Metal Gear Solid 3: E3 2K4 Video
- Thread starter Farid
- Start date
David_South#1
Banned
Vysez said:_phil_ said:
One of the most impressive thing was that cloud of grasshoppers (or something else) on a bridge.
The flamethrower is also something worth to see.
After checking some stuff I think they're bees trained using a conditioning responce system.
http://www.honeybee-research-berlin.de/column/research.html
I hope some of you caught the name at the end of the trailer. ~ Kuwabara
http://forum.pcvsconsole.com/viewthread.php?tid=10841
I checked out the IGN MGS3 videos, and the game is still rife with texture aliasing. I wish more devs would use trilinear filtering on PS2, because it really DOES work - as witnessed with Jak2 for example. Well, for all I know Jak2 might be the ONLY example!
Well, anyway, it works.
So why aren't they doing it? Konami coders = teh dumbasses? Don't think so...
Well, anyway, it works.
So why aren't they doing it? Konami coders = teh dumbasses? Don't think so...
^ ^ u mean mip-mapping...?
Guden Oden said:I checked out the IGN MGS3 videos, and the game is still rife with texture aliasing. I wish more devs would use trilinear filtering on PS2, because it really DOES work - as witnessed with Jak2 for example. Well, for all I know Jak2 might be the ONLY example!
Well, anyway, it works.
So why aren't they doing it? Konami coders = teh dumbasses? Don't think so...
I think because there's a tradeoff. Is Jak2 texture heavy?
PC-Engine said:Guden Oden said:I checked out the IGN MGS3 videos, and the game is still rife with texture aliasing. I wish more devs would use trilinear filtering on PS2, because it really DOES work - as witnessed with Jak2 for example. Well, for all I know Jak2 might be the ONLY example!
Well, anyway, it works.
So why aren't they doing it? Konami coders = teh dumbasses? Don't think so...
I think because there's a tradeoff. Is Jak2 texture heavy?
I guess it all depends. Cutscenes have very detailed textures for a PS2 game, in-game i could say it's better than most PS2 games, but still PS2 level if u know what i mean. One thing is for sure, there is A LOT of variety, even though the textures might not be the most detailed ever. There are just a lot of them to choose from.
And i do think GO was talking about Mip-Maps, if he was complaining about texture aliasing.
It has nothing to do with filtering, it's an issue of how hardware selects miplevels on pixel level - which is what gets rid of texture aliasing (assuming the game is using mipmaps).Guden Oden said:
To solve this "correctly" you would need to recalculate mip-koeffiicent and set an extra register on per polygon basis, which is both bandwith and calculation overhead (not to mention it adds mess to your shader code).
Trilinear will get rid of mipmap banding (the dreaded horizont blur line from many DC games).
Mipmapping will not fix pixel crawl in textures alone, no matter how good the hardware is at selecting mipmaps if the filter algorithm used is bilinear. It'll be full of "crawling ant" pixels anyway as soon as a texture undersamples.
With trilinear, the situation will be much much different... And, it doesn't seem to me as if MGS3 mipmaps at all, textures get that peculiar interference pattern that is typical with undersampling that you can see in Silent Hill 3 for example by hopping out the bathroom window at the start of the game and then looking along the brick wall as it stretches down the alley. There aren't any mipmap boundaries shown on that wall, and neither are there in MGS3 that I could see.
Jak2 however shows it IS possible to do smoothly filtered textures on PS2, even with dense geometry, so technically it can be done.
With trilinear, the situation will be much much different... And, it doesn't seem to me as if MGS3 mipmaps at all, textures get that peculiar interference pattern that is typical with undersampling that you can see in Silent Hill 3 for example by hopping out the bathroom window at the start of the game and then looking along the brick wall as it stretches down the alley. There aren't any mipmap boundaries shown on that wall, and neither are there in MGS3 that I could see.
Jak2 however shows it IS possible to do smoothly filtered textures on PS2, even with dense geometry, so technically it can be done.
Guden Oden said:
Actually, Guden, you are talking about a problem that can be fixed with Mip-mapping. Trilinear will only get rid of the step between different mip maps.
Trilinear on its own will not make texture aliasing better. Mip-mapping on the other hand will.
Actually London, since trilinear *requires* mipmapping, and works by interpolating two mipmap levels, what happens is that crawlesy-antsies are greatly/virtually eliminated (on hardware that does full mipmap interpolation - curiously this does not include geforce fx series chips).
As the viewing angle increases, rather than getting antsier and antsier, it gets blurrier instead which I actually much prefer.
As the viewing angle increases, rather than getting antsier and antsier, it gets blurrier instead which I actually much prefer.
Guden Oden said:
Your point in your initial post was "I hate texture aliasing".
Solution is: Mip Mapping.
Easy enough?
Yes it will - proper selection of miplevels is all you need to get rid of texture pixel crawl.Guden Oden said:
However, if you use bilinear alone you will get banding across mip-levels(very obvious mip boundaries), and mip-popping if polygons are facing you directly and move forward/backward slowly enough. Trilinear works to fix these transitions between mip-levels, it has pretty much no effect on texture aliasing.
Actually this has been described before on the forum, but I don't feel like working right now, so you're in luck
PS2 HW implements a simplified mip-selection function on pixel level that doesn't take into account pixel coverage - as a result polygons with high inclination will "always" have texture aliasing if you don't fix things manually.
As I've said in previous post, it's possible to recalculate inclination factor for the mip-map function on per-polygon basis and correct it(which will give you results equivalent to that on other hardware), but it will cost you.
Because of the asociated cost and added complication of your shader programs, most people will just try to fudge the factor looking for a value that works best "most of the time", and perhaps manually correct it for certain objects that have worse issues then others.
Now, I don't know what exactly Jak2 does, given that their camera is limited to certain type of motion relative to terrain they have an easier time to just fudge factors for most cases then say, a game with full freedom of motion would.
To speculate a bit further based on what I know of their tesselation tech - I think it's likely that they could also correct the factor on level of their T-Frag(or whatever they are called) primitives, which is groups of several (up to 16?) polys, which is MUCH cheaper then doing it for every poly, while the accuracy should be pretty close, depending on topology of the mesh.
You will get the same pattern when mip-factor is tuned for wrong inclination. It will effectively hide mip-boundaries too though, since they are usually pushed towards material edge (not doing much good).
It's also safe to say that JAK2 is saved by the fact that MipMapping blurs the textures quite aggressively after a certain level, which makes texture aliasing almost inexistant...
Trilinear filtering only helps keeping the transitions "natural". Although i still saw some parts with mip-map transitions.
That's why Jak2 looks so "clean" compared to most other PS2 games.
Trilinear filtering only helps keeping the transitions "natural". Although i still saw some parts with mip-map transitions.
That's why Jak2 looks so "clean" compared to most other PS2 games.
marconelly!
Veteran
If MGS3 will have the same image quality as MGS2 or ZOE2 that fine by me. I was never bothered by any annomalies in them, and while I do see some texture aliasing in MGS3 videos, it doesn't seem to be nearly as often or enough to be annoying to me.
marconelly! said:
IQ in MGS2/ZOE2 was OK...
Best yet is still Jak2. Or any of the pro-scan games of course.
Fafalada said:
Then I've never seen any PC game with proper mipmap selection EVER, because there is definitely pixel crawl on polys with high inclinations unless trilinear is enabled. Of course, those polys turn to a blur instead if aniso filtering isn't activated also, so it's not perfect. Still way better than the crawl though.
Actually this has been described before on the forum, but I don't feel like working right now, so you're in luck
Hhahaha, okay, thanks a lot for the most excellent explanation on PS2 intricacies. I must have missed that other time it was talked about because the details were new to me.
WHAT, you mean you don't disassemble other people's code to learn their secrets?
Anyway, there is the freelook mode too of course where the camera behaves like in any FPS, and I can't say I noticed more aliasing there than I did in the regular game. Maybe they use different techniques in the two modes.How DOES that work anyway? Models don't tesselate, they just pop between low/high poly version (in a very obvious manner I might add), but buildings seem to be handled much more smoothly.
Material edge?
When moving forwards/backwards, it's usually super easy to notice mip boundaries. However, I can't find any indications at all of mipmapping in that particular part of SH3. Also, since red bricks tend to turn into brown muddy crap in the smallest mips, it's usually rather easy to spot, but in SH3 the texture looks exactly the same all the way as it sparkles along the wall until it fades away into the background fog.
^ ^ because i don't believe SH3 uses Mip-Mapping. And if it does, well it's no JAK2. Oh, Faf explained.
Not sure about PC games right now(kinda used to seeing only aniso on PC these days), but in many DC games it was quite common to have next to no texture aliasing with no trilinear.
The polys still turn to blur, and the line where it happens is a real line, not a soft transition like with trilinear, but that's about it.
It's usually faster to wait for them to write a paper about it
It's a simple subdivision scheme, where each primitive is progressively tesselated from 1-16polys (or something to that end) depending on distance. Don't know how extensively they use it though, in Jak1 I know it was used for terrain and water among other things.
Sorry, some things aren't easy to explain in English for me
What I meant is the same thing you get when playing with LOD bias on PC cards - you can push the mipboundary farther away, at the cost of getting more aliasing. At some point the boundary won't even be visible anymore.That said, SH3 seems to use mipmapping selectively, so some surfaces use it and others don't. But even with mipmaps, the floor can still get that kind of aliasing thanks to the way GS handles things.