Andriasang has a brief one:
http://www.andriasang.com/e/blog/2010/04/02/triace_aska_framework/
I suspect some site will do a detailed translation in a few days.
I heard RoF is a complex and difficult game.
The Game Technology discussion thread *Read first post before posting*
- Thread starter Shifty Geezer
- Start date
Andriasang has a brief one:
http://www.andriasang.com/e/blog/2010/04/02/triace_aska_framework/
I suspect some site will do a detailed translation in a few days.
I heard RoF is a complex and difficult game.
A liittle more details from siliconera.com on the same Japanese article:
http://www.siliconera.com/2010/04/02/resonance-of-fate-hair-is-technologically-impressive/
http://www.siliconera.com/2010/04/02/resonance-of-fate-hair-is-technologically-impressive/
I continue to believe that MT framework works badly with the ps3 hardware, not seems exactly 'elastic' (5 gb of installation, bad fps & lag or too much tearing, 'dynamic qaa' & so on ); I'm pretty sure even the 2.0 version not will change so much, in that perspective. I will be glad to wrong for the future.
Last edited by a moderator:
IIRC this is the first time a framework engine has V lock disabled on the PS3 ver.One would think that after their good work in DMC4 they'd have gone through a lot of optimizations for the PS3.
That's a pretty poor assumption or even an attempt at correlation considering the difference in technology between the two games... Exactly why are you implying that they haven't done "a lot of optimizations".
hmmm 5k just for hair ? and to that include all the additional polygons required for the face & you have a lot of polygons to render while in close up shots of the head. How's the performance like ?
I think he's referring to the multiplayer LP2 demo. Look in the frame rate analysis topic to see the shocking difference.
That said, DMC4 was fantastic and was in fact a hair better on PS3 (cutscene frame rate, LOD bias and some better looking shaders). RE5 got a bad rap unjustly (at least post patch), unless one goes through with a frame rate analyzer but that's just kicking in a placebo effect. LP1 on the other hand had lots of gamma issues and the uncapped frame rate affected the controls noticeably.
Can someone compare textures, alpha effects, lighting, shadows etc.? Because PS3 demo looks just awful
They're identical in that respect, but clearly scaled down from the main campaign demo to accommodate 16 player matches. It's hardly awful though, and I think the texture fidelity is a noticeable step up from RE5.
Billy Idol
Legend
I was just thinking about the facial animation stuff.
I recently played Heavy Rain and was quite dissapointed with respect to the movement of the mouth while the characters talk or try to express a feeling.
I recently also played GOW3, which has a nearly perfect mouth movement and ME2, which seems to be as good! I remember (although it is not a fresh memory) that the ol Heavenly Sword had great facial animation too!
Is there an explanation why this aspect in Heavy Rain dissapoints? As this is a rather ambitous project by Quantic Dreams who seems to be motion capturing specialists (which shows if you ask me in the character movement department, which was rather good looking) and expressing emotion, voice acting and this stuff is one of the important things in this game (should be!!)...it must be really difficult to get the movement of the mouth done right!!
Are the animation made by hand or by some automatic (or maybe semi-automatic) process? Is it easy, but takes much time and thus costs a lot of money!?!?
I recently played Heavy Rain and was quite dissapointed with respect to the movement of the mouth while the characters talk or try to express a feeling.
I recently also played GOW3, which has a nearly perfect mouth movement and ME2, which seems to be as good! I remember (although it is not a fresh memory) that the ol Heavenly Sword had great facial animation too!
Is there an explanation why this aspect in Heavy Rain dissapoints? As this is a rather ambitous project by Quantic Dreams who seems to be motion capturing specialists (which shows if you ask me in the character movement department, which was rather good looking) and expressing emotion, voice acting and this stuff is one of the important things in this game (should be!!)...it must be really difficult to get the movement of the mouth done right!!
Are the animation made by hand or by some automatic (or maybe semi-automatic) process? Is it easy, but takes much time and thus costs a lot of money!?!?
^I think its the actors who are to be blamed...not really a fault of game.
Just watch that video clip that had Scott's voice actor in a movie playing detective & you'll see that its how that guy talks & the game managed to capture his essence exactly as he did it. But you do have some odd scenes where you character's mouth will do a weird movement before talking [Like Ethan's wife "yawning" before she asks Ethan to look after Jason at the mall]
Just watch that video clip that had Scott's voice actor in a movie playing detective & you'll see that its how that guy talks & the game managed to capture his essence exactly as he did it. But you do have some odd scenes where you character's mouth will do a weird movement before talking [Like Ethan's wife "yawning" before she asks Ethan to look after Jason at the mall]
Blaming the actors is b******, sorry.
Facial animation is an extremely complex issue, especially in games where massive amounts are required, compared to even a full CGI movie. There are basically two possible approaches to solve this problem:
Use straight motion capture for all the facial animation. This requires processing a massive amount of mocap data - cleaning, tweaking, reworking etc. As far as I know both GTA4 and Heavy Rain took this approach, and probably Assassin's Creed 2 as well; what's more interesting is that the outsourcing company involved has been Image Metrics in all these cases.
So the biggest factor is probably the general budget of the cleanup work, which had an obvious effect on the quality. We all know GTA4's budget was huge compared to the other two games. Although it's also worth noting that the general quality of the facial rig that IM receives from their clients is also a factor here.
The other approach is a procedural one, where you build a set of gestures and drive them by adding timed keys to the animation or voice in the game. For example Mass Effect does this; they do automatic lip sync to the voice file and the designers add gestures (that aren't only facial animations, but full body poses, short movements etc). There's a specialized off the shelf software for this that I've seen recently called FaceFX.
This approach can also use mocap data for the facial gestures and expressions, but these are very basic elements that are combined during runtime for the full animation.
So, all in all, I'd say that the issues with Heavy Rain are probably a result of combining less then perfect character rigging with mass produced facial animation data.
Also note that GOW3 has a lot less facial animation, being mostly an action game, so they had less work to do as well, and in turn more time to polish what they had. Although Uncharted 2, another game with high quality facial animation, has a lot of cinematics too, but I guess ND was able to fit the necessary work into their schedule.
Facial animation is an extremely complex issue, especially in games where massive amounts are required, compared to even a full CGI movie. There are basically two possible approaches to solve this problem:
Use straight motion capture for all the facial animation. This requires processing a massive amount of mocap data - cleaning, tweaking, reworking etc. As far as I know both GTA4 and Heavy Rain took this approach, and probably Assassin's Creed 2 as well; what's more interesting is that the outsourcing company involved has been Image Metrics in all these cases.
So the biggest factor is probably the general budget of the cleanup work, which had an obvious effect on the quality. We all know GTA4's budget was huge compared to the other two games. Although it's also worth noting that the general quality of the facial rig that IM receives from their clients is also a factor here.
The other approach is a procedural one, where you build a set of gestures and drive them by adding timed keys to the animation or voice in the game. For example Mass Effect does this; they do automatic lip sync to the voice file and the designers add gestures (that aren't only facial animations, but full body poses, short movements etc). There's a specialized off the shelf software for this that I've seen recently called FaceFX.
This approach can also use mocap data for the facial gestures and expressions, but these are very basic elements that are combined during runtime for the full animation.
So, all in all, I'd say that the issues with Heavy Rain are probably a result of combining less then perfect character rigging with mass produced facial animation data.
Also note that GOW3 has a lot less facial animation, being mostly an action game, so they had less work to do as well, and in turn more time to polish what they had. Although Uncharted 2, another game with high quality facial animation, has a lot of cinematics too, but I guess ND was able to fit the necessary work into their schedule.
Billy Idol
Legend
Thanks for the explanation!
I hope that Quantic Dreams focus on this issue for the next game!
Guys I have a question, I understand that in Xbox 360's case avoiding tiling & having the framebuffer within the eDRAM itself provides the developers lots of bandwidth for alpha, so what I wanna know is does this automatically means that when you tile you have a significant reduction in the amount of bandwidth that's available ? what are the other disadvantages of tiling ?
No, tiling doesn't affect bandwidth to the EDRAM. MSAA and/or MRT, if anything, _improves_ the bandwidth. Then again, you're almost never EDRAM-bandwidth-bound, so this is not a big consolation.
Tiling makes code a bit more complex, and makes you handle a bit more geometry due to overlap.
Then why do I read in almost every DF article (whenever the eDRAM is mentioned) that "by constraining the framebuffer in side the fast eDRAM the developers are left with near infinite amount of bandwidth for alpha"
I'd have to see that quote in context. Even when you use tiling, the framebuffer operations are performed inside eDRAM, only the workload for the whole framebuffer is performed piecemeal.
This is a long time ago and I'm not a 360 developer so others here can correct me, but from what I remember, you can put a framebuffer in your EDRAM, and once it is there, you can use a limited set of instructions to manipulate it at full bandwidth speed (256GB/s).
However, you can't do just everything there right up to the point of showing it to the user - it has to go from there to somewhere else at 1/8th of that speed max, and I believe that's a 32GB/s bandwidth that is shared among both GPU and CPU. Therefore, it's not suitable for all types of render pipelines. For instance deferred rendering techniques can't always benefit from this. Tiling complicates that issue further.
Also, if you are combining different framebuffers, there's not enough room in EDRAM to do all that, which means the image has to come out of there for post processing, and that often impacts the AA that is possible.
Again, this is just from reading and memory, so hopefully someone with more hands-on knowledge can correct me if I'm wrong.
However, you can't do just everything there right up to the point of showing it to the user - it has to go from there to somewhere else at 1/8th of that speed max, and I believe that's a 32GB/s bandwidth that is shared among both GPU and CPU. Therefore, it's not suitable for all types of render pipelines. For instance deferred rendering techniques can't always benefit from this. Tiling complicates that issue further.
Also, if you are combining different framebuffers, there's not enough room in EDRAM to do all that, which means the image has to come out of there for post processing, and that often impacts the AA that is possible.
Again, this is just from reading and memory, so hopefully someone with more hands-on knowledge can correct me if I'm wrong.
@Shifty
This is just one of the few articles, since it was recently written it was easy to find couldn't remember specific details about rest articles
If its enormous fillrate available either ways then why is the article suggesting that since SE is keeping the framebuffer within the eDRam they should have lots of fillrate at their disposal..?
This is just one of the few articles, since it was recently written it was easy to find couldn't remember specific details about rest articles
http://www.eurogamer.net/articles/digitalfoundry-final-fantasy-xiii-face-off?page=2
If its enormous fillrate available either ways then why is the article suggesting that since SE is keeping the framebuffer within the eDRam they should have lots of fillrate at their disposal..?
Similar threads
