One question I don't have answer is if we can use tileable resources feature of DX11 somehow, that would allow much improvements as you could set one gigantic texture for all the buildings without hogging all VRAM or much any hogging actually, 8K textures or more with performance of 2K or less of texture size. re-use same objects few times more that city could cover whole map size of WCUSA.
It really comes to what map maker then wants to achieve, if on high details you could run a map with gtx1080 and better only, that would allow quite big city, for example from WCUSA if we remove forest, race track etc. With some good optimization I could make many times bigger city like this, there would be no problem at all, but there would be 'few' compromises: It is possible to build bigger city than WCUSA, hugely bigger actually, but it will always be a compromise of something, using batchable objects without textures will help a LOT, but you like to look solid colors instead of textures much? However you can bog down even 8086k with about 3000 drawcalls, because not each call is equal, you can put lot of heavy calls using many textures and low amount of polygons for each separate object, or you can combine some objects and textures as well as use game engine's batching of meshes to your advantage like in case of JRI.Ĭase of WCUSA is showing very well how game engine's batching abilities were not possible to utilize more, so to help CPU to chew stuff for GPU, larger objects were used as modern GPUs can chew 10 000 000 polys just fine, yet CPUs are still at same what there was 5 years ago. Now this is more generic stuff than any specific reply to anyone. is that engine can put only about 3000-4000 drawcalls out on most machines and you have to fit everything on sight to that.
Click to expand.You can go around of that problem, but then amount of ram is another thing, editor keeps crashing a lot with 16GB ram and when putting more than 60 000 or so objects on map.
0 kommentar(er)
