HTPC Credentials - YouTube and Netflix Streaming

Streaming of 4K OTT content with YouTube and Netflix has become mainstream. HDR displays have also started becoming affordable. In order to evaluate YouTube HDR streaming support, we have the Mystery Box's Peru 8K HDR 60FPS video as our test sample. On PCs running Windows, it is recommended that HDR streaming videos be viewed using the Microsoft Edge browser after putting the desktop in HDR mode.

HDR content on YouTube is encoded exclusively in VP9 Profile 2. On systems supporting hardware accelerated decoding of VP9 Profile 2, Edge streams and decodes the HDR stream. The DeskMini A300 passes this test, and we can see HDR YouTube videos, as shown in the picture above.

Various metrics of interest such as GPU usage and at-wall power consumption were recorded for the first four minutes of the playback of the above video. The numbers are graphed below.

Decoding 4Kp60 VP9 Profile 2 videos consumes between 70 and 90% of one of the decoder engines. We also see some Direct3D loading of the order of 70% initially (when the stream being played back is at a lower resolution and needs to be upscaled to 4K). Once the stream stabilizes and the system starts decoding the 4K stream, the Direct3D loading settles around 50%. The average at-wall power consumption for streaming 4K YouTube HDR videos is around 33W.

4K HDR streaming in Netflix requires support for hardware-accelerated decoding of HEVC Main 10 videos. Microsoft removed the in-built HEVC decoding capabilities of Windows 10 in the 2017 Fall Creators Update, and replaced it with an extension that had to be downloaded from the Microsoft Store. Without the extension, playback is restricted to 1080p non-HDR streams encoded in H.264. In addition to the decoding capabilities, the system also needs to support PlayReady 3.0 DRM. In our tests with the Radeon 19.4.2 driver release, we ran into major problems.

Our usual testing flow involves the installation of the HEVC Video Extensions and the Netflix apps from the Microsoft Store. After configuring the Netflix login credentials, we attempt to play back Season 4 Episode 4 of the Netflix Test Patterns title. Unfortunately, the DeskMini A300 behaves in an unexpected manner when subject to the above steps. As soon as the playback of the title starts, the screen goes blank and the display driver crashes.

Windows reverts to the basic display driver, and the Netflix title continues to play back. However, instead of the expected 4Kp60 HEVC Main 10 stream (with HDR), we only get a 1080p30 AVC stream. In order to prevent the crashing of the driver in the course of our evaluation, we opted to uninstall the HEVC Video Extensions app, and make use of only the AVC decoder built into Windows. Without the HEVC Video Extensions, the driver has no problem with standard 1080p Netflix streaming. The OS screenshot facilities obviously can't capture the video being played back. However, the debug OSD (reachable by Ctrl-Alt-Shift-D) can be recorded.

The (avc) entry corresponding to the Video Track in the debug OSD, the Framerate (30.000), and with the A/V bitrate details (192 kbps / 5800 kbps) indicate that we are only getting a low quality 1080p stream compared to the best available one (4Kp60 16 Mbps HEVC Main 10 HDR stream). Similar to the YouTube streaming case, metrics such as GPU usage and at-wall power consumption were recorded for the first three minutes of the playback of the title. The numbers are graphed below.

1080p30 AVC playback is a walk in the park for modern computing systems. We see that the Direct3D loading is only around 30% and the decoder usage is less than 20% in the steady state. The former is mainly due  The at-wall power consumption is around 24W for this workload.

HTPC Credentials - Display Outputs Capabilities HTPC Credentials - Local Media Playback and Video Processing
Comments Locked

88 Comments

View All Comments

  • Haawser - Sunday, April 28, 2019 - link

    @oliwek That's the 310, it's an Intel barebones system, not Ryzen.
  • sor - Saturday, April 27, 2019 - link

    If you sign up for Newegg notifications you’ll probably have one within a week. I got one of mine about three weeks ago and the other last week.
  • yankeeDDL - Monday, April 29, 2019 - link

    I'm just going to throw it as a suggestion.
    I understand the purpose and the rationale in comparing similarly priced models, and all relatively recent/available, however ... I think it would add an enormous value to "normal" users to be able to somewhat put things in perspective with slightly dated hardware.

    I am not saying that we should be able to compare the Ryzen 5 2400G with an 80486, but, to give you an example, I have an A10-8700P and I have been considering an upgrade, but it seems really difficult to find a way to get an idea of just how much faster the 2400G is.
    The A10-8700P is certainly not efficient, but it does have 4 cores, and a decent iGPU, already based on the GCN. There's no question that the 2400G will trounce it in efficiency, but is it a worth upgrade?

    It is just an example, to indicate that having also 2-3 previous generations in the comparisons would not be such a bad idea, in my opinion. Just a thought.
  • Irata - Wednesday, May 1, 2019 - link

    If you are looking for this type of comparison, planet3dnow de has one between an Athlon 5350 (Jaguar) vs Athlon 200GE. It's in German but maybe you can get useable results using Google translate.

    They also did a review of the Hewlett-Packard 17-ca1004ng notebook with a Ryzen 5 3500U comparing it to its Bristol Ridge predecessor.

    Both show just how far AMD has come with Ryzen.
  • mikato - Tuesday, April 30, 2019 - link

    Ganesh, you didn't talk about noise at all. I feel like that should always be included in these mini PC articles. You have 3 pages on HTPC credentials and 1 page on Power Consumption and Thermal Performance but no mention of noise. How does it compare to the Intel NUC8i7BEH that you mentioned had a noisy fan with any load? Or the ZBOX CI660 which was fanless but had a dull whine and a strange grating sound though it wasn't noticeable beyond 10 feet? Did you think that optional CPU heatsink/fan they included was adequate?

    By the way, 10 feet isn't close enough. How about 6 feet? If you give one noise impression from 10 feet, you could move a few feet closer and check what it is at 6 feet as well... for future reviews.

    If you keep doing these Mini PC reviews, I will keep reading them!
  • Lucky Stripes 99 - Sunday, May 5, 2019 - link

    I was also puzzled at the lack noise testing. One of the reasons I'm looking at mini-STX and thin mini-ITX systems instead of a NUC is because of the fan noise many of those NUC systems suffer from.

    Some folks have been installing low-profile Noctua CPU coolers in their Deskmini systems and they report that they are "nearly silent", even while under heavy load. I'd really like to know how silent that really is.
  • werpu - Monday, May 6, 2019 - link

    I have a ryzen 2400g mini itx system with the flat noctua, and it is very silent, even under heavy load it is almost not hearable. I however had to add 2 smaller noctuas later in my system because the vrm would get rather hot with my standalone noctua and the case I was using. Still a very silent system and definitely less noise than a nuc.
  • mikato - Tuesday, April 30, 2019 - link

    At first I thought Bean Canyon was the best way to go for an HTPC with the performance and power efficiency. But with an idle power usage of 11.24 watts for this DeckMini A300 vs the 8.45 watts of the NUC8i7BEH with Bean Canyon - that difference is so little. As my HTPC, it would be left on and most of the time it would be idling. And the "as configured, no OS" prices in your reviews for these were $465 for the DeskMini A300 and $963 for the NUC8i7BEH...

    Let's see, your DeskMini A300 had:
    G.Skill Ripjaws F4-3000C16D-16GRS 2x8 GB (newegg $100)
    Western Digital WD SN500 500 GB (newegg $70)

    While your NUC8i7BEH had:
    G.Skill Ripjaws F4-3000C16D-32GRS 2x16 GB (newegg $185 - your review listed RipjawsV F4-3000C16-16GRS but I couldn't find that anywhere and the photos in your NUC8i7BEH review showed it was F4-3000C16D-32GRS instead so I think your specs table was incorrect)
    Western Digital WD Black 3D NVMe SSD (2018) 1 TB (newegg $238 for the SN750, older 2018 one costs more)

    Let's see... that's $185-100=85 and $238-70=168. 85+168=253
    So your combined storage and memory choices for the NUC8i7BEH cost $253 more.

    And the difference in your overall "as configured, no OS" prices was $963-465=498.

    So if I remove the difference in your storage and memory choices, I calculate that the DeskMini A300 is still $498-253 = $245 cheaper!

    Sure the A300 doesn't look as nice or compact as the NUC, but that is some solid money.
  • Irata - Wednesday, May 1, 2019 - link

    The Intel models are usually configured with higher end components vs. the AMD based models. Same for laptops where $600 AMD based notebooks go up against $1,600 Intel ones .

    I guess it's a hold over of the "AMD = budget" days.
  • mikato - Thursday, May 2, 2019 - link

    Well, it's not a laptop. What do you think might be a higher end component here? This is basically a case, motherboard, and power supply, right?

Log in

Don't have an account? Sign up now