Yeah its completely swept under the floor which is a failure to market and communicate true optimization paths in my opinion… Even today in UE5.5 look at this.
DX11 SM5 - 4.26ms starting cost (235fps)
DX12 SM5 - zero changes other than rhi results in (but why even run dx12 with sm5?)
Note: the memory footprint difference
29% drop for literally nothing
DX12 SM6 - zero change other than lumen is now enabled since its locked behind sm6
Another 73% drop from DX12 SM5.
That’s a total loss of about 123% of the initial starting ms
I feel this is a failure to capitalize on an area where so many ppl who have asked and been yelling that UE5 is too heavy could have simply been told (fall back to dx11 as older hardware drivers and chips are better suited for dx11 and sm5)
It all honesty it was our “secret weapon” for running really cheap pixel streaming instances with Lumen SWRT GI without the need for HWRT capable machines… this change really hurts our business model. Direct impact on operating cost. With the starting overhead budget we could squeeze 4 users per instance in with no issues… We will be lucky to hit 60 fps with lumen now with 2 users per instance unless we go g5 and thats pretty sad.