Hardware Accelerated H.264/265 [De|En]coding - New Mac Pro

Heya gang -

I'm a bit ignorant of driving FCPX (ok, ok ... a LOT ignorant) and I'm hoping I'm just doing something wrong or misunderstanding something. Is FCPX supposed to be able to make use of the hardware h.264/h.265 encoders on the AMD Vega II GPU? I thought it was supposed to, but I might be completely mistaken. I know it will use Intel's QuickSync on the laptops and other Macs that don't have Xeons. But what about the AMD hardware? Can it use that as well? Or is it just using the hardware for other rendering tasks?

My inputs and outputs are both 4K h.264 (or h.265 for output). I know that h.264 isn't the best of CODECs to edit in, but it's what my software produces. So I have to make the best of it. I did some simple re-encoding with my new Pro today, using a 5 minute 4K/59.94 h.264. Outputting to both h.264 or h.265 was basically real time; IOW around 5 minutes. That's pretty good, but I was sorta hoping for it to be a bit quicker; hardware encoding should have halved that time, give or take.

Am I barking up the wrong tree here?

For the record, hardware:

• Mac Pro 7,1 w/16-core Xeon
• 96GB RAM
• 1TB system drive (not using that for editing)
• AMD Vega II GPU
• 8TB of an NVMe RAID01 volume spread across 2 HighPoint PCI-E cards (this is the drive I use)

The source footage was a 5 minute 75Mbit/sec h.264 file, 4K/59.94
Output 1: 30Mbit/sec h.264 file, 4K/59.94
Output 2: 25Mbit/sec h.265 file, 4K/59.94

Thoughts/comments?

Thanks.

I may be way off the mark here, but in my experience FCP is written to take advantage of iMacs more than anything, since those are the computers that actually sell. I have a Mac Pro (trash can) and it offers zero hardware acceleration at anything simply because they chose to put Xeons in it for some reason. I'm not sure what Xeons really bring to the table when it comes to editing, but I often found myself wishing for one of those i-processors instead because at least in that case I could have some acceleration for encoding and decoding. I don't know if the new Mac Pro has any hardware acceleration. The Afterburner card seems to make very little difference at all in the tests I've seen. Certainly not $2K worth of difference. But that's only for ProRes, which already plays back just fine. Supposedly the AMD CPUs are much better and cheaper than what Intel has been offering lately. Hopefully Apple switches up.

Can you run a benchmark or statistics program on your Mac Pro to see what is going on in your computer system? Having said that I thought the T2 chip was supposed to help encode and decode video.

jasonvp wrote: ... new Pro today, using a 5 minute 4K/59.94 h.264. Outputting to both h.264 or h.265 was basically real time; IOW around 5 minutes. That's pretty good, but I was sorta hoping for it to be a bit quicker; hardware encoding should have halved that time, give or take..

I think the T2 chip handles H.264/H.265 (HEVC) encode/decode acceleration. Normally it should be faster than real time for regular 8-bit H.264. However all codecs are different, even within a given type like H.264. Previous Macs and FCPX versions were very slow on 10-bit HEVC output. I don't know about your new 7,1 Mac Pro (Cheesegrater 2.0).

IF the issue is encoding performance, separate this from rendering. Duplicate the timeline as a snapshot, open that, strip all effects with Edit>Remove Effects, then export to 4K H.264 using Share>Master File>Settings>Format:Computer, Video Codec: H.264 Faster Encode, Resolution: 4k. Then repeat the same test, exporting from FCPX to 8-bit HEVC.

Joema,

I don't know about the T2 chip but I know Intel's Quick Sync cannot help playback and render all versions of H.264 as seen in the video below. I am sure Intel's Quick Sync will get a revamp soon.

joema wrote: IF the issue is encoding performance, separate this from rendering. Duplicate the timeline as a snapshot, open that, strip all effects with Edit>Remove Effects, then export to 4K H.264 using Share>Master File>Settings>Format:Computer, Video Codec: H.264 Faster Encode, Resolution: 4k. Then repeat the same test, exporting from FCPX to 8-bit HEVC.

My tests had no rendering to be done; no effects or anything whatsoever. It was literally taking a 4K/59.94 h.264 file and writing a new, identical copy of that same h.264 file. Then also transcoding it to h.265. With hardware encoding, that should have been faster than it was.

I did follow your suggestion after doubling the length of the footage (literally using 2 copies of the same clip) to ten minutes. Sure enough: it took about 10 minutes to write the h.264 file and quite a bit longer to do the h.265 file (somewhere around 13-15 minutes). FPCX doesn't allow me to create "Master File" with h.265, so I had to create a Compressor output instead.

Based on what I'm seeing, it doesn't matter which encoder and setting I choose (Faster/Slower, h.264/h.265). It appears as though there's no access to the AMD hardware encoder with FCPX; at least not with the Vega II GPUs.

arc wrote: I don't know about the T2 chip but I know Intel's Quick Sync

Unfortunately, Intel's QuickSync isn't applicable at all here. The Mac Pro has a Xeon chip in it, and those don't have iGPUs incorporated in them. Therefore: no QuickSync.

jasonvp wrote:

arc wrote: I don't know about the T2 chip but I know Intel's Quick Sync

Unfortunately, Intel's QuickSync isn't applicable at all here. The Mac Pro has a Xeon chip in it, and those don't have iGPUs incorporated in them. Therefore: no QuickSync.

I am well aware that the Xeon CPUs do not have Intel's Quick Sync. That is why I stated I don't know about the T2 chip. I knew the Mac Pro would have to use the T2 chip. I was letting the OP know that The T2 Chip might not playback all H.264 video files like Intel's Quick Sync. There should be a way to get the stats of the Mac Pro. Having said that the video below might be of some interest.

arc wrote: I am well are that the Xeon CPUs do not have Intel's Quick Sync. That is why I stated I don't know about the T2 chip.

OK, understood. I don't believe the T2 chip is at play here at all. As far as I can tell, the encode/decode performance is due to the GPU hardware. And unfortunately the Vega II GPU has an older encoder and decoder that isn't nearly as capable as what's available on Team Green. They work, but slowly unlike NVidia's which go through h.264 and h.265 like butter.

The encoder that'll come with the new W5700X MPX module (whenever...) should be a lot faster.

Having said that the video below might be of some interest.

I saw that when he launched it; his presentation style grates on me though. Anyway, I don't have the stuttering during the playback that he has.