*merged/spinoff* for the Neverending Killzone Discussion on Graphics

[Lucid_Dreamer]

I don't believe he ever said that textures don't get blurred at all. And frankly, he could tell us a lot of things under the guise of not being allowed to talk about the implementation.

All mlaa based games are less sharp due to the simple nature of current mlaa, it will soften texture details to an extent.

Argue with that. That's post #97 in this thread.

^ yeah it's this level of ignorance that I'm talking about. Has no place in these forums.

Edit:

Oops sorry, thought I was posting directly under Lucid, he's the one I'm referring to.

Backing up what you say is not ignorance. I proved what I said. Are you saying proof has no place in these forums? BTW, where is your proof, again?

 

[T.B.]

Haha! You see that's a really good point. Can our MLAA blur textures? Sure. Does it have to? Nope. Can *anyone* judge if I'm totally making this up, or what it really means? Well, yeah, everyone who has access to the Edge Perforce server, but there are not many of those on this forum.
(To those who are: Hi!)

You guys should really, really relax a bit. This is a public internet forum. A place of fun, speculation and conjecture. Nobody here will ever be able to make any statement with any level of authority unless he's actually seen the code, talked to the engineers, and *understood* their reasons for what they did.
Very few people who do know can talk about things in public at a level of detail that makes judging stuff possible.

Or, to quote the great DeanA, "Yes, but then again you have no idea what they are *really* doing."

And with this, I'm out of here.
See you next time I just can't shut up.

 

[patsu]

^ yeah it's this level of ignorance that I'm talking about. Has no place in these forums.

Edit:

Oops sorry, thought I was posting directly under Lucid, he's the one I'm referring to.

Heh heh, see what I meant by reactions and emotions when people lashed out quicker than you can type a proper response ? ^_^

Also, reading his post again and I don't know where he got most of his.........delusional comments from. His claims about me are unfounded and flat out retarded.

Aye, it's a combination of all our responses.

You should read some of your own posts, sometimes. I see you like to take things out of context and fabricate your own point from misinformation (like you did with my post about the "virtual particle collisions" in another thread). My post about clock cycles was to back up my point about sometimes just having more cycles to use appose to always having to compromise on features. However, clock cycles can effect the direct criticism (insinuating that it's false and trying to downplay the technical information) of the "3x polygons of KZ2" in KZ3 made by Ranger.

I think kagemaru said he skimmed through the thread in broad stroke. So he might have missed some of your points, and caused more misunderstanding. OTOH, it's probably better to calm down before posting.

 

[corduroygt]

T.B. have there been any checkins/checkouts to your "area" in the Perforce server since March 2010? Can you at least tell if the MLAA code in GOW3 is identical to KZ3, and how about UC3?

 

[patsu]

Let me guess… like I said, an uneducated one:

Haha! You see that's a really good point. Can our MLAA blur textures? Sure. Does it have to? Nope.

You'll need more memory for the second case (Don't blur textures).




… and he chickened out.

 

[kagemaru]

I think kagemaru said he skimmed through the thread in broad stroke. So he might have missed some of your points, and caused more misunderstanding. OTOH, it's probably better to calm down before posting.

I kept up just fine in the original KZ3 thread and made my posts there. None of my posts or comments have anything to do with this thread, I'm not going to assume things until I have properly read what I'm replying to. :smile:

I understand why the post got spun off, but it's obviously created some confusion to what discussion I'm referring to.

I'll sum up my point:

Saw some legit critisism over the lighting in KZ3 from someone who I think is a level headed poster. Few members come up in KZ3's defense. My post about the ps3 was related to all of that and all of "that" happened in the other thread, not this one.

I personally have never really posted much about KZ3 other than my opinion of MLAA and how I prefer the hazy look QAA provided. I think the game look gorgeous and have said it before.

Now we have Lucid and his drivel. I'm not misunderstanding anything, he just doesn't have a clue to what he's talking about.

 

[Lucid_Dreamer]

I think kagemaru said he skimmed through the thread in broad stroke. So he might have missed some of your points, and caused more misunderstanding. OTOH, it's probably better to calm down before posting.

You're right, Patsu. I shouldn't let misinformation and personal attacks influence my posting, but unchecked misinformation tends to get to me. If I'm misinformed, I expect the evidence to show that. Then, I will apologize. It seems that others don't follow that rule. I just don't understand why. I thought that's what this forum was suppose to be all about.

Now we have Lucid and his drivel. I'm not misunderstanding anything, he just doesn't have a clue to what he's talking about.

Yet, you have provided what evidence to support your theory? It seems you are the one without a clue, since you have only posted words. However, if I don't have a clue, you should be in that same boat with me. Maybe, I should be on the deck and you can be in the galley. Deal?

 

[joker454]

Haha! You see that's a really good point. Can our MLAA blur textures? Sure. Does it have to? Nope.

You just made things confusing again Although it would explain what I'm seeing. I just tried Gow3 and it's mlaa does not seem to be quite the same as say LBP2's. Of the two games, I notice more overall texture softness and less effective edge smoothing in LBP2, whereas Gow3's mlaa implementation is way better. KZ3's seems to be somewhere in the middle. I'm guessing that the level of aggresiveness at edge smoothing can be tweaked to suit amout of available spu time? Likewise for any ill effect on texture details, it can vary from none to some depending on available spu time and perhaps memory? If you can't answer that directly then just say "Hey the weather is nice in Cleveland today", then no one will know that you leaked anything

Or how about this, when confronted in a dark alley by a mean looking individual who looks you dead in the eyes and states:

1) mlaa makes use of normals to locate edges
2) mlaa makes use of depth to locate edges
3) mlaa can be made to have no ill effect on texture detail but it takes more spu time and/or memory

Would you answer 'yes' to any of the above?

 

[JB9861]

GoW3 is probably way better because they apply MLAA after blurring.

 

[jlippo]

GoW3 is probably way better because they apply MLAA after blurring.

This should be easy to check.
If it's done after blurring effects, everything affected by DoF is quite jagged when compared to MLAAd surfaces. (due to the magnification of non AA surface.)

 

[Turok]

Argue with that. That's post #97 in this thread.

Ct03 was referring to T.B when he said that "textures didn't get blurred at all." One user suggested you brush up on your reading comprehension, and given your attitude in the thread, I'm inclined to agree.

Ignore 1080p, 7.1 LPCM, DTS, true S3D, Sony's MLAA, more techniques happening at the same time, more high rez textures, etc. Geez! I'm not so gullible.

Which games display a 1920x1080 resolution, while using stereoscopic 3D, MLAA, and 7.1 sound at the same time?

 

[Lucid_Dreamer]

Ct03 was referring to T.B when he said that "textures didn't get blurred at all." One user suggested you brush up on your reading comprehension, and given your attitude in the thread, I'm inclined to agree.

Actually, it's the other way around. Maybe you should consider brushing up on reading comprehension. Mine is just fine. I posted about this, first. My post was in response to "All mlaa based games are less sharp due to the simple nature of current mlaa...". That earlier post had nothing to do with CT03. CT03 responded to that post by saying , "I don't believe he ever said that textures don't get blurred at all." T.B.'s first response was "Nope." His second response was "Can our MLAA blur textures? Sure. Does it have to? Nope."

Maybe you were reading out of sequence or something, because that's pretty cut and dry.

Which games display a 1920x1080 resolution, while using stereoscopic 3D, MLAA, and 7.1 sound at the same time?

Did I say something about using all techniques at the same time or using more techniques at the same time? This is ironic, isn't it? BTW, HDMI 1.3 is only suppose to be capable of 720p/60fps in true S3D.

 

[JB9861]

This should be easy to check.
If it's done after blurring effects, everything affected by DoF is quite jagged when compared to MLAAd surfaces. (due to the magnification of non AA surface.)

You won't find such a stark contrast in the game because another blur conceals aliasing. So no it doesn't hinder MLAA's ability to detect edges.

 

[joker454]

Actually, it's the other way around. Maybe you should consider brushing up on reading comprehension. Mine is just fine. I posted about this, first. My post was in response to "All mlaa based games are less sharp due to the simple nature of current mlaa...". That earlier post had nothing to do with CT03. CT03 responded to that post by saying , "I don't believe he ever said that textures don't get blurred at all." T.B.'s first response was "Nope." His second response was "Can our MLAA blur textures? Sure. Does it have to? Nope.

He is saying that mlaa with no texture detail loss is possible and I believe him (only on the PS3 version, not the Ati version) since he is on the mlaa team. But we still don't know if any existing games are doing it that way. Just like games with 16x af are possible, but we have possibly never played any with it due to various reasons. In the meantime I see softness in mlaa based games and I want to find out why it's there. I presume that some implementations like LBP2 aren't using the full monty of the mlaa code, maybe because they didn't have enough spu time or whatever. But I'd like to confirm that if we can. I'm guessing that in it's most simple form maybe the mlaa library goes only by color buffer and has a smaller sampling set. That version would take the least amout of spu time and be the easiset to drop into existing games. Likewise I'm guessing the best version also takes normals and depth into account with a large sample set, hence needing the most spu time and likewise being the most complicated to drop into a game as you have to make sure all your buffers are setup in a way that makes then easy to funnel over to spu. Maybe they have other settings in the middle to fit whatever your engine, memory and spu profiles are to make it as easy as possible to drop it into existing games.

But that's just my guess going by what I see in current mlaa based games. If anyone else knows otherwise or has other speculations feel free to post as I'm curious to know what's going on with the current crop of mlaa games because they clearly do not look the same to me.

 

[patsu]

Ask your ex-colleages. ^_^

They have the library and documentation.

 

[joker454]

Ask your ex-colleages. ^_^

They have the library and documentation.

Only know one that may be toying with mlaa. Plus once I do ask means I have to step away from this thread.

 

[patsu]

If secrecy/NDA is the concern, I doubt you'll get more from T.B. in a public forum. You have better chance asking creative questions behind the scene.
[size=-2]… and then come tell us.[/size]

 

[Shifty Geezer]

I presume that some implementations like LBP2 aren't using the full monty of the mlaa code, maybe because they didn't have enough spu time or whatever.

I can't say I've noticed texture blurring in LBP2, but as you say it's MLAA results aren't comparable with other titles, so they are definitely doing something different.

One way to selectively target edges, which isn't at all complex, would be to select triangle edges as a mask. Do a plain vanilla geometry pass, edge detect, and derive from that which pixels should and shouldn't be post-processed. Or perhaps, MLAA the whole scene but mask the results so only the edges are composited with the unprocessed original.

 

[Turok]

Actually, it's the other way around. Maybe you should consider brushing up on reading comprehension. Mine is just fine. I posted about this, first. My post was in response to "All mlaa based games are less sharp due to the simple nature of current mlaa...". That earlier post had nothing to do with CT03. CT03 responded to that post by saying , "I don't believe he ever said that textures don't get blurred at all." T.B.'s first response was "Nope." His second response was "Can our MLAA blur textures? Sure. Does it have to? Nope."

Maybe you were reading out of sequence or something, because that's pretty cut and dry.
.

You quoted CT03 and then quoted Joker, when CT03 wasn't referring to Joker at all, but rather TB.

 

[patsu]

One way to selectively target edges, which isn't at all complex, would be to select triangle edges as a mask. Do a plain vanilla geometry pass, edge detect, and derive from that which pixels should and shouldn't be post-processed. Or perhaps, MLAA the whole scene but mask the results so only the edges are composited with the unprocessed original.

I was thinking along the same line. Also, can they save time by keeping intermediate data when the SPUs cull the triangles ?