SPU usage in games
- Thread starter betan
- Start date
Patsu man these questions are answered just a couple of posts ago.
(~#169)It was Naughty Dog.
Patsu man these questions are answered just a couple of posts ago.
It was Naughty Dog.
It wasn't a single company though, right? You don't think Insomniac, being good buddies with Naughty Dog, and generally being considered masters of SPU at this point, and responsible for a great game that was extremely complete and ready at launch, have contributed anything to Edge in this area? Not to mention that I'm still convinced that Edge has bits contributed from several teams, first AND second party.
archangelmorph
Veteran
I hope you're not serious!?
R&C seems to have MUCH higher poly counts and better quality shaders than alot of games in the pipe.. Are you refering to the PS2 version or something?
I'm pretty sure the insomniac guys had already implemented similar or comparable tech before the edge tools were made available.. That's about the only reasonable excuse I can see for them to reject it.. (I could be wrong though mind u..)
I thought Insomniac weren't doing triangle culling on SPEs, and/or other graphics setup tasks.
Why there plenty of reasons not to use that kind of technology?
Nothings free from a performance or memory point of view, if your bottlenecked by a particular area of GPU, it might be advantageous to use. If not it might just slow you down and waste memory.
Insomniac (in particular I've chatted to Rob a bit) are clever chappys, I have total confidence they know and used whats best for them.
Patsu man these questions are answered just a couple of posts ago.
You never know the exact answer since no one from Codemaster or PSSG confirmed that they used/did not use EDGE. All we know for sure is they used PSSG. Hence, the question to double confirm since joker454 is a dev.
But the bottomline is not everyone uses EDGE to achieve good framerate and impressive visuals. e.g., Heavenly Sword, Resistance, R&C (Yes, I think it's high poly too because it has many active objects on screen), and perhaps Dirt.
You never know the exact answer since no one from Codemaster or PSSG confirmed that they used/did not use EDGE. Hence, the question to double confirm.
Well, I can confirm right now that PSSG doesn't use EDGE at all. Whether Codemasters themselves used aspects of EDGE above and beyond their usage of PSSG, I don't know, but I suspect not.
archangelmorph
Veteran
Aww shux..
I tried to cover my back by saying "comparable tech" (as in comparable with regards to using separate methods to achieve similar performance gains..) but it doesn't look like it worked..
It wasn't a single company though, right? You don't think Insomniac, being good buddies with Naughty Dog, and generally being considered masters of SPU at this point, and responsible for a great game that was extremely complete and ready at launch, have contributed anything to Edge in this area? Not to mention that I'm still convinced that Edge has bits contributed from several teams, first AND second party.
Well all I can say really is in terms of external contributions, DeanA probably has a fairly complete sense of the scope of contribution from developers - whether he will feel comfortable speaking to the topic who knows. As far as Insomniac themselves, the impression I've gotten through a couple of conversations is one of 'technology independence' in terms of the way they're pursuing their tech, but we'll further look into at some point both their level of use and contribution to EDGE.
If a game already has frustum culling (I wanna see a gamme that doesn't have that!), good world/meshes partitioning (so that we don't draw a lot of stuff that is completely out of the screen) and a decent lod system (so that we don't draw a lot very small/vanishing primitives) most of the savings Edge culling can give to you are due to backfacing triangles culling.
IMHO one can achieve similar results on non skinned meshes just partinioning meshes in clusters that share similar normals (within a certain threshold angle) and then cull them at cluster level.
This would be less accurate than a per triangle test, but it be would at least an order of magnitude faster as you just need to skip subsets of your index buffer.
One could just generate a new optimized index buffer via memcopys, lazier implementations are also possible where just multiple draw commands are issued to draw all the visibile clusters, this wouldn't even require to modify your index buffers! Also the additional memory to implement this system would be negligible (probably 16 bytes per cluster, or something like that..)
It looks also to be extremely simple to implement..mhhh..maybe I should!
[EDIT] Forgot to add that this idea would be feasible only if you draw indexed triangles lists, strips would make everything so much more complicated. who needs strips anyway?
Marco
IMHO one can achieve similar results on non skinned meshes just partinioning meshes in clusters that share similar normals (within a certain threshold angle) and then cull them at cluster level.
This would be less accurate than a per triangle test, but it be would at least an order of magnitude faster as you just need to skip subsets of your index buffer.
One could just generate a new optimized index buffer via memcopys, lazier implementations are also possible where just multiple draw commands are issued to draw all the visibile clusters, this wouldn't even require to modify your index buffers! Also the additional memory to implement this system would be negligible (probably 16 bytes per cluster, or something like that..)
It looks also to be extremely simple to implement..mhhh..maybe I should!
[EDIT] Forgot to add that this idea would be feasible only if you draw indexed triangles lists, strips would make everything so much more complicated. who needs strips anyway?
Marco
Last edited:
And why would they? They have 10 world class engine developers on their team and judging by their tendency to hit the hardware directly with low level assembly I doubt they would even benefit from EDGE technology whatsoever. If anything, their games would probably take a hit by using it.
I'm not certain on the usage of EDGE at Codemasters, but their NEON engine was built with technology obtained from Sony's cross platform multi-core optimized PSSG graphics engine. The PSSG engine, incidentally, is free to all licensed PS3 developers.
As you know graphic engines are always evolving and the graphic engine used for Resistance pales in comparison to the current iteration of the engine being used in R&C and Resistance 2. As an example of this evolution, the physics processing power of the engine has improved 400%.
As for the visual fidelity of Resistance (which I happen to think is very good, but always a subject of personal preference), well, you need to take into account the obvious sacrifices made to support 40 player multi-player online matches. Unless I'm mistaken there hasn't been a console FPS that even comes close to supporting that many online players
Xenon don't take this the wrong way, but I find this post annoying as hell.
Have you even been reading the thread?
Pardon me, but why then are these people in the Resistance credits?
Mark Cerny, Pål Engstad, Naty Hoffman, Manny Ko, Swaminathan Narayanan, Jason Hughes, Jon Olick, Cort Stratton, David Simpson, Jason Scanlin, Boris Batkin, Alex Hastings, Christer Ericson, Rob Wyatt, Pierre Dufresne, Vassily Filippov, Ben Diamand, John Folliard, Brad Byrd, Keith Bruns, John Morgan, Ben Weston, Eric Lengyel
Well I'd hope Rob Wyatt was in the credits ;-)
As has been stated many times, when your published by Sony you get to use a lot there development teams for things. So those teams quite rightly get a credit (from some of those names, I'd say thats the ICE and SCEA TnT teams). Doesn't mean Insomniac aren't good themselves, just that if theres a good solution you have access to, and that suits your game, you'd be silly to not use it.
For us who are published by SCEE (WWS), having ATG as your first port of call for problems and tech is great. For example both HS and KZ2 have benefited greatly by having access to ATG. And in HS's case its only fair to give DeanA a personal shout out and thanks, as apart from being the main architect of our animation system, has helped solve many an obscure bug all over the code base even when the friday night pub has been calling.
And to get back to the Edge questions, thats what Edge is about. Providing some of the Sony WWS tech out to developers who aren't published by Sony. Its basically the same as if EA decide to give bits of Renderware away for free.
Excuse me? Just because their name is in the credits does not automatically mean they contributed to the creation of the engine in Resistance or the application of any EDGE technologies to the game. Mark Cerny, for example, is a consultant with Insomniac which is probably why he was thanked in the credits.
Similar threads
- Locked