The Game Technology discussion thread *Read first post before posting*
- Thread starter Shifty Geezer
- Start date
Ah, right. I guess a bigger deal was made of LC in the run up to it's release, probably because of it's Half Life setting. DoD was a pretty cool game.
Do you have any idea what those portals cost in the rendering pipeline?
hm... You're dynamically placing camera views, so I suppose just render to texture/off-screen buffers. The hard part I think is getting things to appear to travel through it seamlessly... Dealing with clipping the model geometry at a surface then seeing it/projecting it on the surface, if that makes sense? That would also imply rendering to 2 or 4 more full res targets (two complete doorways in co-op). Since it's dynamic, it'd probably have to be done in extra render passes?
*shrug*
*shrug*
Couldn't instancing work here? Basically create a copy of the scene and place it at the other side of the portal but maintain a relationship that allows you to update the copy in sync with the original (and vice versa - instancing in offline 3D apps means basically applying multiple transform matrices to the same object so if you manipulate it there's just an additional reverse transformation).
It might appear like a backwards approach, instead of looking at the same geometry from different POVs you use multiple pieces of geometry, but it could simplify a lot of things, like actually moving through the portal. Even if you open both portals you only need to handle three instances of everything and the levels are fairly simple...
Then again, in coop this means up to four portals, with split screen rendering that's 5 instances and two POVs...?
Would be nice to hear an engine programmer's opinion...
It might appear like a backwards approach, instead of looking at the same geometry from different POVs you use multiple pieces of geometry, but it could simplify a lot of things, like actually moving through the portal. Even if you open both portals you only need to handle three instances of everything and the levels are fairly simple...
Then again, in coop this means up to four portals, with split screen rendering that's 5 instances and two POVs...?
Would be nice to hear an engine programmer's opinion...
This is what they say about portal rendering in the first game:
And..
They also mention that for a short radius around the portal the physics model is less accurate to handle transitions easier. They say generating a portal costs around 10ms (not on what hardware though), and earlier versions took about 500ms.
(From the in-game commentary).
And..
They also mention that for a short radius around the portal the physics model is less accurate to handle transitions easier. They say generating a portal costs around 10ms (not on what hardware though), and earlier versions took about 500ms.
(From the in-game commentary).
Last edited by a moderator:
honestly L4D/L4D2 has the same crappy IQ and probably implementation of the source engine as portal 2 on the 360.
you get heavily aliased and ( lower resolution? ) edges when things are hit with some kind of light source. play around with the flash light in the game almost reminds me of when they do half rez alphas on ps3
http://xboxlivemedia.ign.com/xboxli...sacrifice-left-4-dead-2-20101007044429045.jpg
well thats not true, orange box came out on the ps3 and there a significant improvement in IQ and performance since then, there was heavy optimization and even recruiting which went into the ps3 version.
you get heavily aliased and ( lower resolution? ) edges when things are hit with some kind of light source. play around with the flash light in the game almost reminds me of when they do half rez alphas on ps3
http://xboxlivemedia.ign.com/xboxli...sacrifice-left-4-dead-2-20101007044429045.jpg
well thats not true, orange box came out on the ps3 and there a significant improvement in IQ and performance since then, there was heavy optimization and even recruiting which went into the ps3 version.
Last edited by a moderator:
Lucid_Dreamer
Veteran
Again, this was not Valve.
That doesn't mean they didn't use the existing code for Portal 2. Seems reasonable to think they would use what they had already and would make improvements on it. Hell, the Orange Box patch for the PS3 made a few stability improvements, if I remember correctly.
I really doubt they would have used another devs code, which was probably a mess. Even after the patch there was still a short delay before every explosion, framedrops everywhere, and the potential to go into a slideshow in some places that only a restart could fix.
Think it would have been much easier to port the latest build of the PC version to PS3 and not worry about fixing up what EA did wrong, fixing the huge amount of bugs, and then having to add the improvements to the engine that have been made since then.
Think it would have been much easier to port the latest build of the PC version to PS3 and not worry about fixing up what EA did wrong, fixing the huge amount of bugs, and then having to add the improvements to the engine that have been made since then.
The difference between disc and HDD seek times is a huge factor for a virtual texturing system. Each page load (or page tile load) needs a seek, and disc seeks can be up to 100ms each, while HDD seeks can be up to 20ms (five times faster). Since each page is rather small, the seek time is dominating the latency (the delay of seeing the surface and loading the texture detail it needs).
XBLA games (larger than 512MB) can require a HDD, but the retail games cannot. That's kind of a bummer for fine grained streaming technologies, such as virtual texturing. It doesn't seem so bad however. There's some minor texture detail popping when you look really closely to detailed surfaces (in the Brink video), but nothing major. Id software's virtual texturing system seems to be working really well. There should be even less texture issues in Rage, since it runs at 60 fps (assuming they do page id rendering at full rate). It's already looks better than Crysis 2 streaming (in Brink).
Have you actually played the game on the 360? The popping and streaming is horrendous.You won't even see a top-level mip unless you've stared at a surface for a few seconds.
hm...
Seems kind of nuts using 4xMSAA with LPP, but I suppose Blur did it too. I wonder what their g-buffer looks like at each phase though. :s
edit: http://www.eurogamer.net/articles/digitalfoundry-the-making-of-shift-2?page=2
Ah right, forgot about this.
http://www.gamesindustry.biz/articles/digitalfoundry-under-the-bonnect-shift-2-part-2?page=2 (gonna need a login for the rest)
Seems kind of nuts using 4xMSAA with LPP, but I suppose Blur did it too. I wonder what their g-buffer looks like at each phase though. :s
edit: http://www.eurogamer.net/articles/digitalfoundry-the-making-of-shift-2?page=2
Ah right, forgot about this.
djskribbles
Legend
FWIW, here's LoT's screenshot comparison.
http://www.lensoftruth.com/head2head-l-a-noire-screenshot-comparison/
Does the PS3 version have SSAO? Other than that, they look very similar.
Performance edge seems to be in the PS3's favor judging by reviewers comments so far, so it seems like it's the version to get if you own both consoles. But the 360 version seems perfectly fine as well.
http://www.lensoftruth.com/head2head-l-a-noire-screenshot-comparison/
Does the PS3 version have SSAO? Other than that, they look very similar.
Performance edge seems to be in the PS3's favor judging by reviewers comments so far, so it seems like it's the version to get if you own both consoles. But the 360 version seems perfectly fine as well.
Last edited by a moderator:
Similar threads
