After reading this article I’m curious to know how the GTX (or even RTX) series holds up against nVidia’s workbench cards (namely the popular P2000) for hardware accelerated transcoding.
I’ve had a server build sitting on a wishlist waiting until I have some money to buy it, and if I can swap out the P2000 for say a GTX 1060 or 1660 then I’d be stoked to be that much closer to affording it.
I searched for comparisons, but could only find stock standard (locked) 1080’s being compared. Sorry if I didn’t look hard enough.
the locking only affects the ENCoder, which is the relatively easy part (cpu’s have been doing x264 encoding for years now).
the (4k x265) DEcoding is the hard part, and that part hasn’t been locked.
so locked or unlocked, the 4k/x265 decoding is the important part and i’d expect the (decoding) performance should similar across same level hardware.
What so far seems to be a more limiting factor to consider (as far as linux nvidia decoding), is that VRAM usage is much higher than windows.
linux nvidia 4k/x265 decoding appears to use ~1gig of video ram per decode. So this means cards with less ram, may not have sufficient ram available to do more than 1 or 2 decodes at a time, regardless of hardware performance.
Whether this is due to NVDEC support being incomplete, or the decoding path differences between linux and windows, remains unknown at this point.
