Tracks Dessau 1949-1956

When one door closes, another opens.
I returned to my second track project, which has been on hold for some years.
It's Dessauer Rennstrecke, a historic track of the 1950s located in East Germany.

I want to do all three major configurations.
I'm not sure about the 1955 configuration, could have been the B or the C layout.
Start/finish in C layout is uncertain too. In 1952 it was in the woods, for 1953 I have no information.

Footage is rare – I have about 100 photos – some things have to be guessed and I have to use some of the trackside elements for all layouts, although I have only evidence for a certain year.
I had quite good elevation data (with an accuracy of 1 meter) and a shaded relief, but especially the motorway access changed a lot. Fortunately the demolished road parts are still visible in Google Maps. Nevertheless there is a high probability for errors.

I have been there in 2016. You still can see the remains of the steep turn in the forest.

Here is a track map and first screenshots.
configurations.jpg
screenshot1.jpg
screenshot2.jpg
screenshot3.jpg
screenshot4.jpg
screenshot5.jpg
 
Wow, can't wait - guys on this board have made AC the best 'historics' sim of all time and an interesting circuit combined with the level of details you added to your previous project will make this an absolute pleasure, I'm sure.

Previews are looking good already, thanks.

You say "high probability of errors" but I say "even higher probability that no-one will know better and will enjoy it anyway"!
 
Last edited:
Wow. The atmosphere on your video is pretty epic!
Congrats on the scenery, looks really good
those 3 trunks make a big difference how they catch the light !!
I wanted to try that... how's poly count ? 3d vs 2d
what you're losing in polygons, you're gaining in space on your DDS, not wasting so many pixels for empty trunks ....
 
Last edited:
Status is as follows:
1952 layout: 70 % of scenery done
1949 layout: 40 % of scenery done
1950 layout: 30 % of scenery done

A first release is scheduled for Easter 2021, maybe for 1952 layout only.

Material count is 24.
The track has 1,760,000 vertices (target: under 2.5 million).
Number of objects: 5674 right now (target: below 1000, joining at the end).
FPS: between 175 and 200 on my new system, on my old system frames have been between 80 and 110 (similar to other tracks, despite high number of objects and polys).

I use different types of 3D tree trunks:
Low poly (far away): 8 tris
Medium (nearby): 32 tris
High (within close proximity): 220 up to 340 tris

Other objects:
Y-tree: 24 tris (standard KSTREE_GROUP type, with loop cut in the middle and pivot point at the bottom)
Tree wall: 12 tris (no shadows, a lot of transparency, no edit of normals, which is unorthodox but looks quite fine at the moment)
A bale of straw: 216 tris (maybe a bit to high?)
Car: 1000 (simple car, most) to 4000 tris (special things like fire truck, only a few)
All cars in scene: approximately 160,000 tris

Spectators have the biggest impact.
Most spectators are right beside the road, many of them within 1 metre to the edge of the road.
That's very special about Dessauer Rennstrecke. Texture planes are not suitable for that in my opinion.
Each person has around 250 tris.
That means: 250K tris for 1000 people.
25 million tris (impossible) for 100,000 people (real number of spectators back in the 50s).
My target: staying under 6000 people = under 1,500,000 tris = 1 spectator per metre

Of course it's possible to do a version with less 3D people (half?) and less cars (half?) for people having problems with performance.

I have no experience with LODs yet, could be a perfect solution too.
 
25 million tris (impossible) for 100,000 people (real number of spectators back in the 50s).
My target: staying under 6000 people = under 1,500,000 tris = 1 spectator per metre

Of course it's possible to do a version with less 3D people (half?) and less cars (half?) for people having problems with performance.

I have no experience with LODs yet, could be a perfect solution too.
You could spread the spectators amongst different _KSLAYERn's. With clever grouping (much like KSTREE_GROUP_n) you could get away with that much, maybe. Or indeed do some LODs.
At least worth a test, I reckon :)


Edit: I did a test with a few tri's. 17 million of them, and it hardly impacts performance on my system. And that's with everything visible at once.
20201207155150_1.jpg
 
Last edited:

Latest News

Are you buying car setups?

  • Yes

  • No


Results are only viewable after voting.
Back
Top