• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA 4000 Series

I took a random 1 minute segment of one of my Shadowplay recordings and tested NVEnc HEVC and AV1 on it with a different CQ value to get a very similar file size.

NVEnc HEVC - CQ 28 slowest preset. File size 235,636 KB.
NVEnc AV1 - CQ 36 slowest preset. File size 235,635 KB.
Overall bitrate is 31.9 Mb/s in both.

Then I took a screenshot of a random frame and did a 1920x1080 crop so you can more easily see the differences in quality. Can you tell which one is which?

Source.png
Test-1.png
Test-2.png
Will check when home tonight!
 
Yes, full screen, zoomed to fill v axis, each image opened in a new tab and then cycled rapidly with CTRL+TAB and saw no meaningful differences just scanning the whole frame as I cycle the tabs.
 
If the pixels aren't changing, i.e. it is a still shot then you don't need much bitrate for it to look like the source. The accumulation of data gets you there.

You get artifacting and blocking in motion (i.e. where new information has to be encoded).

Combined that with a very high bitrate for a 1080p video, and you won't see a difference.
 
Last edited:
If I was to guess I'd say encode 2 is av1, slightly crisper background and it has a higher cq
Encode 1 is AV1. Both of them lose quite a lot of detail but NVEnc AV1 is definitely worse than NVEnc HEVC, at least in the Handbrake snapshot build. I don't know if it's something they can improve with it being hardware based, but it seems that the main purpose is to avoid royalties in HEVC. I guess that's why they didn't compare AV1 to HEVC in their blog post.


Yes, full screen, zoomed to fill v axis, each image opened in a new tab and then cycled rapidly with CTRL+TAB and saw no meaningful differences just scanning the whole frame as I cycle the tabs.
Look at a specific area like the wall, you should be able to easily see differences.
 
Encode 1 is AV1. Both of them lose quite a lot of detail but NVEnc AV1 is definitely worse than NVEnc HEVC, at least in the Handbrake snapshot build. I don't know if it's something they can improve with it being hardware based, but it seems that the main purpose is to avoid royalties in HEVC. I guess that's why they didn't compare AV1 to HEVC in their blog post.
Bit weird, I've been hearing all this stuff about av1 being the 'saviour' for low files sizes etc keeping good quality but to me at least it's the worst one of the two at the same file size/bitrate....based on your images (and I'm sure there will be others with different outcomes) the marketing is more about saving money from paying royalties...

Was there a difference in the amount of time it spent encoding?

You're right in the fact it could be down to the hardware/software integration at this point with everything being quite new (think it was similar with nvenc when first added) but considering how hard it's being pushed you'd have thought it would be fairly far along before being added... although I did read av1 is only now being added to ffmpeg etc which might be part of it.
 
AV1 was always marketed for its streaming capabilities as they say you don't need as high a bitrate to get the same quality as H265 and thus file sizes are smaller so can stream higher quality faster or upload faster and save storage space etc.

The other upshot is if you have a 40 series, then you get dual hardware encoders so can encode two files at the same speed (apparently)

I'm just having a play around seeing when quality drops off noticeably but the dual AV1 encoders is really cool, it was unclear if the dual encoders were just for AV1 or any codec, now having tested it out, I can see it is only for AV1, if I put two tasks, one AV1 and one H265 in the queue, then both encodes will start but the H265 one will encode at 60fps or so whilst the AV1 is over 115fps, bot two AV1 tasks in the queue though, and both run at +115fps

Probably best leaving number of encodes in one go to just 1 if not encoding exclusively in AV1 I guess :p

Edit* It's actually faster if you just do a single encode at a time too, so there's clearly a single piece of component sitting between the dual encoders resulting in some reduced speed if you encode two in one go.

Edit 2*

Using the Slowest preset encoder mode shows some interesting results. H264 is still fast using it, but H265 and H264 are noticeably slower in Slowest. AV1 remains unaffected. So it's clear that AV1 can perform better as a result of using the more efficient (slower)encode preset, but not actually encode slower still compared to H265. The file size though at the same CQ between them both weigh in favour of H265 of H264 as you can see:


5oHYhGx.png
 
Last edited:
Back
Top Bottom