My desktop is soon due for an upgrade. It’s held up pretty well but one area of pain is slow iteration with increasing project size. My last project took a minimum of 17 seconds to enter play mode. My current project is a lot smaller, but is already up to 5 seconds (an empty project is 2 seconds).
Profiling shows there’s no single culprit but a lot of time is spent in serialization and reloading assemblies. I don’t use lightbaking, so that’s not an issue.
I’ve experimented with assembly definitions and they helped, but not very much.
Current machine is an i5-6500 3.2ghz with 16GB RAM, SSD drive.
Question: If my #1 goal is fast enter play mode, what specs really matter? Number of cores? Single core performance? SSD speed?
Interesting! I’m not using ECS, but skipping domain and scene reload changed enter play mode time to effectively 0. It sounds like there are some side effects that might trip me up (non serialized variables not being reset), but if I can remember them, it would be worth it for the faster iteration time.
Choose an SSD which uses a DRAM cache, otherwise you can get poor write speed. Writing the currently open scene to a temporary location is one of the things which occurs when entering play mode.
The biggest task in Unity I’m aware of which really benefitted from higher core counts is CPU based lightmap baking, but you already said that doesn’t matter, and all the cool kids seem to be moving to GPU based anyway. Unity keeps getting better at taking advantage of multiple cores, but the main thread is still where things tend to get bottlenecked. So good single core performance is still very important in the Editor.
But the recent benchmarks coming out for Ryzen 5000 CPU’s seem to have ended the choice between single core performance and more cores, even the choice between AMD and Intel. It looks like you get the best available single core performance and just have to choose how many cores you can afford.