Which GPU GT 1030 or GTX 1050?

[Guest asrequested]

Here's how they compare

 

https://www.nvidia.com/en-us/geforce/products/10series/geforce-gtx-1050/

 

 

I'd say the 1050 is probably good for your use

[Guest asrequested]

I love how they say is supports 8k but the HDMI is only 2.0b. 

 

But it does have DP 1.4. So it actually does support 8k

Edited May 15, 2018 by Doofus

[Grom8 2]

Hahaha, thanks again.

 

I'm probably confusing to talk to messing up the terms of stuff, this is all very new to me. Thanks for helping fix this issue and give me something to read about

 

Grom8

[Waldonnis 148]

I love how they say is supports 8k but the HDMI is only 2.0b. 

 

But it does have DP 1.4. So it actually does support 8k

 

They do support 8k...as long as you're not expecting to output it to a monitor via HDMI     I can't even imagine using the encoding side for 8k video, though, due to the quality and compression differences compared to software encoding or pro hardware.  Outside of the rare demo-type videos on YouTube, the only 8k stuff you'd likely find is raw camera footage, and who would want to encode that using nvenc??  Seems silly unless all you're doing is posting it on YouTube (which will re-encode it anyway) or for home-video type applications.  From a gaming streaming/recording situation, though, I guess it makes more sense - recording games that are spanning multiple monitors or that have ultra-wide aspect ratios and such.

 

As for 1050 vs Ti...probably won't make a difference either way.  The "media block" is actual hardware on the card rather than the individual "cores" doing the work, so the gaming capabilities don't matter.  Bus speed and bandwidth do, though...although in this case you'd probably only be looking at single-digit percentage performance difference (if any at all) between the Ti and non (most of the Ti's improvements were more aimed at budget gaming than media playback/encoding).  IMO, you're limited to two streams anyway (and the hardware itself isn't why it's limited to two), so don't bother spending more than you have to if you're only using it for transcoding.

[Guest asrequested]

Yeah, I think they are doing it just to make it seem more desirable, even though it can't actually be used. 4k has barely got a foot hold. And yet they've made HDMI 2.1, 10k  ready. I'm using USB-C port over Thunderbolt 3, which gives me 8k bandwidth. Not that my GPU supports 8K, though. The main reason I use it is so that I don't need an active adapter for the DP to get 60 FPS. So much better than HDMI.

[Waldonnis 148]

Yeah, I think they are doing it just to make it seem more desirable, even though it can't actually be used. 4k has barely got a foot hold. And yet they've made HDMI 2.1, 10k  ready. I'm using USB-C port over Thunderbolt 3, which gives me 8k bandwidth. Not that my GPU supports 8K, though. The main reason I use it is so that I don't need an active adapter for the DP to get 60 FPS. So much better than HDMI.

 

I think it may just be a side-effect from the other changes in the card or block architecture.  It may not be taped out to have a resolution limit, but rather the resolution limit could be determined by the bandwidth and processing capabilities..then they just fix the max "reliable" value in the driver or firmware so it can't run into a situation where the user may exceed the physical capabilities.  Even older cards can handle displaying higher resolutions than most would expect (I run a 2160p monitor and 1080p monitor off of my 970, for example, and have run more concurrently at times), but if I tried high-poly-count gaming on both at the same time, the performance would suffer greatly because the card can't push that much data through.  The encoding block probably isn't much different, albeit with more firmware limitations in place since it's a fixed-function block.

[Guest asrequested]

I guess it's like an engine. Design it, build it, and then see what it can actually do. Then apply a limiter, if required

[PrincessClevage 175]

I have a gt1030 (fanless/silent) and it plays 4K content fine local and remotely with all ET quality features turned on.

[yepitro1986 0]

That's why I asked if a particular software was required. I figured ffmpeg wasn't fully utilizing the ability. And why I said application not development. As you say, it's been around for a while, but would seem that not much actually takes full advantage of NVEncC. But it's good to know that the potential is there. I don't have any Nvidia GPUs. I've been using iGPUs, for years. Less parts, less fans.

and your IGPU success transcode HEVC 10 bit without any problem?

[Guest asrequested]

and your IGPU success transcode HEVC 10 bit without any problem?

I switched my server over to a Threadripper, last year.

 

https://emby.media/community/index.php?/topic/51347-server-gpu-acceleration-or-cpu-processing/page-5&do=findComment&comment=507643

 

My other machines have iGPUs. I wasn't having good success with GPU transcoding.

[yepitro1986 0]

I switched my server over to a Threadripper, last year.

 

https://emby.media/community/index.php?/topic/51347-server-gpu-acceleration-or-cpu-processing/page-5&do=findComment&comment=507643

 

My other machines have iGPUs. I wasn't having good success with GPU transcoding.

 

okay maybe  GTX1050 alternative solution

[Guest asrequested]

okay maybe GTX1050 alternative solution

The 1050 should work well enough, for two GPU transcodes.

[Grom8 2]

Hi again,

 

Is there a guide for NVIDIA Nvenc setup? I'm getting better performance on a Pentium with quick sync, than a 1050.

 

I have the NVIDIA driver but after that is it just plug-in-play? My CPU will get to 100% load with quick sync (0% GPU usage) and manage to play 20 seconds of 4k before freezing to load the next part, when I use Nvenc the CPU usage drops to 10% and the GPU gets around 15% but the video transcodes slower as well as bugging out transcoding 6 seconds then repeating the cycle (video playback is frozen).

 

I'm finding it really hard to find information on a initial setup, I doubt I have done it correctly.

 

Grom8

Edited May 17, 2018 by Grom8

[yepitro1986 0]

Hi again,

 

Is there a guide for NVIDIA Nvenc setup? I'm getting better performance on a Pentium with quick sync, than a 1050.

 

I have the NVIDIA driver but after that is it just plug-in-play? My CPU will get to 100% load with quick sync (0% GPU usage) and manage to play 20 seconds of 4k before freezing to load the next part, when I use Nvenc the CPU usage drops to 10% and the GPU gets around 15% but the video transcodes slower as well as bugging out transcoding 6 seconds then repeating the cycle (video playback is frozen).

 

I'm finding it really hard to find information on a initial setup, I doubt I have done it correctly.

 

Grom8

 

my friend had GTX 1070 transcode HEVC 10 bit HDR very good only consume 20% CPU 

[Guest asrequested]

Try disabling decoding.

[Grom8 2]

I did that previously and I tried again and always seem to get this at the end of the ffmpeg logs...

 

frame= 875 fps=6.9 q=12.0 size=N/A time=00:00:36.73 bitrate=N/A speed=0.291x

frame= 880 fps=6.9 q=12.0 size=N/A time=00:00:37.14 bitrate=N/A speed=0.293x

[segment @ 000001e882dc0780] Opening 'C:\Users\Worker1\AppData\Roaming\Emby-Server\programdata\transcoding-temp\9228be56e325412644760119de7d7eb3.m3u8.tmp' for writing

frame= 885 fps=6.9 q=12.0 Lsize=N/A time=00:00:37.18 bitrate=N/A speed=0.292x

video:329013kB audio:1310kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown

[aac @ 000001e8827d9260] Qavg: 63493.941

 

The video I was trying to play played smoothly till 9 seconds then froze, and in the emby dashboard I was constantly looping from 10s to 20s then back again.

Edited May 18, 2018 by Grom8

[Grom8 2]

Here is some ffmpeg logs...

ffmpeg logs.txt

Edited May 18, 2018 by Grom8

[Guest asrequested]

This is the section you want to look at.

Stream mapping:
  Stream #0:0 (hevc) -> overlay:main (graph 0)
  Stream #0:4 (pgssub) -> scale (graph 0)
  overlay (graph 0) -> Stream #0:0 (h264_nvenc)
  Stream #0:1 -> #0:1 (truehd (native) -> aac (native))
Press [q] to stop, [?] for help
frame=    0 fps=0.0 q=0.0 size=N/A time=-577014:32:22.77 bitrate=N/A speed=N/A    
[matroska,webm @ 000002485b003580] sub2video: using 3840x2160 canvas
[segment @ 000002485b642720] Opening 'F:\Transcoding Temp\transcoding-temp\defb384a07b58372edc59cca6641b3f80.ts' for writing
Output #0, segment, to 'F:\Transcoding Temp\transcoding-temp\defb384a07b58372edc59cca6641b3f8%d.ts':
  Metadata:
    encoder         : Lavf57.83.100
    Stream #0:0: Video: h264 (h264_nvenc) (High), yuv420p, 3840x2160 [SAR 1:1 DAR 16:9], q=-1--1, 99809 kb/s, 23.98 fps, 90k tbn, 23.98 tbc (default)
    Metadata:
      encoder         : Lavc57.107.100 h264_nvenc
    Side data:
      cpb: bitrate max/min/avg: 99809608/0/99809608 buffer size: 199619216 vbv_delay: -1
    Stream #0:1: Audio: aac (LC), 48000 Hz, stereo, fltp (24 bit), 384 kb/s (default)
    Metadata:
      encoder         : Lavc57.107.100 aac

You're trying a 4k HDR. That's a tall order. What happens if you try a 1080 HEVC? And/or don't use subs?

[Grom8 2]

I run into a similar issue except freezing every 10 seconds but moving forward this time. There must be some sort of setup/setting im not using correctly.

 

Intel Quicksync at 1080p makes my CPU run at 80-85% transcoding 1080p but video playback is triple the FPS (33fps). The movie can be watched comfortably at this setting.

Nvenc creates lag while stopping and starting (11 fps) with the GPU at 5-15% load, surely there is something wrong here because GPU utilization is so low and for some reason isnt increasing to compensate for the poor performance.

 

If a Pentium can get successful playback of 4k HDR with lowish frames and having to stall every 20 seconds but the movie being "watchable" from start to finish (100% load),  then how come a GPU which is far better at this task unable to play video after 10 seconds (still 10% GPU load) because it completely freezes?

Edited May 18, 2018 by Grom8

[PrincessClevage 175]

I run into a similar issue except freezing every 10 seconds but moving forward this time. There must be some sort of setup/setting im not using correctly.

 

Intel Quicksync at 1080p makes my CPU run at 80-85% transcoding 1080p but video playback is triple the FPS (33fps). The movie can be watched comfortably at this setting.

Nvenc creates lag while stopping and starting (11 fps) with the GPU at 5-15% load, surely there is something wrong here because GPU utilization is so low and for some reason isnt increasing to compensate for the poor performance.

 

If a Pentium can get successful playback of 4k HDR with lowish frames and having to stall every 20 seconds but the movie being "watchable" from start to finish (100% load), then how come a GPU which is far better at this task unable to play video after 10 seconds (still 10% GPU load) because it completely freezes?

On emby servers transcode section what settings do you have configured? Do you have ssd for transcode temporary path?

Edited May 18, 2018 by PrincessClevage

[Grom8 2]

Thank you for the settings, I tried them and I still get the freezing at 10 seconds while looping in dashboard. Its weird though I still have <8% GPU usage while transcoding video

[Guest asrequested]

That's why you have the issue. The CPU can't handle the load. You're GPU, for some reason, isn't being properly utilized.

[Waldonnis 148]

I'm wondering why 2160p HDR-flagged HEVC is being transcoded to 2160p yuv420p h.264 at all.  It's probably going to look terrible since it's not being tonemapped to yuv420p (8bit) and very few devices even support 2160p h.264 anyway (at least not many do it well).  With UHD content becoming more common, the use of HEVC encoding really needs to be looked at soon.  Yes, it's a big issue to tackle, but it's not going to go away.

 

That being said...

 

Here is some ffmpeg logs...

 

There's a lot going on in the log.  It's not being told to use hardware to decode the video stream, so the CPU is decoding the HEVC video, along with decoding/rendering/overlay of the subs (also CPU), audio transcode (CPU de/encode)...you get the picture.  Basically, you're still going to see a lot of CPU use in situations like this no matter what.  It still shouldn't be as slow as the indicated encoding speed, but there could be multiple reasons for this and the ffmpeg loglevel isn't verbose enough to tell what may be going on.  You can try running it manually and increasing the loglevel so you can tell if it's actually initialising the hardware properly.  There was an nVidia driver problem late last month that caused some problems for many people, so it may be worth checking if you got bitten by that (I know I did; broke a few of my scripts when the driver decided to unload itself randomly).

[Guest asrequested]

I was wondering about that. A few things in that log, I didn't recognize.

[Jdiesel 1431]

This may be an unpopular opinion but 4K HDR transcoding is not ready from primetime yet. This is no fault of Emby but rather a combination of factors. Do yourself a huge favor and convert your 4K HDR remuxes to a more proven format that will direct play on the devices that currently require transcoding and keep both copies. Emby already does a good job of handling multiple versions so the 4K HDR videos will direct play on the devices that support it and devices that don't can use the pre-converted version. I have made a second 5Mbps 1080p, h264, Rec.709, 2-channel aac and multichannel DD audio tracks version from all my 4K HDR movies. The average file size of the second version is about 6GB and they direct play on pretty much every device I have.

Edited May 18, 2018 by Jdiesel