LAMMPS performance on AMD Zen/Ryzen CPU

Dear LAMMPS users,

I am currently planning to build a new Desktop machine to be used for LAMMPS simulations. I'm planning to use an 8 core CPU (consumer level) combined with 1-2 NVIDIA GPUs (also consumer level). The NVIDIA GPUs are pretty much a requirement for the GPU package (CUDA seems to be much better than OpenCL). When it comes to the CPU however, the choice is not so obvious. AMD has released the Ryzen CPUs, and an 8-core Ryzen CPU costs less than half of Intel's 8-core CPUs. On paper AMD's Ryzen CPUs should perform similar to and sometimes even better than Intel's, especially in multi-core/threaded workloads.

So I wonder if any of LAMMPS users are using/have tested AMD's Ryzen CPUs, and can say anything about the performance with LAMMPS? (preferably on Linux)

I've seen some benchmark results on Tomshardware (AMD Ryzen 1800X review), where LAMMPS benchmarks are a part of the SPECwpc benchmark suite (Scientific and Engineering Computations), and Ryzen CPU generally performs better than Intel's. However these are performed on Windows and I'm not sure how optimized "their" LAMMPS executable is with respect to either of the architectures (Intel's and AMD's). Also, the SPECwps benchmarks are only for the CPU, and does not use the GPU acceleration available in LAMMPS...

Based on what I've seen on the Internet so far, the raw computational power of AMD's Ryzen CPUs seems to be comparable to or even greater than Intel's. But since I will be using the GPU package, the combination of CPU and GPU becomes important, and it pretty much boils down to how well AMD's Ryzen CPU play along with NVIDIA GPUs.

BTW: I am aware that there are many factors that will affect the simulation performance, specially for a hybrid CPU/GPU setup. I am mostly interested in other people's experience (if any) with LAMMPS on Ryzen CPUs, and whether it really is comparable to Intel's 8-core CPUs.