Warspite wrote:If you use DXV content, all content is decoded on your videocard, there is no parallel work for your CPU, and no way to make good use of CPU multi threading.
That's not really true. When compared to h264, decompression indeed does take a whole lot less time from the cpu, but it's still there.
GMAS wrote:So I'm using a HP Z620 with 2 Xeons (so I have 16 cores). Yesterday we did a test with 40+ layers and many slices. Totaal resolution counts up to about 5k...
We saw that indeed only 1 core is doing the main load, some others do minor work, the rest is just goofing around.
With a gtx 1080 ti we get about 35fps. So I guess I'm missing clocks speed to get higher fps. The Xeons can go to 3.3 GHz in turbo...
I think you're guessing in the right direction. What i think you're seeing is that the 1 core (our renderer) is working as fast as it can, and there just isn't enough work to offload onto other cores. (Especially if there's a LOT of slices) 3.3GHz is pretty decent though and i dont think you'll find a machine anywhere that can easily do 40+ layers. In fact, 35fps for 40 layers sounds pretty good to me
Spoiler: We've got some work planned for some cases where we've seen renderer utilization drop below 70%. This might improve your situation as well.