Proper patch notes to come at a later time, but the major updates are as follows: Local Server multi-threading across planets and AIs LZ4 compression for server mods AI improvements and multi-threaded Client and Server improvements Some details put together by mikeyh here: https://wiki.palobby.com/wiki/PTE_Testing
This sounds pretty exciting, hoping the "pte" pushes to "stable" as soon as possible, so everyone's experience can benefit from it (assuming no new bugs are found)
I always thought multi-threading per planet seemed like a low hanging fruit in regards to optimization. I'm was a bit surprised that Uber didn't add it previously, although I don't know much about the previous multi-threading architecture. Any benchmarks or estimated performance gains with the new architecture?
I'd guess there are two reasons for this: a) stability concerns. Implementing multithreading that works reliable is a gamble b) server costs. To actually use the multithreaded server for online games hosted on uber servers they would need to basically multiply their server costs by whatever amount of cpu cores a single game will use.
Well the best I've managed is a game on Pax Prime (5 planets) with 5 AI in an FFA with no slow down at all during the game, so I'd say that was a big improvement
For b) in my benchmarking with linux servers the opposite was true. Multi-threading allowed the operating system to make better use of cores spreading the load... even when under stress. Memory usage also dropped due to other changes. The pressure point with multiple games per server is still typically going to be memory.