so i have a race map and if i set it to Atomic then the game just lags out forever, doesnt matter if it loaded, the ping is at constant 10k+ with huge lag spikes
and if its nonatomic then it works just fine no problem
It happens on all 4 of my maps
Specs:
3060 ti
5600x ryzen
32 GB Ram
Expected behavior
Well i expect it to function as it should without crashing my game
A private message is associated with this bug report
Thank you for your report. We are examining the place file you sent and will get back to you soon.
Is this issue something that started happening recently, and if so, can you tell us approximately when it began?
honestly im not sure, i just did it to use it on my map cuz atomic is better for loading a race map, and it just happend, do note i cant put nonatomic on 3 on my maps at the same time, they all must be Default aswell, if i set 2+ nonatomic it crashes
I mentioned in another thread on atomic models which you also posted on that we have some improvements in the works which will improve performance when dealing with large atomic models. That is also true here, but I have some additional advice for your case, as I am concerned you may still have issues even after those improvements arrive.
Your repro file contains a single atomic model which includes pretty much all of the instances in the entire world. As you’ve seen, this pattern has significant performance drawbacks, but it also minimizes benefits from enabling streaming, as you’ve set basically the whole world to replicate as a single atomic unit. I would strongly recommend not doing this. You will find that having many, smaller atomic models performs better than having one much larger one.
Trying this with the provided repro place, if I made the root model non-atomic and made all of the child models atomic (including grouping part children of the root model into another atomic model), the performance was greatly improved.
While we will be making efforts to improve performance of atomic models with many part descendants, I believe that carefully considering smaller models and marking them atomic, persistent, or default as makes the most sense sense will yield not only better performance, but will also benefit clients who will no longer need to have the entire world loaded at all times in order to participate in your experience.
how many parts should i have per model for the most optimized experience?
and does this work for nonatomic aswell?