HTPC Credentials - Local Media Playback and Video Processing

Evaluation of local media playback and video processing is done by playing back files encompassing a range of relevant codecs, containers, resolutions, and frame rates. A note of the efficiency is also made by tracking GPU usage and power consumption of the system at the wall. Users have their own preference for the playback software / decoder / renderer, and our aim is to have numbers representative of commonly encountered scenarios. Towards this, we played back the test streams using the following combinations:

  • MPC-HC x64 1.8.5 + LAV Video Decoder (DXVA2 Native) + Enhanced Video Renderer - Custom Presenter (EVR-CP)
  • MPC-HC x64 1.8.5 + LAV Video Decoder (D3D11) + madVR 0.92.17 (DXVA-Focused)
  • MPC-HC x64 1.8.5 + LAV Video Decoder (D3D11) + madVR 0.92.17 (Lanczos-Focused)
  • VLC 3.0.6
  • Kodi 18.1

The thirteen test streams (each of 90s duration) were played back from the local disk with an interval of 30 seconds in-between. Various metrics including GPU usage and at-wall power consumption were recorded during the course of this playback. Prior to looking at the metrics, a quick summary of the decoding capabilities of the Radeon RX Vega 11 is useful to have for context.

While the NVIDIA and Intel GPUs support decoding of 8K streams also, AMD has decided to limit support to 4K. 8K is mainly of academic interest right now, and the range of supported codecs is more important. Fortunately, the driver reports supports for MPEG2 (up to 1080p), H.264, HEVC (both 8b and 10b), VC1 (up to 1080p), and VP9 (both 8b and 10b).

All our playback tests were done with the desktop HDR setting turned on. It is possible for certain system configurations to have madVR automatically turn on/off the HDR capabilities prior to the playback of a HDR video, but, we didn't take advantage of that in our testing.

VLC and Kodi

VLC is the playback software of choice for the average PC user who doesn't need a ten-foot UI. Its install-and-play simplicity has made it extremely popular. Over the years, the software has gained the ability to take advantage of various hardware acceleration options. Kodi, on the other hand, has a ten-foot UI making it the perfect open-source software for dedicated HTPCs. Support for add-ons make it very extensible and capable of customization. We played back our test files using the default VLC and Kodi configurations, and recorded the following metrics.

Video Playback Efficiency - VLC and Kodi

The key aspect to note here is that VLC is unable to take advantage of the hardware acceleration for VP9 Profile 2 videos. Instead, it relies on software decoding for that stream. Otherwise, we see the hardware decoder being used (with usage ranging from 18% for the 480i60 MPEG2 streams to as high as 75% for the 4Kp60 HEVC Main 10 video). D3D loading of up to 65% is triggered due to scaling requirements for non-4K videos. The at-wall power consumption is between 25W to 35W for the hardware accelerated codecs. Software decoding of 4Kp60 VP9 Profile 2 videos drops lots of frames, and the recorded metrics for that stream do not present any useful information. Kodi, on the other hand is able to play back all the videos in our test suite with hardware acceleration. We see the decoder usage go as high as 90% for the 4Kp60 10-bit encodes. The at-wall power consumption is between 30W and 35W for all the streams.

MPC-HC

MPC-HC offers an easy way to test out different combinations of decoders and renderers. The first configuration we evaluated is the default post-install scenario, with only the in-built LAV Video Decoder forced to DXVA2 Native mode. Two additional passes were done with different madVR configurations. In the first one (DXVA-focused), we configured madVR to make use of the DXVA-accelerated video processing capabilities as much as possible. In the second (Lanczos-focused), the image scaling algorithms were set to 'Lanczos 3-tap, with anti-ringing checked'. Chroma upscaling was configured to be 'BiCubic 75 with anti-ringing checked' in both cases. The metrics collected during the playback of the test files using the above three configurations are presented below.

Video Playback Efficiency - MPC-HC with EVR-CP and madVR

Similar to VLC, MPC-HC / LAV Video Decoder is unable to play back the 4Kp60 VP9 Profile 2 video with hardware acceleration. In fact, with the default configuration (EVR-CP), the video just shows a blank screen (and there is hardly a significant change in the at-wall consumption for the stream). With madVR as the renderer, and the LAV Video Decoder set to D3D11, we do manage to get the software-decoded (avcodec, as reported by LAV Filters) stream on the display. The madVR DXVA configuration works decently without dropping frames for 1080p and lower resolution videos. The 4K videos see significant frame drops, as evident from the D3D Load in the above graphs. For the Lanczos configuration, 1080p and higher resolution videos with a frame rate of more than 50 fps suffer from playback issues. These results are along expected lines for an integrated GPU. However, they are significantly better than the performance of madVR with Intel's integrated GPU.

Overall, we can recommend Kodi 18.1 as the playback software for local media on the DeskMini A300. Other open-source softwares such as VLC and MPC-HC / LAV Video Decoder do not play well with the current drivers for certain types of video streams.

HTPC Credentials - YouTube and Netflix Streaming Power Consumption and Thermal Performance
Comments Locked

88 Comments

View All Comments

  • abufrejoval - Friday, April 26, 2019 - link

    Very nice review!

    But surprisingly little relative change (relative to publicity...) from the previous (major) iteration, which I interpret as the Kaveri vs. Skylake Iris Plus that I own and tested, A10-7850k vs. i5-6267U.

    Intel still seems to never use more than 15Watts for the CPU, yet manages scaling single to 4GHz at great IPC while it manages to sustain admirable Hertz even at multi-core constant loads, taking a nice sip of cool on every little stall. AMD seems to retain a much more linear efficiency curve where clocks and cores just eat power, while the difference at the wall plug is much smaller in this iteration (was 3:1 for exactly the same performance on my old systems).

    The good thing is that on a device like this, peak power is much less important than on a notebook, so it’s ok, as long as maintains quiet on constant peak and (finally) reaches acceptable idle: Here I see a lot of progress on AMD's side, Intel has much less room to beat itself.

    For graphics, bandwidth is so crucial and I wonder what the AMD could do with a bit of eDRAM, HBM or even a lower-power variant of GDDR5… but I guess the latency issues could kill browser performance and that is unfortunately a large chunk of what buyers would want these for…

    Still dreaming of a way to put well-proportioned APUs in a scalable system with 1x/2x/4x configs… With storage and RAM no longer eating box space, 75/150/300 Watt configs could be relatively small yet remain quiet.

    Speaking of idle power and quiet, this is where I get interested in the AMD. The NUC is great in everything but noise on peak load, but it would really only take a replacement top and a Noctua to make it great… There is so much space behind these giant 4k screens, nuc/NUC can become a little pointless.

    Good Linux support is where I am getting concerned. Current reports praise AMD on their Linux vision… but progress seems a very different story and one where Intel (sorry Charly), really shines, even Nvidia seems better in practical terms (sorry Linus). I’m also somewhat disheartened by power management there: Not sure I’ll be able to reach 10 Watt of idle on CentOS or Ubuntu *and* Steam/Vulkan performance comparable to Windows (it’s actually gotten quite good on bigger Nvidia GPUs, even GPU pass-through to a Windows VM is kind of fun).
  • sor - Saturday, April 27, 2019 - link

    As I mentioned probably as you were typing this, I ran Ubuntu straight out of the box and am getting nearly 50% better FPS than this review on Dota 2. Full vulkan support and max settings. Pleasantly surprised, I am used to having to tinker with drivers.

    Notably, I don’t think this would have been possible 8 months ago as only newer kernels have the good AMD support built in.
  • Pishi86 - Saturday, April 27, 2019 - link

    This is not exactly a fair comparison. You are comparing a desktop AMD chip with and a mobile Intel chip. Its kind of like comparing an i3 8100 vs a Ryzen 5 3500u. AMD's Ryzen 5 3550H and Ryzen 7 3750H would have been more competitive. These chips are about as fast as the 2400G, but with an maximum TDP of 35w. There are some reviews on Notebookcheck and these chips are consuming just over 70w underload. This is with a 15.6 1080P screen and a power hungry Radeon RX 560X. The power consumption and battery life is actually better than an i5 8300H and 1050 combo with an identical. Check out the review below.

    https://www.notebookcheck.net/Asus-TUF-FX505DY-Ryz...

    The truth is the onboard Vega on Ryzen is a very powerful iGPU held back by memory bandwidth. Unrestrained, its probably 80-90% as powerful as an RX 460. It has 640-704 Vega cores which are clocked higher (1.2-1.4GHz) than the 896 cores in the RX 460. Vega's IPC should be a bit above Polaris's.

    I agree with you Linux support is spotty, I am a Linux user myself and I am in the market for a new laptop, but I may have to buy Intel despite its weak iGPU. Unfortunately, you can't find anymore Iris powered laptops these days (outside the macbook pro). Also, even though its improved AMD's video decode/encode is not as efficient as Intel's. I am not even sure if Nvidia is as efficient as Intel in video playback. Having that said I would not trust Intel's UHD graphics powering a 4k monitor, which is what I am in the market for.
  • Pishi86 - Friday, April 26, 2019 - link

    Does anyone know if you could get a 3rd party power supply that's more than 120w? I mean 150w might be good, if AMD releases 95W APUs in the future. A 120W PSU might limit CPUs abover 65W.
  • Lucky Stripes 99 - Sunday, May 5, 2019 - link

    Yes, there are third party power bricks available that can supply more current. Just keep in mind that the power regulators on the motherboard may not be rated for that higher current and that you could shorten its life or run into stability issues if you attempted to use a more power hungry processor (assuming if the BIOS would even make it past POST with an unsupported processor).
  • Haawser - Saturday, April 27, 2019 - link

    Good luck buying one...Only EU retailers I could find seem to have sold out within hours. Still, will keep trying. As this is exactly the sort of SFF I've wanted since Ryzen APUs came out.
  • ganeshts - Saturday, April 27, 2019 - link

    On Amazon and Newegg, Computer Upgrade King seems to have lot of ready-to-go models with the DeskMini A300 ; Eg: https://www.newegg.com/Product/Product.aspx?Item=9... (Just FYI - I have no idea about the reputation of this retailer. Just came up during my search on Google)
  • Haawser - Saturday, April 27, 2019 - link

    @ganeshts- Again, out of stock. Personally I think ASRock, Lenovo, HP, Zotac and everybody else that manufacturers SFF PCs have greatly underestimated the number of people looking to buy Ryzen APU based systems. And with the improved 3000 series (12nm Zen+ with soldered HS) soon available, the barebones will be even more sought after.
  • ganeshts - Sunday, April 28, 2019 - link

    Shows in-stock for me when I added to cart : https://i.imgur.com/YWbYlJ6.png
  • oliwek - Saturday, April 27, 2019 - link

    For people in NL or BE, I bought mine from here, delivered promptly : https://www.megekko.nl/product/2321/237330/Barebon...

Log in

Don't have an account? Sign up now