PinkDot on 23/5/2020 at 20:51

---

I wonder if anyone came up with an idea of creating a benchmark mission, so everyone could establish where they are in terms of the New Dark performance on their machine.

The reason behind it is, that with the increased capabilities of New Dark and new high-res, high-poly resources like texture or model packs, some players find certain FMs too heavy. Part of it can be blamed on the old architecture of the engine, partially it is due to the fact, that some of us still play on a relatively old machines (mine is like 10 or 11 years old now, with some minor upgrades along the way). Some people complain, that certain missions are unplayable for them due to custom resources or massive and detailed open spaces etc...
It would be a good idea if we could collect data from such a benchmark and know what percentage of the community can enjoy a mission, if it's made by somebody with a high-end machine, using all the heavy resources and complex brushwork.

The way I would see it would be something like this:
- a mission with detailed interiors and large exteriors (rooftops, forest etc...) - both with lots of objects
- extremely detailed architecture to push the New Dark limits - let's say 15K-20K of polygons on screen
- a few dozens of high poly AI (it's relatively easy to create ones by sub-diving the .3ds models in a 3D application and re-exporting to the engine - nevermind the look)
- lots of different textures (300+)
- particle and tweq effects - torches (model tweqs are performance eaters), some rotating stuff, fire FX, rain (or snow) and fog etc..

The mission itself would meant to be heavy on purpose, so let's say only on the bestest computers you'd get a decent framerate - the goal is to really push the engine and your graphics card.

Then the mission would contain a teleporting cameras set up (or camvators) for everyone to have consistent views/camera angles.

The benchmark would have to be run from Dromed with the show_stats enabled, so a person could read and make notes of the frame rates for each camera angle. I don't know if the frame rate could be displayed in actual game and I don't think it could be captured and output somehow through a script, so probably manual notes would be the only way to go. Bit awkward, but better than nothing.

The benchmark could be run twice:
- with all the vanilla (stock) resources only
- with one of the high-res mods (like HD mod or ESRGAN texture mod - to be decided) and Model Enhancement Pack

Then we could set up a thread with a kind of a poll with multiple frame rates ranges as answers, so people could vote for the average frame rate they got from the tests. This way we could quickly tell what's the community's average results etc... Of course, in the thread itself, folks could be more specific, giving us their machine specs and more detailed frame rates, let's say for each of the 10 (or so) cameras from the benchmark.

Results could be also provided in another way - maybe a pre-made spreadsheet, where you only fill in the fps fields for each test and then you get some nice graphs out of it, which you could post etc... And such results could be also collected by a person for further statistics etc..

I think this would help mission authors better fine tune their levels, before they get blamed for making a mission unplayable for 90% of people etc...

Let me know your thoughts. Personally I don't have a great experience with benchmarking, so hope to get some useful ideas.

PinkDot on 25/5/2020 at 10:22

---

Ah, that's a good news - thanks nightshifter.