Hello everyone,
I haven't done one of these for a while and now that I have a few days off I decided to do a new one, with a test sample that I have never used before. The source for this test is an excellent 10-bit, HEVC, 4k, HDR file named "LG Chess 4K Demo". I don't remember where I go this and it's too big to attach to this forum but a google search should turn it up.
The mediainfo for this file is as follows:
Video
ID : 257 (0x101)
Menu ID : 1 (0x1)
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : Main 10@L5.1@High
HDR format : SMPTE ST 2086, HDR10 compatible
Codec ID : 36
Duration : 1 min 52 s
Width : 3 840 pixels
Height : 2 160 pixels
Display aspect ratio : 16:9
Frame rate : 59.940 (60000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 10 bits
Writing library : ATEME Titan KFE 3.7.3 (4.7.3.1001)
Color range : Limited
Color primaries : BT.2020
Transfer characteristics : PQ
Matrix coefficients : BT.2020 non-constant
Mastering display color primaries : Display P3
Mastering display luminance : min: 0.0500 cd/m2, max: 1000 cd/m2
A few notes about this test, I used the latest build of Staxrip, which doesn't support this file, so I was forced to use the latest build of Shotcut to convert it to something Staxrip does support. When i loaded it into Shotcut, it asks if i want to convert it to an edit friendly format, which btw is the proper behavior, not like some half-baked, overpriced software that wants to convert everything to "ProRes-like".
I was given 3 choices and I chose the highest quality which resulted in a source file with this mediainfo:
Video
ID : 2
Format : YUV
Codec ID : V_MS/VFW/FOURCC / ULH2
Codec ID/Info : Ut Video Lossless Codec
Codec ID/Hint : Ut Video
Duration : 1 min 52 s
Bit rate : 2 016 Mb/s
Width : 3 840 pixels
Height : 2 160 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 59.940 FPS
Color space : YUV / YUV
Chroma subsampling : 4:2:2
Compression mode : Lossless
Bits/(Pixel*Frame) : 4.056
Stream size : 26.5 GiB (98%)
Writing library : Lavc58.54.100 utvideo
Default : Yes
Forced : No
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709
As a word to the wise, when I first did this test I did something stupid, something that I should know better since I have been using hardware encoders since they first came out and have posted numerous here that I have done on both Windows and Linux.
With the Intel encoder, on both Linux and Win 10, the OS used here, the default Intel drivers work for hardware decoding and encoding, but on Linux they only allow use via vaapi and on Win 10 they do not allow the use of all the features.
I had gotten accustomed to updating the open source drivers on Linux with the intel-media-sdk stack but on Windows I took it for granted that the default signed drivers from MS were the ideal ones and I also assumed that when I updated Windows 10 it was installing the ideal drivers.
I realized I had made a mistake because no matter what I did I could not get certain features to work, such as trellis, with the Intel encoder, features that I know are supported on Ice Lake.
I updated the Intel drivers with the latest drivers from Intel's website and things were fine, but I did end up having to redo the tests.
For this test I focused in the Intel encoder's CQP mode, because Intel has said that this mode, by far, produces the highest quality.
I used Staxip+QSVEnc and what I did was enable all the options for both H264 and H265 and then let the driver disable the ones that were unsupported. Intel's encoder is tricky, as is QSVEnc, some options do not work in combination with other options, for instance look-ahead does not work with mbbrc, and neither work with CQP.
One note, Intel's encoder has AQ enabled all the time, according to the docs, and there is no way to disable it. Mbbrc, based on the docs description, is similar to mb-tree; look-ahead only works with H264 for all modes except CQP, Intel's HEVC does not support look-ahead.
For a baseline comparison I created an x264 file with the slow preset and tune film and I created an x265 file with the fast preset; on this Ice Lake powered laptop both these encoders with their respective settings ran at between 4-5 fps.
X264 and x265 used crf 23, the Intel H264 Film and the Intel H265 used QPI 23, QPP 25 and QPB 28 and the Intel H265 ICQ used ICQ 23.
The Intel encoder was about 4-5 times faster than the software encoders with significantly less cpu load and thus much lower electricity use.
I did not test fixed function mode, which should be much faster and even lower power draw, I may repeat the Intel tests with fixed function enabled and post those files here as well.
Try StreamFab Downloader and download from Netflix, Amazon, Youtube! Or Try DVDFab and copy Blu-rays! or rip iTunes movies!
+ Reply to Thread
Results 1 to 1 of 1
Thread
Similar Threads
-
Ice Lake hardware deinterlacer test
By sophisticles in forum Video ConversionReplies: 0Last Post: 24th Jan 2021, 00:06 -
Fresh x264 10-bit vs Ice Lake hevc qsv 10-bit
By sophisticles in forum Video ConversionReplies: 5Last Post: 7th Jan 2021, 17:06 -
More Ice Lake Quick Sync test samples
By sophisticles in forum Video ConversionReplies: 13Last Post: 11th Nov 2020, 10:08 -
Ice Lake QSV Review/Test!!!
By sophisticles in forum Video ConversionReplies: 19Last Post: 1st Oct 2020, 10:14 -
Spectre and Melton Preformance Issues - Ice Lake vs Coffe lake
By Snakebyte1 in forum ComputerReplies: 0Last Post: 21st Mar 2018, 10:20