chedap

(yet another thread about optimizing the map by turning details into func_statics) So, I have a somewhat complex interior scene with 1300 drawcalls that runs at about 50fps on a dated Core2DuoE6300/RadeonHD4890 machine. Apart from curved surfaces and models, it's entirely made from worldspawn brushes. Adding AI to the scene costs 5 fps per guard. Maybe I can improve that? I remember that detailed brushwork is bad for pathfinding and so I decide to spend a day taking said interior apart piece by piece, replacing visible sides of brushes with patches, projecting textures onto them and simplifying the sealing geometry to this: The actual scene is visually indistinguishable from before. At this point everything is still worldspawn, though aside from sealing brushes, everything is patches/models. On that same PC, viewports in DR start to stutter a bit (render window msec counter has doubled), but in-game it's still 50fps, still around 1300 drawcalls. The fps drop when adding AI (1, 2, 3 guards) is exactly the same as before despite a much simpler ws-brushwork. So, let's group patches into func_statics and add noshadows to them. Attempt 1: let's try to keep modularity intact. I group together walls consisting of panels + wallpapers + trim at the top. I group ceiling panels (6 patches each). Same with the columns+arches. Testing it in-game reveals a 10fps drop and 1800 drawcalls from the same spot. Huh. To be expected, I guess, since (I assume) each func_static calls for each texture independently. The scene allows us to add noshadows to most of the walls/ceiling, and that'll give us a whole 1 (one) fps increase. Eh. Attempt 2 then: to hell with modularity, let's just group all patches with the same texture to suit drawcalls better. Maybe then we can make trims non-solid. Testing in-game, 1450 drawcalls. While an improvement over modular approach, that's still 150 extra drawcalls (and 5 fps less) compared to just worldspawn patches, and turning select surfaces non-solid seems to not have any effect at all. So it all boils down to this: compared to my initial brushwork, I spend more time, have worse performance in DR, and gain no tangible benefits. Am I missing something else here? So far it seems to me that the effect of complex sealing geometry on performance is a bit overblown, and modular approach is quite inefficient unless you export all func_statics as models first.

See here. A driver rollback apparently solved it.

TDP had like 50% of missions take place in 'horror' setting, at least partially, so if anything, the word "classic" is not the right one to use. I wouldn't describe them as horror either, they're more like scenarios a thief-class character could find himself in (in rpg terms). As for the reasons they weren't as popular as burglary-style missions, it's probably just that the gameplay mechanics have a different focus overall (although without the character being fragile no horror would work, so at least there's that).

They're red on 2.01 (just checked). Are you sure it's not some test build stuff?

Another not very obvious thing is that only one surface of a visportal matters, make sure you didn't instead align nodraw with the doorways. Also check out Komag's on basic and advanced visportals.

Uhh, there was no need to restart for me. I was able to just load a quicksave and continue on, these random deaths didn't happen at some defined moment. Which, btw, suggests that it probably isn't a pathing issue, at least if you don't disturb their default patterns (I mostly ghosted).

Not sure where you got the idea of 512x512 being a limit, there are plenty of textures that are 1024x1024 (both diffuse and normal) among the TDM defaults, and I bet it's possible to go way higher.Both the statue and the door handles (that you mentioned in another thread) are objects player is likely to get very close to, so using 1024 isn't entirely unreasonable compared to, say, wasting the same 1024 squared on a spoon even if its size is about the same as a door handle. In general, you don't want new objects to feel out of place in an environment TDM is currently capable of. (I don't know the inner workings btw, so my advice doesn't hold any authority. It's just guidelines I use myself)

You should just try using it more. Increases blackjack and pickpocket range, allows to pick the locks just out of reach, and more.Can't think of a reason not to make chest base non-frobbable though, would definitely smoothen the experience.

Here, found it (on my own hard drive, no less). Edit: checked if it's shopped, just in case. It isn't, here's the source. Oh, sure, they aren't flawless and other than them being somewhat bloated I have my own grudges (you probably don't have to deal with their various regional restrictions). Comparatively though, they still are about as good as it gets, and unfortunately I think if GOG had to deal with big international releases alongside retail launch days, they'd either die or become as 'flawed' as Steam.

?

This is got way off-topic, but, wesp5, could you find that interview with Tim Cain about Activision 'freezing' final build of VtMB? Can't seem to find it and don't even remember how it looked like(some scanned page?). I bet you remember better. But yeah, I don't see how comparing Valve to Facebook is appropriate. Valve's finances aren't based on advertising and collecting personal data. So in context of VR it's not much of a dilemma of who I would trust more.

See, I felt almost the same when I first went through training mission and St. Lucia, but after effectively unbinding the block button I was able to take on up to 3 enemies. Just strafe around and back away, keep a thrusting attack (accurate and doesn't hit walls) prepared, wait for an opening and aim for the head. Try slaughtering your way through The Rift or Samhain Night for practice.

Got several mission fails when some unknown NPC dies. Thought it might be the guard by the well drowning, but the amount of damage shown in the console makes that somewhat unlikely. And minor stuff: out of bounds from the balcony and this very obvious invisible barrier (here I tried to measure the depth with a candle).

Got the "body with that name already exists" playing card error. Search found two bugtracker tickets with no updates (1, 2), I can provide the savefile if that could help. There's only one card player I can think of on this map. Might as well post minor issues. Out of bounds on builder church. Out of bounds between inventors and watch station. Some weird lighting in the lobby. And a tiny little piece of see-through geometry here. Great mission by the way! Proper thieves' highway and all that. The WS series is like a humongous cherry on top when playing in alphabetical order

Well you could also try going to ThiefEngine.ini under [systemSettings] and disabling ambient occlusion, DoF and motion blur. There are also values for shadow resolution, but you'll have to find suitable numbers yourself. Doubt it, it just makes higher-res objects and textures pop-in a bit more noticeable.

The thread-starting post has a fix that seems to cause undesirable effects. The one I've used is similar: go to Thief\ThiefGame\Config\ThiefGame.ini and under [Engine.WorldInfo] replacemFog=(mbEnable=TRUE, mbEnableOutdoor=TRUE, mbEnableColor=TRUE with mFog=(mbEnable=TRUE, mbEnableOutdoor=FALSE, mbEnableColor=FALSE Before/after. Some (hand-placed?) fog will still remain though, architect's house in particular.

TDM isn't an engine though, modified idtech4 is. TDM is that + a set of mechanics and goals + a large asset library + an imaginary universe. Which is more than a lot of things generally referred to as "games" have. I don't get why it's so hard that it needs 10 pages of discussion.

Temple of Tides FM for TMA had some sort of periscope windows into an underground tomb. But you should steal the idea anyway.

A change as radical as sending a pickpocketed guard into alert mode is likely to ruin the balance on already existing maps. Same with 'only pickpocket while walking' suggestion - there are plenty of situations when waiting to end up behind a guard between his patrol cycles is a strategic decision (noisy floor / brightly lit room / lots of NPCs around). I also don't think 'hightened senses' right after pickpocketing is something a player can expect without knowing about this mechanic beforehead, it's rather counterintuitive. Ideally, it would work like this: NPCs only notice their item is missing when they try to use it - obviously, most common case is reaching for a key to open a door, but there may be creative uses like reading a letter or something. Then their reaction would depend on their alertness state, afaik there should already be a system in place that gradually raises suspicion with each doused torch / door left open / noise (which don't by themselves warrant a 2+ alert state). If prior to discovering an item missing AI wasn't alerted, they could maybe enter a peaceful 'search for item' phase, wandering around corners of the current room (without drawing a weapon) for 5-10 seconds, with "must have dropped it" barks. In a perfect implementation, player would be able to drop the key nearby for the guard to find it and pick it back up to their belt, adding to full ghost challenge. Now, I don't know exactly how much work would that be, if at all possible. I'm not even sure how NPCs handle locked doors currently, I think I managed to lock some of them in, but guards seem to always be able to open locked doors even after they don't carry a key anymore. And some flexibility probably wouldn't hurt, either through menu options (AI vision and lockpicking are there, after all) or item flags in DR for mappers to use. I think he meant the lady behind the counter, with a "perfumed key".

One answer would be: it doesn't. It stutters on high-end machines. Also it's a modern engine that takes full advantage of GPUs and everything's done with models. You can observe how they budgeted detalization: they can afford 'redundant' geometry for individual bricks in cramped interiors, but exteriors are much more simple comparatively (pavement is just a believable use of parallax). Exteriors also have noticeable LOD-models and simplified geometry in places player isn't likely to notice. Examples.

This has been collecting dust on my hard drive for a while now: And watching it now I just realized that it isn't actually raining, yet that water flow suggests there's a flood going on. I assume there's no rain for performance reasons though. Still, you might want to make it more consistent one way or another. (also, moving that chair a bit from the window would probably be nice) As for suggestions for a facelift, I feel the commons quarter could use some more lights on the streets. Currently most walls are in complete darkness, making the scene somewhat flat. But other than that, I think it's a great mission with a nice flow to it, I don't think there's that much to improve really.

Patch 1.1 introduced SSAO+AA incompatibility when SSAO would render on top of fog, patch 1.2 fixed that. I didn't really notice problems aside from that, but then again I'm not very sensitive to flaws that only appear in motion, (screen-tearing, aliasing etc). I think it might be another effect of them layering desaturation+green filter+fog on top. It gives the whole scene unnatural feel, and dynamic lights don't behave the way you intuitively expect them to.You don't realize it's Fallout 3 levels of bad until you open a screenshot in an image editor. Example: in asylum, male/female wards have blue/pink line of tiles along the bottom of the walls. But in the game it all blends together as gradients of green.

Absolute values seem to only be available for individual faces, not for brushes. Grid size defaults to 2 for me on restart, unaffected by setting it in the menu. Don't know what's up with that.

Made an organized gallery of Thief 4 screenshots without shroud, fog & color filters obscuring the visuals: The City (exterior architecture) Interiors (from lower to upper-class) Uninhabited (service & tech, ruins etc.) (minor spoiler warning, I guess)

Helpful, thanks. I made them adjacent, but not intersecting, and what I got can I guess be described as odd and moving.Doesn't matter really, just that the tutorial is unclear. It's working fine with the global ambient light extended to the whole playable area. Seems like it can effectively triple the total amount of tris if the brushwork is complex enough, and all those tris will take part in shadow calculations. From my basic understanding at least. It's just a basic thing to put into habit before it's too late.