not. Even when MS does what you think it right (like not forcing everyone to get a HDD) you still are negative.

Having a bias is fine... but when you start making up numbers to draw "speculative" comments, well, that is not technical and that seems more like version's pipe dreams and really add nothing to the discussion.

And like I said, I expect the PS3 to be better at physics overall by the end of the gen. But I am not going to pretend to make up numbers and speculate on them; and I certainly see a difference between a feature--like FD or MSAA+HDR--NOT being indicative of performance, gaps, or cutting, but instead ARCHITECTURE.

It would be crazy, without solid specs and benchmarks, to say because Xenos has HDR+MSAA it is better than RSX or people have to "cut" stuff to make up a gap. It is a feature difference related to the design and architecture. Ditto FD. So while CELL has its SPEs chewing up FD code Xenon will be doing things the SPEs totally suck at. But I personally don't see a need to insert those types of comments unless we are discussing the archtectural strengths and designs... in this specific thread the logic is taking leaps beyond the data.

 

[Acert93]

This whole PhysX stuff is way overhyped IMHO...

They look pretty dead in the PC space IMO. Too expensive, too late. And eventually Intel/AMD or NV/ATI will offer the same thing as a pack in and leverage their revenue channels and dev support and *poof* no more PhysX.

 

[dukmahsik]

1. Speculation on make-believe numbers is not speculations!

2. You are interpretting. From the above statement you seem to clearly indicate the issue is a gap (in performance) and thus resulting in the "cutting" of a feature.

This assumes it is a performance issue; instead the question seems to be *architecture*. The feature dropped, fluid dynamics, has specific needs to perform at a high level. The 360 architecture and the Novodex SDK seem to hit a wall here.



Your letting your Sony bias ass/u/me too much. This could very well be true--but what we are hearing specifically is about a tension between how Novdex does FD and the 360 architecture.

You are the one inserting comments about 1. make believe numbers and 2. expanding the issue beyond what has been officially stated.



And that is the tendancy of your posts. What is the PS3 doing and the Xbox 360 not. Even when MS does what you think it right (like not forcing everyone to get a HDD) you still are negative.

Having a bias is fine... but when you start making up numbers to draw "speculative" comments, well, that is not technical and that seems more like version's pipe dreams and really add nothing to the discussion.

And like I said, I expect the PS3 to be better at physics overall by the end of the gen. But I am not going to pretend to make up numbers and speculate on them; and I certainly see a difference between a feature--like FD or MSAA+HDR--NOT being indicative of performance, gaps, or cutting, but instead ARCHITECTURE.

It would be crazy, without solid specs and benchmarks, to say because Xenos has HDR+MSAA it is better than RSX or people have to "cut" stuff to make up a gap. It is a feature difference related to the design and architecture. Ditto FD. So while CELL has its SPEs chewing up FD code Xenon will be doing things the SPEs totally suck at. But I personally don't see a need to insert those types of comments unless we are discussing the archtectural strengths and designs... in this specific thread the logic is taking leaps beyond the data.

post of the year

 

[seismologist]

Unless I'm mistaken we still haven't seen proper realtime fluid dynamics on the PC. I imagine this is due to similar architectural limitations as in the 360 case (hence the need for a PPU).

The fact that Cell lends itself well to these computation heavy tasks should come as no surprise to anyone. Yet some of you act like you're afraid of what might be possible with Cell...it's actually pretty funny.

 

[Carl B]

So while CELL has its SPEs chewing up FD code Xenon will be doing things the SPEs totally suck at. But I personally don't see a need to insert those types of comments unless we are discussing the archtectural strengths and designs... in this specific thread the logic is taking leaps beyond the data.

I don't disagree with your logic concerning the whole PhysX debate Acert, but at the same time let's not shortchange the SPE's using the 'common knowledge' of their weaknesses, when indeed we don't know what they'll be capable of working on five years down the line either.

 

[Dave Baumann]

This is why MS should have included a PPU. DeanoC was wrong when he said a PPU is not needed for the xbox360. Xbox360 needs a PPU so it can handle fluid dynamics.

Why can't the graphics handle fluid dynamics?