Comparison: AV1 software vs IntelARC hardware Accelerated AV1 vs x264/x265 - scored with ssimulacra2

[@BTA_KeepItFun]

Question! Has anyone tried to encode the original file I included here with some of the same settings I did? Perhaps: 10bit SVT-AV1 30-3? I would be interested to hear what CPU you ran and what time it took to encode that 20s clip. =)

[@JohnDoe-qu9jr]

This video has been an incredible resource in understanding what all the knobs and dials do when encoding stuff. Thank you so much!

[@peterschmidt9942]

Thank you for taking the time to put this together. I've recently been ripping my Blurays and finding it a little daunting as there doesn't seem to be a lot of info out there either on hardware vs software settings. The encoding time of software vs hardware is significant compared to normal DVDs.

[@BTA_KeepItFun]

You're welcome =). Trial and error for your preferred results (size, time, quality, etc). For everyone with cheap electricity and plenty of time to run encodes I would recommend going with software encoding for the best compression/quality. If in a hurry and size really does not matter (we all run out of space eventually) go with higher file sizes and enjoy the speed that hardware encoding offers. Currently, I am doing my "master" encodes with 30-3 SVT-AV1 10Bit for archive/re-use purposes and I use Intel Arc hardware encoding for small preview versions of those "master" ones. *I'm not ripping Blu-rays/movie media, so I'm happy that I don't really need to go investigate film grain and how to match and re-create every single movie title's chosen visual effects with Handbrake. Some choose to spend hours-days-weeks to make the best possible media rips and that's okay, as long as you're happy with what you've done!

[@peterschmidt9942]

@@BTA_KeepItFun Cheers. Currently if I'm converting DVDs I'll use software. Even though software encodes probably still give better results with bluray, I think I'll use hardware with a bigger file size. They're still coming down to 25%-30% of the original and I haven't noticed much difference. I just can't leave one encode for 24-48hrs for a movie.

[@TransformHypnosis]

INCREDIBLY indepth and valuable data. THank you so much!

[@BTA_KeepItFun]

You're welcome!

[@XilenceCinema]

Thank you very much, this insanely detailed comparison is the best I've seen so far. Great work!

[@webtax]

People sometimes go crazy with hardware encoders, with but it's always been kinda of the same: HW encoders usually target lower quality, so they only make sense for people that pump out videos every day, and are not concerned with quality. For actual video storage where size and quality is the concern, software was and is still the way to go. This behemoth of work you did slams with facts and lamplights on not only that this keeps true for AV1 & intel, it also shows us the curve of where could one choose to land a good compromise for this decision, congrats & thank you

[@user-ru3il1br8j]

Thank you! Your efforts are very appreciated! As far as I understand, the size of AV1 also depends on the “denoise” and “film grain synthesis” settings, the impact of which is also easily measurable using ssimulacra2 (unlike VMAF). But these values apparently need to be selected for each video individually.

[@BTA_KeepItFun]

You're Welcome. I also believe so. There was too many variables to compare so I kinda skipped these all together. Tried to keep this simple enough. :)

[@gimpinator]

thank you for all the hard work and effort, much appreciated 🙂

[@oxkaioxo]

love you

[@Teh-Penguin]

What a resource. Thanks for doing this! Why is the Excel file 2.5gig?

[@BTA_KeepItFun]

Nice to hear! All the video files are included there for people to compare themself. Excel file is small alone :)

[@jacobmar2797]

very helpful. thank you! :D

[@Bryan-T]

After staring at the spreadsheet with the video paused I'm sure I would use preset 7 as it's much faster than 3 and only a small hit to quality. I would've loved to see a few x265 encodes with a faster preset than placebo. I don't think anyone actually uses that setting and I can't get a grasp of how much slower av1 is, realistically. The intel option is good if you value time over file size, but that seems to always be the case with hardware encoding. Hardware encoding has always made bloated files.

[@BTA_KeepItFun]

This makes sense, but for extreme cases using faster preset here 7 or x265 fast for example you'll sacrifice minor details Handbrake does not understand need to be there every frame. For example the snowflakes flickering on this video ( it is not very very visible/hurtful here), water slowly moving, bird flying on sky and other small details like these (for me) get too often "removed" from encoded file with faster presets. This is very dependant what the source files are. (Might have called them sauce files few times on the video..) Well I'm quite picky how much detail I'll accept to loose from source, but also want file sizes that make sense in the long run and are reusable without noticing it was not source.. Also visual/perceived quality can be very different than technical data driven comparison. As you have now all encoded files you could compare Quality setting 30/40 preset 7, with matching IntelAV1s. It's a tricky game what is "best". Perhaps if 7 is satisfying enough for your use case would you rather go with HW-Accelerated AV1 and perhaps give up little bit more space on the storage, but gain extremely fast encoding and good quality? My first comparisons included more x264/5 and also other quality settings (constant bitrate for example) quickly got to conclusion that Constant Quality is superior. Also did not include much of x264/5 other presets on this as I wanted to go with "best" x264/5 vs. AV1. I got Intel Arc expecting quite matching quality with faster encoding speeds and learned what HW-Accelerated actually is good at, speed with lesser quality. Now I know better. Just a bit of rambling to perhaps make a few thinking moments also delightful to see that my research has value for you =).

[@TylerDurden-up8hz]

Nice work, but what gpu are you using to encode AV1?

[@BTA_KeepItFun]

Intel Arc A750

[@Bryan-T]

Very thorough! Could I get a direct link to the spreadsheet? The link you provided requires downloading a 2.5GB file. Thanks for the hard work in putting this together.

[@BTA_KeepItFun]

Thank you kindly. Getting this out of my head in English and also being clear was a lot harder than I thought. For next similar type video I'll need to write proper script to make overall flow better. Good idea to separate Excel file from the whole pack. I'll do that tomorrow! :)

[@Puhnu]

Kassu spelade med mej vill du i nåste wekkoslut

[@DimbleWally]

Thank you for this comparison; this is exactly what I was looking for!

[@BTA_KeepItFun]

Awesome! Great to hear My effort to this is helpfull. :)