I did hardware acceleration for 1080p on a 4th gen Intel for a bit (Windows) years ago. I have helped others get quicksync setup on an intel 7th gen (Windows) headless servers. I have worked other forms of hardware acceleration VAAPI for a bit (Linux and an older AMD GPU)
Now, I have been running a virtualized environment (VMware esx 6.7 u3 ryzen 1700 (8 core 16 thread) 32 GB ECC memory, 1.5TB nvme memory (for vm system VHDs), 48TB disk storage (33 usable after ZFS chops it up) that is passed to a freenas VM, and Emby server is hosted on a headless Linux VM (runs on 8 threads (2 socket 4 VCPU per socket) and 8GB of RAM. At max I will have 6 clients. My nominal is 3 streams at once plus 2 live tv recording at once. CurrentlyI software transcode. Just about all the different hypervisors out there are optimized to efficiently use resources and share across guest VMs.
I am working on my 4K library and when I transition I will be testing for throughout. Right now I am limited by ESXi free licensing to 8 threads per VM. I may get VMUG licensing to expand thread count per VM or look at hardware pass-through, a mid level PCIE GPU ,and driver hacking the GPU to the emby VM.
Mid year this year, I am considering rebuilding my home server when the Ryzen 4000 series 8 core 16 thread #700X model drops (improvements would be pcie 4, 8+% IPC, +200Mhz clock speed, double the L2/3 Cache in comparison to the 3000 series). For me this would be moving forward 3 generations of Ryzen which is pretty substantial. I am also considering moving the xenserver or HyperV as my hypervisor instead of ESXi).
Sent from my iPhone using Tapatalk
Edited by Tur0k, 01 March 2020 - 12:48 AM.