A scientific experiment - Cell does T&L, with MAG

[Lucid_Dreamer]

You can hardly see anything. The ghosting is a mess, making it hard to work out how much is happening and how many are on screen on once and how complex the models are. It's certainly not smoothly animating. There's also just a moment at the beginning where there's clearly ~40 players on screen given health bars, but then they all spread out and you only see a dozen or two on screen at once. There's no measure of LOD or environment complexity versus other titles (you divided the vertex budget between characters and scenery, so if your game targets less than players, you can invest more in scenery and vice versa) so even having 40 characters at a time might not be seriously pushing polycounts.
Being multiplayer is immaterial to the requirements of TnL unless you want to design your system to account for GAF coordinated parties. You design assets for a given quality, provide suitable LOD, and manage accordingly regardless of whether troops are positioned by AI or other people.

Here's a crowd vid for you:

OMG! Only the powerz of teh Cell can make this happen! Except it's a 2006 PC game, not only drawing and animating thousands of troops, but managing their AI too. Of course, that's because the GPU is handling the TnL leaving the CPU to calculate AI...

There's nothing remarkable about MAG's vertex throughput or graphics only made possible by the Cell, AFAICS. It'd make far more sense to use Cell to help with vertex culling prior to RSX TnL (players obscured by other players or scenery need not be drawn), using all the hardware available instead of having the vertex units idle.

Let's not get ridiculous here. If the requirement is just characters, controlled by A.I.; Lair has thousands of enemies controlled by A.I. in 2007. 6:40secs

http://www.youtube.com/watch?v=BLdTn8Z1AM0

Here is one with 15,000 A.I. controlled objects/characters on screen.

Each individual chicken has its own behavior model interacting with other birds. The simulator was demonstrated to provide realtime (30fps) performance with several thousand chickens. In fact, when the number of chickens was increased to a total of 15,000 birds the Cell B./E. processor was still able to perform the simulation with interactive speed, but the graphics rendering was not able to keep pace, even on a stateoftheart NVidia GPU and started dropping 2 out of 3 frames, resulting in 10fps "sluggish" video output.

www.simbiosys.ca/science/white_papers/eHiTS_on_the_Cell.pdf

I love the way some of you guys try to act like I'm making this stuff up, or it's act like it's nothing.

EDIT: Also, 256 players (MAG) is still in the Guiness Book of World Records as the most? Why is that still around? Hell, why was it a record in the first place? According to you, this should've been done on another platform in 2006 or so.

 

[Mize]

Can someone tell me the point of this thread?

Does anyone actually play MAG anyway? I seem to remember games being devoid of players after about a month and no single-player.

Oh, and can someone please, please explain the method and procedure (no, not the same) of this scientific experiment?

 

[Shifty Geezer]

Let's not get ridiculous here.

The thread opened with the most ridiculous statement, that looking at MAG screenshots we can see how Cell is used for TnL.

www.simbiosys.ca/science/white_papers/eHiTS_on_the_Cell.pdf

I love the way some of you guys try to act like I'm making this stuff up, or it's act like it's nothing.

That paper has nothing to do with anything. The thread, ludicrous as it is, is that Cell has enabled MAG on PS3 thanks to taking on the job of transform and lighting. The arguments should be about that. The argument of lots of characters being impressive, one you were supporting, doesn't stand to logical scrutiny - you yourself find other games rendering more characters than MAG, so what's its claim to fame? MAG isn't doing anything out of the ordinary or incomparable to games using TnL on RSX (at least, no-one has presented a proper argument showing that). Bringing up random Cell papers is illogical - it's known to be good at certain workloads and it should be very good at TnL too. It just makes little sense to use Cell for that and not RSX, and there's certainly nothing to the OP. So please either present info that MAG is doing something impressive regards TnL workload, or a white paper on the technology of MAG that proves its using TnL, or otherwise refrain from random Cell trumpeting.

EDIT: Also, 256 players (MAG) is still in the Guiness Book of World Records as the most? Why is that still around? Hell, why was it a record in the first place? According to you, this should've been done on another platform in 2006 or so.

Clearly you don't comprehend the argument. 256 players is impressive in its online execution and not the TnL of the visuals.

 

[kagemaru]

Awww my post was edited. I honestly thought that was a harmless joke.

Let's not get ridiculous here. If the requirement is just characters, controlled by A.I.; Lair has thousands of enemies controlled by A.I. in 2007. 6:40secs

http://www.youtube.com/watch?v=BLdTn8Z1AM0

I wouldn't really call enemies standing around and waiting to be slaughtered as "A.I.". Also you're ignoring the point about AI and environment complexity that Shifty just told you about. Yes, there looks to be a lot of enemies in that area in Lair, but again they were just standing there, the environment was a barren field, and the soldiers on the ground were mere specks, probably composed of a handful of polys each.

Half joking here, but you have to realize your point falls apart when you start using Lair to support it.

Here is one with 15,000 A.I. controlled objects/characters on screen.

www.simbiosys.ca/science/white_papers/eHiTS_on_the_Cell.pdf

I love the way some of you guys try to act like I'm making this stuff up, or it's act like it's nothing.

That quote still tells us nothing. How complex was this "behavior"? How complex were the models for these chickens?

All I get from that quote was the GPU was the bottleneck, and that's about it.

I think most of us don't think you're making stuff up, but maybe mistaken is a better word. Also we act like it's nothing because it really is nothing.

Can someone tell me the point of this thread?

Does anyone actually play MAG anyway? I seem to remember games being devoid of players after about a month and no single-player.

Oh, and can someone please, please explain the method and procedure (no, not the same) of this scientific experiment?

I played it for a night at a friends house after him and I had a bit of fun with the beta. I liked the concept and thought the core idea was pretty good. But the game lacked direction, even with their system in place, and the performance was pretty poor when the scale actually got impressive.

 

[Shifty Geezer]

Can someone tell me the point of this thread?

You're right. It's utter garbage. I was just going to ignore it, but it's an embarrassment to the board. "Look at MAG. It's clearly using TnL on Cell. Cell's clearly no good at floating point ops. Power7 is better." not even a coherent line to the discussion - are we supposed to be discussing Cell's lack of float power (there's a Cell performance thread for that, and it's a pretty perfect fit for TnL), or MAG's use of Cell for TnL, or Power7 compared to Cell?

 

[willardjuice]

I honestly thought that was a harmless joke.

Maybe the Sony worshipping altar you have grants your PS3 super powers to render games flawlessly?

And just to make clear, fanboy statements like that are not "harmless jokes" and will be punished (when I'm less busy...).