I hope Epic does the same test and fixes it. This was a brilliant test and sad to hear about the meta humans
5 Likes
Did anybody of you see any different performance in UE5.3?
I recreated my earlier test in UE5.3 and noticed a dramatic difference. Whereas before, adding even a few duplicates of Ada (even minus hair) would cause significant drops in frame rate, my system seemed completely unaffected by the first dozen or so clones. By the time I had reached 24 copies, the frame rate was beginning to show strain, but it was still running at a healthy 80 fps.
With 48 copies, it was running at 50 fps.
So, way better results overall. I can’t say for sure, however, whether this is due to better performance of UE5.3 over UE5.0, or if it’s due to updates on Metahuman.
One thing I did not test the first time around, however, was just to what extent is the poor performance a product of being run from within the editor. It’s always a possibility that, even though Metahuman was running poorly in the editor, it may still run perfectly fine in the packaged build. The slow fps could be due to editor functionality running behind the scenes to aid in production, which would not be present in the packaged build. So, I packaged my UE5.3 project with 48 Adas and in the packaged build, the fps was hovering between 112 and 120.
(the ticker is a gizmo I use to monitor fps during gameplay)
Lastly, just to verify what kind of performance I’d get with the old Metahumans on an older version of UE5, I opened the U5.0 project and ran multiple clones again. This time, I threw in the towel after 18 copies, because the frame had dropped to a dismal 15 fps.
I tried to package the U5.0 project to see if it carried over to the final version, but unfortunately the packaged version came out with some kind of weird compression or filter over it, and I did not know how to remove it. The displayed fps was in the 155-to-160 range, however, I’m guessing a lot of that is due to image being heavily compressed.
So, unfortunately I wasn’t able to confirm how well the older version works when packaged with nearly two dozen clones. But my guess is, if the UE5.3 test is anything to go off, that they will play much smoother in the final packaged build than they did in the editor.
4 Likes
bro, you are a legend, thank you
Doing performance testing in editor is not ideal and basing the results on just FPS is not ideal as well.
okay, make your test then =/
1 Like
Thank you so much for testing! 
Really helpful.
What a sad comment
2 Likes
do the new 5.3 skeletal editor tools allow you to achieve the blender delete bones pipeline entirely in engine?
I all honestly, you should have checked quad overdraw and shader complexity.
Unless your CPU bound?
Just took a peek at the new Metahumans:
^ from UEFN 5.4
Check my reply in the other thread. In the new 3.0 MetaHumans they added a Metahuman actor component which allows you to change “corrective” animation playback.
This component is currently being held hostage in UEFN (UE 5.4). It’s not in 5.4 preview. Bridge will not let you import 3.0 Metahumans into UE 5.4 preview.
1 Like