peter_spy

One more thing, on DX 11 single thread test I was able to get around 35000 drawcalls before framerate dropped. On the same machine I can't get past 3500 DCs in TDM. Isn't it OpenGL limitation then, rather than the hardware?

I don't think it's drivers, as I already saw how performance changes on different CPU/GPU setups. Good thing is, it seems all parameters (tris, shadow tris, and now DCs) seem to scale with hardware. The only problem is that people usually change mobo / cpu / ram combo much seldom than e.g. GPU. That's why I see better results for tris / shadow tris with better graphic cards, rather than any change with DCs.

Well, that's interesting, because I cannot seem to "break the rule", regadless of hardware I use. Framerate always drops with DCs around 3500. That said, the "newest" CPU I have is already 5 years old. I wanted to get new mobo / CPU / RAM combo this year, but prices have been... unfavourable, for quite some time now.

I think we've talked about that a bit before, but is it actually possible to increase the number of drawcalls per scene? I can't find that exact entry in the Wiki, but I'm almost sure it stated that the limit was 1500 for Doom 3 and then it has been raised for TDM to around 3000. Is it possible to raise it again? Current hardware can handle much more than that, and because of the way TDM content was/is created (a lot of tiling materials), it would be much easier for mappers to create detailed scenes, with the limit raised to e.g. 10k.

I'm not super proficient in entity classes yet, but IIUC, it should work like actor class tree in Unreal and other engines: you start from general and go to specific. So from light to a certain light, then to light with certain model, and then to light with certain model and certain particle.

50 000 units equals roughly to 1,3 kilometers. Not sure how you would have to use this space to make something playable, and with high graphical fidelity, before hitting any performance limits first.

AFAIK, brush will always split whenever they encounter a vertex, edge, or face of another brush. The only way to avoid this, is to replace brushes with static meshes, and use brushes only as an outer seal for your rooms (with caulk). That's one of the reasons why modern games moved to static meshes exclusively. Brushes are used only in prototype phase and then replaced by models.

That might eat up performance pretty quick too, both because of number of emitters and particles generated, and/or because of transparency overdraw. A lot depends on how well the emiter is made. One or two static particles would probably be fine, but I imagine placing 2 or 3 such chandeliers in a scene might cause a slowdown. Edit: too slow one particle would probably be a good solution.

Yup, TDM updated in real time with what you do in DR, but with game time stopped.

It would be awesome for DR to actually use TDM renderer, but I too thought it would be way too much work. Especially correct lighting, fog, and particle preview would help a lot. That said, I have sort of workaround for fog light bug: I just copy it and delete it for light placement phase, so lights in DR show up correctly. When I finish placing and adjusting lights, I use Paste to get my fog back, and then tweak things using the game itself.

I was working on a marble floor material for my main location, but that didn't go well. Can I perhaps interest you in a wet rock instead?

And for reference, this model is 4000 triangles (for both sides of the entrance): And as you see it's not super efficient, there are some loops and divisions on the arch that could be removed, as they don't contribute to the overall shape much. In that regard, 3000 tris for a simple arch seems like much. Maybe you should use some more general guidelines, f.i. I rarely use more than 16-24 sides for a cylinder, as smoothing group does the rest for me. That would mean 8-12 sides for an arch. Typically that number goes down for smaller details, f.i. you can make thin bars with rectangles and one smoothing group, to fake a cylinder shape. But these have to be pretty small, so player's won't notice that. All in all, it's good to establish a general number for your big architectural pieces and go down from there. Modern games use similar, slightly higher values, so if you feel like 24 sides is still blocky, you can start from 32 etc. Also remember that low number of tris/polygons isn't for performance reasons only. If you plan to fully unwrap your models, that also means less polygons to deal with during the unwrap phase, which means less time spent on that (which is always better). What is probably more important, low number of tris will also mean that you can use your model for casting shadows, instead of making a separate shadow mesh, and that is definitely a time-saver.

Stats like that don't tell much by themselves, as model triangles are multiplied by the number of lights you got in the scene, ambient_world included. If you want to know whether you model is efficient, show us your topology, either in modeling app, or in game with showtris command. Ideally, you'd want to use only as many polygons/triangles as you really need to convey a certain shape, and fake the rest with your smoothing group and normalmap. Another important thing is the number of materials, you should keep it as low as possible too.

Huge collection of photoscanned medieval doors: https://sketchfab.com/demoon/collections/ref_medieval_building_portada

Pavillon concept for a display piece:

Going through a whole map might be a bit long, but talking about creative process and sharing tips on creating stuff, like some particular impressive effects in your map or general ideas, that should be handy. Any technical advice for other mappers would be a good idea IMO. I feel like there's not enough knowledge about making models, materials, and that kind of stuff in the community, and I'd be happy to share that knowledge too.

I'm not sure if this is particle editor bug, or just me not using i correctly, but it looks like the editor preview and the ingame version require different values. In my case, I needed two particle cards, intersecting each other, like here: But to make them look like that, I actually had to create something like this in Particle Editor: Particle definition, in case I did something wrong: { count 1 material textures/do/fx/candle_flame01 time 999.000 cycles 0.000 timeOffset 0.000 bunching 1.000 distribution rect 0.000 0.000 0.000 direction cone 90.000 orientation y speed "0.000" size "2.500" aspect "1.000" angle 90.000 rotation "0.000" randomDistribution 0 boundsExpansion 0.000 fadeIn 0.000 fadeOut 0.000 fadeIndex 0.000 color 1.000 1.000 1.000 0.000 fadeColor 0.000 0.000 0.000 0.000 offset 0.000 0.000 0.000 gravity 0.000 entityColor 1 } { count 1 material textures/do/fx/candle_flame01 time 999.000 cycles 0.000 timeOffset 0.000 bunching 1.000 distribution rect 0.000 0.000 0.000 direction cone 90.000 orientation x speed "0.000" size "2.500" aspect "1.000" angle 90.000 rotation "0.000" randomDistribution 0 boundsExpansion 0.000 fadeIn 0.000 fadeOut 0.000 fadeIndex 0.000 color 1.000 1.000 1.000 0.000 fadeColor 0.000 0.000 0.000 0.000 offset 0.000 0.000 0.000 gravity 0.000 entityColor 1 } }

Pre-recorded answers to questions may be an idea as well, although a talk is always more dynamic than an interview. For our podcasts we used Skype, with everyone recording their track locally (we did a "sync clap", like in movies, to facilitate mixing). I've heard Google Hangouts aren't bad for that purpose as well. I've seen your videos Obs, and I didn't have problems understanding you. Sure, like most of us non-natives, you have accent colored by your mother tongue, but that's how it is. Guys from Flipped Normals are Danish and Swedish, AFAIR (they live in London though). There are people with far worse accents giving GDC lectures: Mappers from all over the world shouldn't be afraid of taking part in the podcast after seeing that I only thought that the host could be English native, but I guess that's less important than e.g. being on the TDM Team (with additional perks of knowing what's going on behind the scenes right now, giving the audience a sneak peek etc.)

If you're collecting images of citysections, you'll actually be better with google maps and street views, as you can get inspired by real-world thing, and not somebody's interpretation of that. Still, games are good to look for modeling tricks, gameplay ideas, or e.g. interior layouts, as those are harder to come by.

I was thinking more about general model fidelity and texturework, than differences in style, but it's true: Carnaca is more "exotic" setting (Latin America, etc.) than Dunwall, which is only present in first and last part of the game.

I have a folder like that as well, but IMO those assets aren't that impressive anymore. Quite a few years have passed, and stuff like Dishonored 2 happened

I don't know about you, but I like listening to people while modeling or mapping. I recently found out two very nice channels, Flipped Normals and Environment Art Podcast: https://www.youtube.com/channel/UC_3bpU4AlXIJJS7KhIQcOgQ https://www.youtube.com/channel/UCQQFz2LkJMoqSFTas-oozOw They're directly related to making content, and you can find lot of interesting points there. Also the people who make those are pretty enthusiastic about what they do, and I must sat it's quite contagious. They effectively make my mapping time better. Now that got me thinking, how about us having some sort of TDM podcast? It doesn't have to be super frequent, an hour once a month would do. A host (Springheel, Spooks, Biker?) would invite different mappers to talk about what they're working on: design, gameplay, and content creation aspects. A commentary on recent TDM releases would be nice too. If there aren't any new ones, we could talk about older missions, other stealth games, or just games we're playing right now; or things we've watched lately, and possibly related to mapping ideas. I've been recording podcasts for several years, so I'm familiar with tech side of things, but I doubt I'm good enough to speak fluently in English (would need some practice, otherwise would stutter a lot). Still, I can help with mixing or making stuff for background videos, if this was to be released on Youtube. Apart from having something nice and informative to listen to, wouldn't that help to advertise TDM a bit more?

That might be the case, I had 60 fps at all times (vsynced, triple buffering).

I used the stopwatch app in my phone. Haven't checked all arrow types though. Yup, I can confirm this for Volta II, 1.8 sec or so. In intro missions (New Job, St. Lucia) it's around 2.6 sec again.

Not sure how you measured the bow draw time, but for me it's 2.4-2.6 sec for both builds (rope arrow).