Somebody posted in one of the reviews I read:
"
1. DX11 can only use a single thread to send commands to the GPU
2. DX12 can use multiple threads (slides showed up to 8, but I am guessing it could use more)
3. DX11 is very high overhead
4. DX12 is low overhead
5. DX12 drawcalls/s possible is way higher - current benches show around a 15x higher number of drawcalls vs DX11.
6. Split Frame Rendering in SLI and CF configurations. Supposedly this will speed things up a lot.
"
If 1 and 2 are true, that alone will account for huge performance boost for all games that support DX12. Think about it - going from single thread to multiple threads that can simultaneously work together to process resource-intensive graphics.
Currently, GPU can communicate with only 1 core at a time, which then either has to tell the other cores what to process OR you just wait for the GPU to change calls a.k.a lag time between phone-calls. Now imagine a GPU that can directly communicate with up to 8 cores at the same time. Opens up a bottleneck..
BTW, it will not do anything for UT99 or CSGO or any game that does not have DX12 support so don't expect going in that it will increase your FPS in UT99 or CSGO all of a sudden