i7 9750H, 630HD IGP, GTX 1660i showdown

Post your testing results with HandBrake.
Post Reply
razorback
Posts: 15
Joined: Wed Apr 25, 2018 7:47 am

i7 9750H, 630HD IGP, GTX 1660i showdown

Post by razorback » Mon Sep 02, 2019 4:19 pm

Hi Friends,

I just got my new Notebook up and running and wanted to find out, what the different coding options are good for.

The Machine:
Notebook ASUS ROG STRIX 51GU
CPU Intel i7 9750H (Hexacore)
Memory 16 GB 2666 Mhz
IGP Intel 630 HD
Nvidia GTX 1660i (6GB)
1 TB SSD NVMe Samsung 970 Pro

Windows 10 Pro 1903
Handbrake 1.2.2

Source: H.264
BluRay Avengers Endgame (ISO) 1920x1080 H.264, Duration 3:01:00, Total Bitrate 32,68 Mbps, 4 Audiotracks, several subtitles, ISO-File Size 43 GB

Target: H.265
MP4, Cropping Auto, Audio DTS-HD PassThru, E-AC3 PassThru, others abandoned, Subtitles foreign only, chapters retained.
Framerate as Source
Constant Quality RF 15

With the above Settings I handbraked the source using (only changing the Encoder Setting)
- Encoder H.265
- Encoder H265 QSV
- Encoder H265 NVEnc

The results were quite interesting:

The Hands Down Winner is NVEnc as well in terms of speed as in terms of quality !!!

H.265 NVEnc, Avg Framrate 215 fps (equals to 20 min encoding time),
Target File Overall Birate(MediaInfo) 15,4 Mb/s, FileSize 19,4 GB
Quality on my LG OLED outstanding.

Second:
H.265 Intel QSV, Avg 95 fps ( = 47 min encoding)
Target File Overall Bitrate 11,1 Mb/s (why so ?), FileSize 14,0 GB

Third:
H.265 CPU, Avg 30 fps ( = 146 min encoding)
Target File Overall Bitrate 10,3 Mb/s (why so again ?), File Size 13,1 GB

The quality loss of the losers is evidently the loss of bitrate - for reasons to me unknown.
As I said - I only have changed the encoder settiings, all others left equal.

And - the speed of the source / target medium does not matter at all..... no difference if on local SSD or NAS.

User avatar
s55
HandBrake Team
Posts: 9545
Joined: Sun Dec 24, 2006 1:05 pm

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by s55 » Mon Sep 02, 2019 4:48 pm

As I said - I only have changed the encoder settiings, all others left equal.
It's not actually that simple. You have infact changed many variables by not changing anything. For example, x265 and the hardware encoders use different rate control methods, meaing RF 22 is not the same as CQ 22.

Also, the encoders all have different functionalities and defaults. It's probably not possible to get an exact match on settings.


Typically. Not always true but holds in common occurrence i'd expect:
  • If you want speed: NVEnc > QuickSync > VCE > x265
  • If you want quality:x265 > QuickSync > Nvenc > VCE (The gap is much closer with H.265 than it is with the equiv H.264 encoders). x264 is extremely well optimised
  • If you want filesize: x265 > QuickSync >NVenc > VCE
It obviously varies a bit with hardware etc.

(I don't have Turing or Navi data yet. Note the 1660 doesn't have Turings new ASIC as far as I'm aware. It's still the older NVenc version. )


So ironically you got the opposite results to what one would expect. That said, it's likely down to the overly inflated filesize.

razorback
Posts: 15
Joined: Wed Apr 25, 2018 7:47 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by razorback » Tue Sep 03, 2019 12:58 pm

Thank you s55 !

of course I do know that in terms of quality/speed/filesize it`s something like a ride on a razor blade :-)
(or the choice between plague and cholera.....)

What I have found out is that when recoding a H.264 Video to H.265 you will get the same quality/fidelity with 66% to 50% of the videobitrate of the H.264 source.

So referring to my original post, at RF 15 Encoder settings h265 NVEnc yields about 50% more video bitrate than both of the others.

That gives me the best off all three worlds (speed, quality and filesize).

What you have mentioned and I still don't understand is this difference in videobitrate between these three encoders.
I would have expected more or less the same result by leaving RF 15 alone - just changing the encoders.

As for my GTX 1660 Ti - HWinfo reports Chipsetname: TU116M

Ah yes - another interesting experience:
During encoding I started Taskmanager and found out that when using the GPU encoders the GPU-memory was barely used (0,5 GB out of 6144 on NVidia, 0,2GB out of 2048 on QSV).
Why so ?

Woodstock
Veteran User
Posts: 3219
Joined: Tue Aug 27, 2013 6:39 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by Woodstock » Tue Sep 03, 2019 1:25 pm

"GPU encoders" is a misnomer - the video encoding hardware is separate from the GPU, so GPU processors and memory aren't used.

razorback
Posts: 15
Joined: Wed Apr 25, 2018 7:47 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by razorback » Tue Sep 03, 2019 1:37 pm

thx woodstock, I didn't know that.

but then: WHEN is GPU memory used ??

Woodstock
Veteran User
Posts: 3219
Joined: Tue Aug 27, 2013 6:39 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by Woodstock » Tue Sep 03, 2019 1:46 pm

Personally, I don't know - while I have video cards with GPUs, I don't have any programs that would use a GPUs.

User avatar
Rodeo
HandBrake Team
Posts: 12209
Joined: Tue Mar 03, 2009 8:55 pm

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by Rodeo » Tue Sep 03, 2019 2:49 pm

Woodstock wrote:
Tue Sep 03, 2019 1:25 pm
"GPU encoders" is a misnomer - the video encoding hardware is separate from the GPU, so GPU processors and memory aren't used.
Erm the VRAM is used, the hardware-accelerated video encoder doesn't have any direct access to RAM that I know of.

razorback
Posts: 15
Joined: Wed Apr 25, 2018 7:47 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by razorback » Tue Sep 03, 2019 3:32 pm

well, that still leaves me out in the rain....

What are those 6GB VRAM used for ?

Gaming ??? which I don't do at all......

mduell
Veteran User
Posts: 6680
Joined: Sat Apr 21, 2007 8:54 pm

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by mduell » Tue Sep 03, 2019 3:34 pm

With an output so large, why not just keep the source and avoid the quality loss?

razorback
Posts: 15
Joined: Wed Apr 25, 2018 7:47 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by razorback » Tue Sep 03, 2019 4:33 pm

what quality loss ??

h.264 @ 32 Mb/s is not better than h.265 @ 15 Mb/s qualitywise but the filesize is only 50% of the original and the encoding for 3 hours playing time in 20 mins (NVEnc RF15) is negligeable.

After all, the VISIBLE Quality is depending on the player (and the TV of course).
For playback I use an OPPO 203 against an LG OLED 55C6V and the results are perfect.
I have a HiMedia Q5Pro as well for ISO Playback and (for my vision) I cannot see any quality degradation when playing the ISO in comparison to the handbraked h.265 .
And to top it all, the OPPO does some Dolby Vision Magic in transcoding the handbraked M4V on the fly to Dolby Vision ------ Stunning Experience :-)))

nhyone
Bright Spark User
Posts: 221
Joined: Fri Jul 24, 2015 4:13 am

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by nhyone » Wed Sep 04, 2019 9:51 am

Hmm, a 32 Mbps 1080p H.264 video file.

I bet it would be efficient (10 Mbps) and look just as nice if you re-encode it using x264 CRF 20.

tlindgren
Bright Spark User
Posts: 238
Joined: Sun May 03, 2009 2:14 pm

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by tlindgren » Thu Sep 05, 2019 4:27 pm

s55 wrote:
Mon Sep 02, 2019 4:48 pm
(I don't have Turing or Navi data yet. Note the 1660 doesn't have Turings new ASIC as far as I'm aware. It's still the older NVenc version. )
No, 1650 has the older Volta encoder, 1660 and up have the new much improved Turing encoder. For the Turing series Nvidia uses the exact same (and same amount of) encoder silicon from 1660 to 2080 Ti.

The Volta encoder perform very similar to the Pascal encoder used in 10xx cards, since 1650 has one Volta encoder it should be similar to say a 1060 for NVenc work (the fastest Pascal card with "only" one Pascal NVenc module).

https://developer.nvidia.com/video-enco ... ort-matrix

User avatar
s55
HandBrake Team
Posts: 9545
Joined: Sun Dec 24, 2006 1:05 pm

Re: i7 9750H, 630HD IGP, GTX 1660i showdown

Post by s55 » Thu Sep 05, 2019 5:32 pm

Cool. Thanks for checking into that

Post Reply