Unified memory just means the memory is on chip and shared between the CPU and GPU. It's not going to have a big impact on performance, if any.
We have no idea about which kind of RAM and frequency Apple is using for its unified memory, so it's a bit early to tell.
The M1 likely has better single core performance, but the Ryzen 9 4900HS has 8 high performance cores, not 4, and all of them have SMT for a total of 16 threads. In heavy multicore workloads, the 4900HS will be faster.
Apple Silicon uses all available threads for multicore-intensive applications, not only the 4 high-performance ones. This cores may be more efficiency oriented, but their combined performance is not negligible. Combining the much greater single core performance of the performance cores (vs. any AMD or Intel core, really) with the added performance of the efficiency cores, the multicore results of the M1 may already surpass the 4900HS.
Also, for some applications, the hyperthreading/SMT doesn't add any extra performance at all. For example (since you mention it) in particle simulations.
I have no doubt that Macbook Pro and Air with the M1 chips will be super efficient, with great battery life and excellent performance for light productivity like web browsing, office, and photoshop. I'm buying one for those reasons because I have a high performance desktop for actual work. They are not comparable to a notebook like the G14 though, especially in heavy productivity like 3D modeling, tile based rendering, particle simulation, or video rendering that can take advantage of CUDA acceleration.
We'll see, but I think it's a bit early to dismiss the M1 as just a "light productivity" capable machine.