Yes. This doesn't strictly mean that something isn't CPU or GPU, just that not all available cycles are being used...something is forcing execution to wait. The video analysis of the alpha that was posted earlier suggests some sort of depth buffer overhead problem, which, from my limited understanding of the potential issues could be responsible for what I've observed.
One thing it's not is an I/O problem; that's negligible.
I tried all sorts of affinity combinations (not allowing the second logical core on a physical core to be used, using one CCX, using one core per CCX, etc and so forth):
Source: https://www.youtube.com/watch?v=TI-QmVKeGiE
It's simply not using all available cycles unless I squeeze all the threads into a very small number of logical processors.
I think the game is waiting around while it spends most of each frame testing and updating the depth buffer. Whatever hardware this is dependent on is evidently becoming bottlenecked before all cycles are used.
Edit:
Source: https://www.youtube.com/watch?v=sZczI65QrbI
Another one, simulating 1440p (with 1080p 1.5x SS) ultra. Approaching the settlement never maxed out any core or the GPU. I thought system memory latency could have been an issue, but the game doesn't seem very demanding in this regard...never going above about 7-8% of the bandwidth HWiNFO can reach in bandwidth limited tests. So, I'm inclined to suspect it's the GPU's caches/memory that's being hammered. GPU memory controller usage only gets to ~50%, but thats still very high for a game and I don't think it includes cache usage.
Videos are still processing, so they are hard to see right now, but they'll eventually be 4k60 when YouTube gets around to it.