The Game Technology discussion thread *Read first post before posting*

I had a strange experience this morning where the framerate was a lot smoother than before particularly in cockpit view (as in smooth as butter). I don't know yet if this is also related to the online issues, to the event I was doing, or to the fact that I was using a DS3 this time. I'm definitely going to look into this tonight.
 
EDIT: And what is this?
gt5.jpg


http://www.thesixthaxis.com/2010/11/24/review-gran-turismo-5/

That looks like a photoshop job, I haven't seen any other pics showing such artifacting. People certainly wouldn't be complaining about '16 pixel' wipers if this was true.
Can anyone confirm?

It is very low post-processing resolution. IIRC 1/16th framebuffer res and possibly lower-res. That means rain effects, smoke, dust will make foreground objects alias badly which brings the memories back to C64 era in this case. I assume it wont be as pronounced in replay (30fps) as in ingame racing and photomode probably does 1:1 post-processing buffer res.
 
It is very low post-processing resolution. IIRC 1/16th framebuffer res and possibly lower-res. That means rain effects, smoke, dust will make foreground objects alias badly which brings the memories back to C64 era in this case. I assume it wont be as pronounced in replay (30fps) as in ingame racing and photomode probably does 1:1 post-processing buffer res.

Oh, I read about that, but thought it was limited to the rain/snow particles looking blocky (like driving into pixellated snow) - didn't know it would affect car geometry.
 
The cars are in fog, which is what's causing it. If you played LBP1 you'd see the same effect when you're cute like Sackboy steps into a foggy light and turns all retro. I don't think I've ever recalled a console more demonstrative of the need for loads of BW than PS3. Fewer polygons, simple shaders, less IQ, are nothing like as jarring as significantly lower resolution buffers.
 
About LOS
They also are terrible drivers, the video isn't even in sync with the other console :rolleyes:
Actually I don't care for GT5/Forza comparision owning only a 360 and still not having bought Forza3.
But that vid is indeed funny I find that both Forza and GT5 are actually pretty impressive lawnmower simulation...
:LOL: indeed they pretty much suck
 
Last edited by a moderator:
It is very low post-processing resolution. IIRC 1/16th framebuffer res and possibly lower-res. That means rain effects, smoke, dust will make foreground objects alias badly which brings the memories back to C64 era in this case. I assume it wont be as pronounced in replay (30fps) as in ingame racing and photomode probably does 1:1 post-processing buffer res.

There is something deeper underlying this than that. Somehow the 'framebuffer' effects have some kind of tricky depth resolution. But there is no need for them to affect either foreground or background objects if they were simple framebuffer effects - you would modify the existing image and add their values to them.

You saw these issues also with depth of field effects in GT5P's 'environments' (the screens where you would see your current car in a nice photo travel type location, or in the pits or whatever). And from what I've seen, these effects even occur (albeit at a less bad resolution) in photo mode.

There is something in the rendering engine that causes this issue, and it clearly can't be solved easily. In this picture the lights effects on the front car are also interesting, in that they are clearly in a different pass, or done differently.

Incidentally, Forza 3 isn't showing the advantage of more bandwidth in this area, as it has almost zero effects. Smoke is generally gone in a second.
 
Particles are really interesting and 'evolved' from Prologue, heat haze effect also makes jaggies in GT 5, but it GT 5:p they didnt.
 
Oh, I read about that, but thought it was limited to the rain/snow particles looking blocky (like driving into pixellated snow) - didn't know it would affect car geometry.

Usually geometry infront and behind effect gets affected. But it's a perfomance tradeoff for having more particles/longer lasting particles.

Regarding framerate and tearing it would be interesting to know difference with same track(s) and using only premium cars (if possible) vs only standard cars and then a mix of both. In the meantime it seems you are not allowed to zoom in to much on standard cars to take pics. A reskin should have been made and not just using GT4 skin.

http://i51.tinypic.com/hvp4bb.jpg

The strange thing is it looks like parts of the car is not geometry but rather decals/sprites. You get these low-res sprite edge jaggies that wont go away with AA.
http://nsa20.casimages.com/img/2010/11/25/101125112452175906.jpg
 
Last edited by a moderator:
1.5x actually not 3x.

3x is correct. Remember these are "video" games. They render multiple frames per second. More frames = more resolution.

PGR4: 1280x720x30hz = 27.64 megapixels/sec (what the eye sees, also what is drawn)
GT5: 1280x1080x60hz = 82.94 megapixels/sec (what the eye sees, also what is drawn)

If you take all pixels that are drawn in one second lay in a straight line, GT5's line is 3x longer.

Because the GT5 is drawing 3x more pixels, the GT5 resolution is 3x of PGR4.

Whether it is best to do 3x resolution via more frames per second or more pixels per frame is separate discussion with other factors but bottom line is still 3x resolution.

PGR4 doesn't uses HDR in gameplay, but in photomode it does, though PGR3 had HDR in gameplay as well, but it ran at a subHD resolution.

Thank you for this confirmation. I felt that replay had much better lighting quality but I was not sure why. I did not know replay was subHD. You have good eyes to see this!

Do you know what is the resolution for replays?
 
Last edited by a moderator:
There is something deeper underlying this than that. Somehow the 'framebuffer' effects have some kind of tricky depth resolution. But there is no need for them to affect either foreground or background objects if they were simple framebuffer effects - you would modify the existing image and add their values to them.

It's just a 1/4 x 1/4 res off-screen particle buffer depth tested against a like-wise 1/16 depth buffer (for proper occlusion). Pretty common technique including artifacts, and not surprising if they want to try hitting 60fps. There's already a big hit to the fillrate incurred with the MSAA modes on RSX as is, nevermind bandwidth.
 
@Arwin

Well this looks like a proper like for like location comparision.

http://www.videogameszone.de/Gran-T...n-Edel-Rennspielen-801563/galerie/1464947/#tp

Better, though not perfect. That first shot for instance, the Forza car is way past the PIT entrance where the GT car is right next to it. For the car comparisons, they could have taken the dealer car select screen, as that one shows the car against various backdrops and more full-screen. Especially the latter would have given a better comparison, I think. Both games also have a way to change the cockpit view to make it match the other better.

But this is a good start.
 
I just saw the LoT framerate analysis & I am surprised to see that the game runs smoother in chase cam (3rd person) when compared to the bumper cam view where you don't even have your car onscreen:
http://www.youtube.com/watch?v=nGXONhFOrkg


3x is correct. Remember these are "video" games. They render multiple frames per second. More frames = more resolution.

PGR4: 1280x720x30hz = 27.64 megapixels/sec (what the eye sees, also what is drawn)
GT5: 1280x1080x60hz = 82.94 megapixels/sec (what the eye sees, also what is drawn)

If you take all pixels that are drawn in one second lay in a straight line, GT5's line is 3x longer.

Because the GT5 is drawing 3x more pixels, the GT5 resolution is 3x of PGR4.

Whether it is best to do 3x resolution via more frames per second or more pixels per frame is separate discussion with other factors but bottom line is still 3x resolution.
Thanks for explaining I didn't think of it that way.

Thank you for this confirmation.I felt that replay had much better lighting quality but I was not sure why. I did not know replay was subHD. You have good eyes to see this!

Do you know what is the resolution for replays?
You misunderstood me there, PGR4 runs at 720p regardless of whichever mode you look at (be it gameplay or replay), its just that HDR is only applied when in photo mode. In case of PGR3 the HDR is applied even when in gameplay but the game (PGR3), runs at a subHD resolution.
 
Last edited by a moderator:
It's just a 1/4 x 1/4 res off-screen particle buffer depth tested against a like-wise 1/16 depth buffer (for proper occlusion). Pretty common technique including artifacts, and not surprising if they want to try hitting 60fps. There's already a big hit to the fillrate incurred with the MSAA modes on RSX as is, nevermind bandwidth.
Do we think PD just didn't 'get' the PS3 architecture, and their mindset was too coloured by PS2? A lot of the compromises we're seeing just dont match PD's previous obsesesion with perfection, and I feel they hit technical hurdles and couldn't find a suitable engineering solution so had to make simple cutbacks. That is, if they had created a rain effect early on and hit the limitations of the hardware, would they have tried a completely different system? I still don't understand why no-one seems to be trying a 2D shader based fog or rain effect. Think like the old Z-based fogging only with some shader maths to produce what looks like partcile movement. It'd have limitations, but in something like a racing game where you're mostly travelling into the spray, the end result should be way better than current lowres transparent overdraw.
 
May they suffered a bit of the same problem TB described in regard to sacred 2 (if my memory serves right). The development started a long time ago and they may have done if not wrong but not optimal choices when they set the basis for the engine of the game. 5 years ago things like spurs and plenty of other pre made libray were not available. They may be when they realize they were facing issues it were too late in the development. But for such a game I don't that that would explain the whole story. I suspect some management errors, the multiple delays are a proof of that.
It looks like they tried to fix things, failed, delayed, fix again, etc. At least that's how it look by the communication surrounding the product launch, "will launch when ready" two "prequels" releases.
At some point should have consider a technical "reset" announce a "huge delay" like the game will be out fall 2010. There was a lot of pressure on the team I guess, it's tough to start on clean basis when you have to release "prequels/demo" (I speal of GT HD / prologue).

But that for the technical aspect I wonder if something else went wrong during the project, exec loosing faith or patience? I don't know but I'm surprised about the content too, old tracks barely reworked, idem for 800 hundred of the cars available. There is something fishy. GT team has to be big either artists don't have proper management and/or tools or the team was not that big or was missing resources of some sort. I really wonder if at some point execs put a stop realizing that may be the way Polyphony was working on the ps2 wasn't scalable to the bigger team/resources a next gen requires and it was close to a cluster-fuck. The exec may have say stop and you have one year to ship and put whatever you have together.

EDIT
I also see an interview with Kazunori Yamauchi a game related tv channel (game one) and he spoke quiet a lot of errors he made during development (mostly thing he should have left out of the game to be implemented on sony next system for example). It sounds a bit weird, I know the man perfectionist never happy but still the game has just released I don't expect him to be like the forza devs that were I would dare to say "full of them self" "overly poud" while speaking of their work. Really I smell something fishy. He also said that the next gran tourismo were to ship "fast". I would not be surprised if he feel like he has to redeem him self of some lacking of the product.
Overall GT6 shipping in two years would not surprise me. They put ninja to get a new engine on its feet in two years, refine their assets and launch.
Sometime more can be less, 1000 cars is not needed to make the best racing game, as well as kart, nascar, etc. I would not be surprised if GT6 is more focused to its business which is GT racing. Actually I expect a lot from GT6, lean interface, less cars but better, outstanding graphics at 720P, they may even consider 30fps if they manage to get the input lag as low as criterion in NFS, huge step in AI, impressive physic and damage model, flawless online.
 
Last edited by a moderator:
They should have gone with 720p 2x MSAA. Maybe it was a matter of pride, trying to keep the game at 1280x1080 2x QAA (according to Al), but I think it required too many sacrifices. Not that it's a bad looking game, from any of the gameplay vids I've watched. The quality of some of the ported content is an issue of man hours rather than anything technical.
 
They should have gone with 720p 2x MSAA. Maybe it was a matter of pride, trying to keep the game at 1280x1080 2x QAA (according to Al), but I think it required too many sacrifices. Not that it's a bad looking game, from any of the gameplay vids I've watched. The quality of some of the ported content is an issue of man hours rather than anything technical.
Which is fishy, five years and supposedly a matching budget, that's a lot of man years of hard work invested on the product.
 
Which is fishy, five years and supposedly a matching budget, that's a lot of man years of hard work invested on the product.

The only thing I can think of is they maybe held off on content creation because they were unsure of the overall direction the tech would take the product. Maybe the content creators did not understand the parameters they had to work within which created a lot of churn in the premium content, wasting man hours.
 
Back
Top