Performance is always a hot topic but it is difficult to address without having an accurate image of the problem. This month I setup logging of client performance and displayed them in various graphs.
This has already proven to be useful. We noticed that exclusive fullscreen had lower FPS than borderless. Rather than guessing, we were able to query the database and confirm it.
See below for various graphs to get an idea of the insights we are seeing & some interesting data.
First up is median FPS across all hardware. It doesn't give a full picture: players with 60hz monitors are included beside 144hz gamers.
A better metric may be comparing average fps to monitor refresh rate to see how many players are getting lower FPS than their monitor supports... something we will look at in the future.
Best Performing Hardware
These charts clearly shows how Rust is CPU bottlenecked (no surprise). What is interesting is how much the extra cache does: a 5800x3D is on average 50% faster than a 5950x!
Please don't use this as an excuse to go buy a better CPU as performance is subject to change... but do use this info so you avoid buying a new GPU for Rust then being surprised when it yields no performance improvement (at this time).
~75% of the playerbase is running 16GB or less of RAM.... perhaps we should really prioritize why Unity is hogging all the RAM.
The majority of players (73%) using 1080p... yet only 2.5% of players using 4k?!
Looking at the most popular CPU & GPU we find that it is similar to steam's hardware survey but differs slightly from Rust. Here are some GPUs as % of total:
- GTX 1650 = 6%
- GTX 1060 = 2.5%
- GTX 2060 = 5.5%
- RTX 3060 = 7%
- RTX 3090 = 0.6%
- RTX 4090 = 0.2%
30% of people using motion blur 👀 only 5% of players using SMAA, 20% of players using primarily windowed mode (not borderless)... all interesting. I thought more players would have turned shadows off.
Our Mac playerbase is very small, with only 0.3% of players.
Yet even stranger are the people still on Windows 8... laptops perhaps?
The next steps are to expand to gameplay metrics. Technically.... we already have some gameplay stats but they are very limited: "total wood gathered", "total kills", etc.
Better gameplay metrics will give us the ability to drill down to specific information such as:
- most popular weapon each day of a wipe
- win ratio of different weapon match-ups
- deaths per size of team
- hourly loot per monuments
- least used items