I used the latest version of Staxrip, updated only some files manually.
Here is the sample file, I cut it out from blu-ray without re-encoding: firstname.lastname@example.org (600 Mbyte)
You can download it here (until a week) https://wsi.li/BEAWwFwfV7wQ
and you can repeat the test too.
REMEMBER: the file sharing service will delete automatically the files after 7days.
X265 2pass SLOW (factory settings) WAS achieved by NVENC h265 with this settings:
With the help of NVIDIA HW image enhancer (very very extra smart über quality HW accelerated sharpener)
ENCODED VIDEO FILES are located there:
X265 2pass SLOW video (88 Mbyte)
NVENC HEVC (88 Mbyte)
NVENC HEVC STAXRIP settings:
THE MOST IMPORTANT SETTING of the enhancer:
It is shocking good!!!!!!
I used only my GPU for image enhancement process and compress, and I achieved the quality of the X265 2pass SLOW.
So a video card with clever settings can create very very very good quality.
Images: Please, download the pictures and enlarge them.
X265 2pass SLOW:
NVENC HEVC pictures
+ Reply to Thread
Results 1 to 30 of 34
Last edited by Comparison; 8th Jul 2018 at 16:18.
If you look more closely at the video samples, NVEnc drops too much detail (it blurs it all away) and produces too many posterization artifacts . It's definitely worse
Sharpening / edge enhancement is probably counterproductive for NVEnc at low bitrates
Around 70 - 80% of the frames of the video, the NVENC is much closer to the quality of the original video source, around 20-30% the X265 is closer.
So I found the quality of the NVENC HW sharpened version much better in most frames than the X265.
That's because this test debunked the very well spreaded old stereotype/myth.Because , maybe nobody tried these exact settings in HW enhancer/sharpener of NVENC HEVC. The reactions are similar when an old faith myth is debunked. Many people don't like "the debunking reality" in the beginnings, until they will try it themselves.
Welcome to the reality!
Without sharpening, the natural NVENC HEVC is much worse and much dimmer than the older backwarder NVENC AVC!!!!
Last edited by Comparison; 8th Jul 2018 at 15:25.
I did download them and you're clearly mistaken.
Instead of viewing side by side, interleave the videos. You can see how poorly NVEnc performs - it's missing all the fine HD details and textures, all the fine grain in the original. It's blurring everything away that makes it "HD quality". Adding a enhancer / sharpener actually makes it worse, more difficult for NVEnc to compress. It would be smarter to add that as a playback option
Not even close here.
This is another test that reaffirms how much better x265 is than NVEnc . Welcome to reality!
Not even close . It's the opposite. Clearly x265 retains more fine details and textures. NVEnc blurs everything away . NVEnc is clearly worse by far . It's not me that needs my eyes checked. Good luck convincing anyone...
Your screenshots aren't even of the same frame. You should learn how to do proper comparisons. Sorry, but clearly you are new at this. One way is interleaving in avisynth, or use something like avspmod to compare the same frames (load each video in different tabs and switch with number keys)
If you're comparing letters and text, you could sharpen x265 too before you encoding. Sharpening make it less like the original; it's a post processing subjective effect . Sharpening makes it more difficult to encode (requires higher bitrate to achieve a given level of quality). New people tend to oversharpen things, but it's conterproductive for compression, especially at lower bitrates
You will eventually learn, good luck with this
I'd have to agree with poisondeathray. The x265 encode retains more detail and has less posterization. And the NVEnc encode is over sharpened. Though the low bitrate is good for comparing, it's much lower than I would use for storing my videos.
I saw a problem with the x265 encoded video: The luma and chroma channels have dropped by about 1 unit, causing the picture to be a little darker and discolored. That might have been caused by the decoder I was using (ffVideoSource()) and its conversion from 10 bit to 8 bit.
cpu beats every time hands down been there done that - just simple facts .... :/
In addition, this is the second thread to the same topic, the "definitve" conclusion about subjective opinions, and the lurid headline (clickbait style) and insulting comment does not make it appear objective.
The secret of optimal quality retention is taking efforts in not losing details at first. Once they are lost, their loss can only be camouflaged, but not repaired. GPU encoders are certainly handicapped in comparison with CPU encoders, due to a limited set of instructions (GPU instructions are not optimal for video encoding demands) and a smaller scope (especially temporal scope across a number of frames), possibly also a limited amount of available memory to store frame content to be compared. You will notice a severe limit of B frames and reference frames in GPU encoded results, both of which are the main advantage factors in modern efficient video encoding.
"Sharpness" is always an attribute of the original material. It can only be decreased by a blurring process (starting with frequency dependent quantization). Applying a filter with such a name does not restore original sharpness, instead it adds artificial contrasts and ringing artifacts (see Gibbs phenomenon). Any objective metric comparing original and copy will identify this as a difference, and differences are defined as "quality loss", regarding objective metrics. Subjective impressions may vary, though, depending on personal preferences.
Last edited by Comparison; 9th Jul 2018 at 05:07.
Marketing vs. years of experience...
How could an encoder separate "unwanted" grain from "useful details"? To achieve that, wouldn't it have to "understand" the video? This was always a challenge to the field of video encoding, and still is ... and until this problem is solved, a video codec should rather pessimistically retain grain than optimistically remove details which may be accidently misinterpreted as "unwanted".
Last edited by Comparison; 9th Jul 2018 at 06:04.
Let me guess ... the good old SAO issue (smooth all out).
Guys! Even old good x264 (2-pass + preset veryslow @ 3796kbps) looks much better than NVENC. I would say even more detailed than x265!
Comparison try something with more details at high frequency and let me know what you got on that NVENC.
It looks much closer to source footage. That's what matters!
Last edited by Comparison; 9th Jul 2018 at 08:14.
I'm talking about blurfest in NVENC! Jacket has been totally washed-out. Not to mention about hands.
Cropped. I do not have to post whole 1920x1080 frame to show you what I mean.
Comparison, NVENC HEVC is clearly better than it used to be and may be the best GPU encoder, especially if you are using an original source or anime. But it is also clearly inferior to x265. I use NVENC all the time as a hobby with the same frontend as you do (Staxrip). And yes I also update the files in the hope that I get better quality out of it. That said NVENC has never been as good as x265. But it is definitely usable and getting a whole lot better.
Encode these files and show us how NVENC beats competition. I do not have NVIDIA card.
BTW. So called "grain kind" is one. x264 not x265.
"Give us this day our daily grain,"
I'm still waiting for those files...
nVidia needs to stick to graphics cards. The dabbling in video has always been craptastic. nVidia has always wanted to be ATI, even after ATI (now AMD) is no longer ATI. When even freeware can outperform you, you're doing something wrong.
Originally Posted by Comparison
Human eye don't see grain on natural objects, grain is just a film industry art.
Clever sharpening can show things which are barely visible (close to non-existent) , even in the original recorded video material. I know it from experience from post sharpening of video camera records, long before you started to notice the simple existence of digital videos.
I think you are a troll, who even write about Politics in a 100% non-political site.
Don't worry about user poisondeathray, he does not belong to the Generation Snowflake. https://en.wikipedia.org/wiki/Generation_Snowflake
Last edited by Comparison; 9th Jul 2018 at 11:26.
What if a movie was shot on physical photo film (might be old, might be for style reasons). Does silver iodide (or similar modern photoreactive substances) not exist in the real world?
And as "clever" as sharpening algorithms may be ... after the encoder destroyed the details due to a too coarse quantization, it cannot be restored anymore. I hope you know what quantization does. Apparently you are able to use Wikipedia to insult people. Maybe you are also able to use it to teach yourself the basics of video encoding.