peter_spy

I don't know TDM assets much, but typically when you make that kind of ornamental stuff, you try to get away with some transparent cards where players won't notice, so you don't model all the complex stuff on geometry. It's more of a balancing act really. Hmm, r_glProfiling 1 works on 2.08. Did something change in the meantime?

@MirceaKitsune A simple illustration of how tris count grows with lighting. What I have here is an object in a caulked room that is around 380 tris. Look how the tris count grew in comparison to the original number. Especially since ambient + fog is one of the most basic setups, for outdoors in particular. The tris count is almost tripled there already.

That gate at the end of the corridor looks like it uses geometry for all the ornamental pieces instead of alphatest parts, so that might be a culprit as well.

Yup, you can use showtris 2 or 3 with showportals to see whether your portals work with your brushwork to cull the geometry: About lights, nbohr is right, they act as multiplier to your number of tris, every object hit by a light gets redrawn. Multiple and overlapping shadowcasting lights are the fastest performance killer. IIRC, the tris also get redrawn if you have additional material stages, i.e. other than diffuse normal and specular. You can also use that gl profiling command to see what takes most time, e.g. DrawInteractions is for lights, ShaderPasses is for materials.

Oh and one more thing about measuring performance, which might be worth a separate article actually: have you tried using frametime instead of FPS counter? IMO this is should be much more useful metric for mappers, mostly because you'll be aware how much overhead you have, at all times, as opposed to FPS going below 60 when something is already very wrong. To use this, in your video settings set your Vsync to Off, set Uncap FPS to On and use highest Max FPS available (400). Also bind a toggle r_glProfiling 0 1 to a key. When you toggle the profiling, you'll see the breakdown of rendered components, but you can ignore most of it and just focus on the uppermost value for the GPU: This is, roughly, how long it takes you GPU to draw one frame. If you want to have 60 FPS in your mission, you need to stay below 16,6 milliseconds. And this way you know, that when you reach e.g. 8,3 ms, you're at approx. 50% capacity of your hardware. Now, let's say you have a section which is barely detailed, but you've already going above 10 ms, then you know that something might be wrong, and it's time to check showprimitives to look for the possible culprit.

It's a bit hard to interpret these numbers without screenshots but the first one is indeed a bit high. Also, either none of your lights are casting shadows, or you have shadowmaps on, since shdw value is 0. I think it's safer to use Stencil Shadows for prototyping FMs, since what I gathered from talks with Cabalistic, shadowmaps are still a very unoptimized feature. If you want a nice compromise between quality and speed, set Soft Shadow Quality to Low or Medium. This way you will have some numbers in the shdw column which is also an important metric.

You mentioned that you don't use commands like r_showprimitives, you may want to use it to check if there is any tangible benefit. E.g. I bet that and these "virtual cubes" do nothing, as they don't have any brushwork to work with. You just seem to repeat what player frustum is doing already.

Last time I checked, and that was around TDM 2.05, TDM could handle 1,5 million tris in one scene on my hardware. That was an isolated test though, results may vary in real FM. You may hit polygon limits for single mesh/model though. I think I had a difficulty working with 30k tris .ase models, and editing their paths in notepad was a nightmare. That probably doesn't apply to .lwo. Still, that's probably more than most of us need.

I'm sorry but your tests seem a bit chaotic to me. Testing stuff in more consistent manner, e.g. each factor in isolation, and with better description, that would definitely go a long way.

How come the second picture has a slight bloom effect, even with bloom off? That looks wrong.

It seems like you've moved the bloom slider from the default position, first values from the left shouldn't give such weak effect. Either that, or it's the matter of color value for the blend add texture. Anyway, it will always be a relative effect to some extent.

Ideally TGAs should be used only for prototyping materials. And one of the downsides of TGAs is that they will be treated by the engine aniso filter, which is subpar to highest quality mipmap filters used with DDS. This affects both game world and 2d menus (e.g. inventory icons made with TGAs will always be blurry in comparison to high quality DDS). Perhaps texture compression options could make their way to video options, although IMO for players the texture size (image_downsize stuff) would be more important. Basically all games have texture quality options, and it's quite weird TDM never had it.

Okay, so the initial setup is 64-bit color on, bloom on, there's one light in the center of the room, noshadows. 3 objects with bloom material (nevermind the weird shape): 15 objects: As you see, the GPU cost is the same, CPU should be as well, maybe there was some fluctuation while I was taking screenshots. The GPU used is nvidia GTX 1060. Edit: Out of curiosity I packed 30 objects here, and that gave me some slightly CPU/GPU utilisation, but I think we can assume that this isn't the most viable scenario

That still looks odd, and the difference between 54 and 57 fps might be coincidental, especially with in-game fps meter, which isn't super accurate. MSI Afterburner tools are better for that (both the fps meter and the hardware load percentage indicators). Also, console commands like r_showprimitives and r_glprofiling should give you some more insight. With lights and emitters for every luminescent surface you will generate more drawcalls and texture calls, than with just one material that has emissive stage. Also, it would be interesting to see several light + emitter combinations in one line, with player looking at the first one, and all the other lining up behind it. That will should you the overview of the impact that particles transparency overdraw might have.

A little study in metals: copper, silver, gold. Tried to use as much as I know from the PBR workflow.

I'm not offended, I'm tired, by you. You use the same old tricks I've seen everywhere else, and ignoring most of the on-topic stuff I already wrote, so it really stopped being a discussion some time ago. All this stuff is exhausting and it distracts me from doing creative stuff that came up with the talks on the margin of this topic, and has a chance to be a positive contribution to the mod, regardless of the actual frob implementation. Stop replying to me, I really don't have anything more to say, and neither have you. Whatever you say next will be ignored by me.

Stgatilov is a TDM dev and he quoted me. So I replied, even though I'm just reiterating what I already said. Since you're just bored^ and already wasted tons of my time, you're not going to get the same treatment. Now excuse me, I have an exciting little project to start

Not all of them, but using them more selectively. Making all 20 bottles on the shelf interactive, just because you can do that in real life, isn't meaningful gameplay. Hiding something behind those bottles or leaving a few of them pickable, in case the player might want to use them as distraction item, is much better design. Nope. If you take a look at both old and new im-sims, it's not like each and every object is interactive. Both Dishonored games and new Prey have tons of static geometry, only stuff that supports the gameplay in a fairly meaningful way is interactive.

I can only help with proper loot items, and core mod stuff, to be precise. Adding more complex models and materials will help solve that problem (human eye will focus on elements that are more detailed or stand out more, *insert relevant 2D art theory*, etc.). Again, IMO you want a UI-based, one-size-fits-all solution to counter design mistakes; these are problems that can't be fixed that way, at least not without consequences I already spelled out. I think these consequences are serious enough to rethink that approach. But obviously, you'll do whatever you decide, I'm covered either way. Now please leave me out of the further discussion, I'm super tired as I spent most of my free day on it. I don't have anything more to say.

On the contrary, it's extremely relevant, because it's how other games deal with this, and TDM doesn't exist in a bubble. People playing TDM don't live under the rock, they play other games too, so they can carry over their expectations just fine. I do. And not only reskin them, but make them from scratch, with better materials and more loot variations. Some objects can replace existing assets, using the same dimensions and the same origin point placement, others will be made as new content. Preparations are already underway. I'm not promising to remake everything, but slowly adding stuff month after month is doable. I already got a preliminary list of objects to work on.

It's funny, because with what I just described you'd essentially have "my thief character knows what to pick up" thing, but being a fully intentional effort from designer, and without making significant changes to anything else. But I think this discussion has run its course already, at least I feel like I was forced to devote more time to it than I wanted to. Let mods cut the meta discussion and leave this thread for actual feedback.

Oh yeah, this has been brought up as a point several times already, but I don't really think graphical complexity is a problem. First off, TDM doesn't look that modern in comparison to other games. And, newest titles have been managing this quite well, with just being selective (and consistent) with objects that are within player agency. You can do the same in TDM: just place static models, or turn off frob for junk models, so they'll react to physics and weapon hits without being pickable.

I believe there's no symmetry here. I think addressing problems at its source (asset design) would be a better option. As I mentioned, there are talks underway, and from what I've seen in DR viewers, there aren't that many core loot models there; quite a few are just variants of one model. So a rework project might be more feasible than you think.

Okay, so let's get down to some details. Imagine a new FM, player walks into a room there's a table and a few brand-new custom objects: goblets, plates, and cutlery. There's a stationary/sitting guard next to a well-lit exit. All interactive objects have proper materials, most look plain and wooden, but there a few that look like silver. All are highlighted by the frob outline when up close, but with no additional info. Player scans the objects and thinks, "okay, silver is probably loot, and there are a few wooden items, so I'll pick up that wooden goblet last, and throw it a bit away from that guard, so I'll distract him and make for the exit." It's a fun mini-game of scanning the environment, trying to guess the rules, and then planning and executing the your plan. Player walks up to a new silver goblet, frobs it, and has a pleasant "yup, I was right, this is loot" moment, and a similar one for the wooden utensils. These small moments of uncertainty and risk vs reward planning are also part of a bigger theme, which Thief series always had, namely the idea of doing all kinds of stuff in a place that that you're not supposed to be in, which is both uncertain/scary and exciting. Now imagine player's thoughts in the same situation when the frob is gold for loot and blue for junk: "okay, objects on the table, blue, blue, blue, gold! press RMB!, blue, blue, blue, next room please." At some point you stop seeing objects, you just look for outline color as it does the job the fastest.

It sure can, and I actually have a recent experience with that, as I've been playing Nier: Automata lately. It was like "Wait a minute, I'm being manipulated here: this absolutely gorgeous music feels elevating and emotional, but the actual gameplay is fairly meh." Seems like they had a great composer, but not so great quest designers. Well, at least boss fights were awesome. So you risk a similar sentiment, plus you'll have to do several people's work yourself. Again, I have nothing against crap missions, we all need to start somewhere, it also shows us our progress. Nothing wrong with that. But trying to correct bad design is designer's choice. Plus I'm concerned about gameplay as it is a bigger change than you think, and it will have consequences, see my previous post. I'd have to post recommended frob shader settings, once I find values that I like. The list already includes stuff like bloom and 64-bit color, and the mission is optimized for stencil shadows + soft shadow quality set to low. It will probably work fine on higher settings and shadowmaps, but this is the setting I find most consistent, performance analysis-wise. I also have a draft of TDM menu rework for the mission, so I need guis to be editable, although this might be too time-consuming to be taken as far as I'd like to. It might end with just custom replacements for current low res backgrounds and effects.