Is x265 half size reduction compared to x264 a real thing ?

[Yertq]

Hi,

I did a few tests with slow preset to compare x265 performance to x264. x265 could not achieve the claimed goal of obtaining the same quality with half size. I'm wondering if x265 is not that good, or maybe it's advantageous only in some specific situations?

[davexnet]

I've seen some good size reduction, I'm only judging it by eye, no special tools.
It seems with x265, the file gets bigger when the preset gets slower. Everyone has a different idea about this,
but I usually start with x265, fast, CRF 22-23 and compare to x264, slow, CRF 18-20

[ConsumerDV]

I did a few tests with slow preset to compare x265 performance to x264. x265 could not achieve the claimed goal of obtaining the same quality with half size. I'm wondering if x265 is not that good, or maybe it's advantageous only in some specific situations?

How do you measure quality?

[Yertq]

How do you measure quality?

This is my method:

1- I i take a PNG snapshot from the x264 video using Potplayer, then i do the same with the x265 video ( both frames must be taken exactly from the same video position).

2- I put the 2 snapshots inside an empty folder.

3- I open 1 of them in full screen mode with Windows Image Viewer.

4- I use the right and left arrows from the keyboard to jump between both frames. This way i can notice any tiny difference.

[Yertq]

I've seen some good size reduction, I'm only judging it by eye, no special tools.

You mean judging after playing the videos, or doing something else ?

[Luke M]

The first thing to understand is that claimed codec improvements only apply to low quality levels. As quality improves, the difference between codecs decreases.

[jagabo]

In my experience x265 does not deliver the same quality at 1/2 the bitrate of x264. It's more like 2/3 to 3/4. And a lot depends on the nature of the source. x265 (and other HEVC encoders) does best with large frame sizes and very low noise video.

x265 (and other HEVC encoders) tend to smooth away film grain and CCD noise, leading to posterization artifacts. You can alleviate that to some extent by using 10 bit encoding (even with 8 bit sources), and 10 bit decoding (even on 8 bit displays).

And of course, I'm not talking about re-encoding an existing a x264 video with x265 and comparing the two. I'm talking about encoding both from the same high quality source, then comparing the two new videos.

[ConsumerDV]

Indeed, here is an illustration how H.265 can use larger blocks:



This short article claims that in some cases H.264 could be more truthful to the original because H.265 removes small details:



This graph shows that for the same quality, meaning the same amount of differences from the source, the bitrates of H.264 and H.265 are not very different. Frankly, I do not see from the graph why the author claims, "a bitrate saving averaging 41% can be achieved in the average case (and just 8% for the difficult case)".

Calculated quality does not necessarily translate into visual quality, as removed noise and speckles may be considered an improvement.



I have not been eager to switch from H.264 to H.265. The former works just fine for me.

[VoodooFX]

Is x265 half size reduction compared to x264 a real thing ?

Yes, in some rare cases, but usually it's ~40%-25% reduction. [My personal subjective measuring]

[Hoser Rob]

In my experience x265 does not deliver the same quality at 1/2 the bitrate of x264. It's more like 2/3 to 3/4. And a lot depends on the nature of the source. x265 (and other HEVC encoders) does best with large frame sizes and very low noise video.....

I was thinking somethine similar ... like IME to get 1080p that really LOOKS like 1080p in x264 you probably need at least 8Mb per second. With x265 it's mpore like 6Mb/sec. Those are rough guesses.

x265 seems to be optimized a lot better for lower bit rates thougfh. That makes sense to me, x264 was largely about BluRay while x265 is about streaming video.

[sophisticles]

"Is x265 half size reduction compared to x264 a real thing ?"

The answer is: it depends.

It depends on what your expectations are, the source, the target quality, it depends on a number of things, much like, broadly speaking, AVC at half the file size of MPEG-2 is real depending on a large number of things.

For me the real question is why would anyone use a software based encoder for anything these days?

Ice Lake's QSV is very good, Tiger Lake improved on it considerably, Alder Lake improved on it even more, Intel now has a dGPU with hardware AV1 support that you can buy for $150, Rocket Lake will also feature the same hardware AV1 encoder, Stax Rip already supports it, NVIDIA's NVENC is great, one has to believe that they too will soon have a hardware AV1 encoder and Hand Brake now supports SVT-AV1, so why would anyone want to use either x264 or x265 for anything other than wasting electricity?

[blud7]

It is for animation and movies with a lot of dark scenes. AV1 is great, but no so great when it can't play on any of my TV sets

[Hoser Rob]

... no so great when it can't play on any of my TV sets

I thinki that's really the one valid reason not to encode to HEVC.

[pandy]

To be honest i never heard before about claim that x265 deliver same quality as x264 but at a half of the x264 bitrate.
However i saw claims that generally H.265 should be able to deliver same quality with half of required bitrate for H.264 but under some conditions - one of this conditions is resolution that should be UHD like (for example 3840x2160) - this is quite clear as H.265 was created with UHD on mind where H.264 was tailored for HD.

So to fairly compare H.265 with claim of half bitrate when reference is H.264 we should probably focus on progressive UHD content then i think H.265 may fulfill half bitrate claim.