VRD Encoders

#1
Hi guys, i was testing out my new Dell Precision M4700 laptop today using VRD Pro, just to get an idea of how fast it would be at encoding compared to my previous Dell Inspiron 2330 All in One desktop pc, which i recently gave to my friend.

The 2330 has an i7 3770s desktop cpu, and gets an average of 8,800 cpu benchmark points, while the M4700 has an i7 3740QM mobile cpu, and gets around 8,300 cpu benchmark points.

I ran 2 lots of tests to see how fast the M4700 was, first test was rendering a 4k XAVC-S file to 1080p using VRD Intelligent Recode, and test 2 was the same outputs using VRD Force Recode,

This gave me a bit of an idea as to how each one compared to the other, as well as how each of the 5 encoders compared to each other, however it was apparent there was barely, if any difference between Default, MC Pro and Software, where as Intel Quicksync was by far the fastest, on average it was 50% faster than the others, and the file size and bitrate the same as the others, and the x264 Encoder had the slowest output time in both tests (not by much) with the same file size and bitrate as the others.

test 1. 125 second video - 4k-25p XAVC-S @ 95Mbps to 1080-25p using Intelligent Recode (output bitrate set by VRD settings)

Default Encoder: output time = 104sec, 180mb, 23Mbps
MC Pro Encoder: output time = 98sec, 182mb, 23Mbps
x264 Encoder: output time = 112sec, 180mb, 23Mbps
Intel Encoder: output time = 55sec, 179mb, 23Mbps
Software Encoder: output time = 106sec, 182mb, 23Mbps

test 2. 125 second video - 4k-25p XAVC-S @ 95Mbps to 1080-25p using Force Recode (with output bitrate manually set to 25Mbps)

Default Encoder: output time = 100sec, 197mb, 25Mbps
MC Pro Encoder: output time = 100sec, 197mb, 25Mbps
x264 Encoder: output time = 106sec, 195mb, 25Mbps
Intel Encoder: output time = 55sec, 195mb, 25Mbps
Software Encoder: output time = 99sec, 197mb, 25Mbps

I was using the Force Recode to 25Mbps before, but i think i will change my 4k to 1080p profiles back to Intelligent Recode, given there is only 2Mbps difference, and nobody can visually see the difference anyway.

QUESTION: can someone please explain the differences between the Default, MC Pro and Software Encoders in the settings box.

I know about Intel Quicksync, and how it can be many times faster than software encoding, depending on which software you are using and how they implement it, and i know VRD's version of Quicksync is quite a bit slower than it is in many other software programs, and i understand that the x264 encoder should be somewhat better than the regular h264, but by how much, as it does take a bit longer to encode using x264 in VRD, but it does give us a slightly smaller file size as well, for the same Bitrate.

Cheers
 
Last edited:

Dan203

Senior Developer
Staff member
#2
In your case Default, MC Pro and Software are all the same encoder. MC Pro is an add-on that costs a couple hundred extra so unless you paid for that upgrade it just drops back to the regular software encoder. (which is also MC, so probably same speed anyway) Default just chooses between the software encoder and the QuickSync encoder depending on how you have the default encoder option set in the Tools->Options->Stream Parameters section. Software simply overrides that option and forces the use of the software encoder.

In v6 we're going to have an option for NVEnc too. If you have an Nvidia card it can encode at 40% faster then even QuickSync. I've done test encodes where I got over 700fps. It's incredibly fast, and to my eyes looks fine.
 

jmc

Active member
#3
and i understand that the x264 encoder should be somewhat better than the regular h264, but by how much, as it does take a bit longer to encode using x264 in VRD, but it does give us a slightly smaller file size as well, for the same Bitrate.

Cheers
Hmm, maybe it is core count... I find VRD's X264 to be more then twice the speed of VRD's h264 encoder.
Thinking... that h264 is core limited. Think I remember someone here talking about that. Probably Dan203.


(6 core 3930 Sandybridge from 2012 FINALLY to move to 16 core ThreadRipper this month)

Ram prices have gone up $200 in two weeks.
64Gigs was $600-700, now $800-900
It's crazy. Hope there is a Black Friday sale!

Boo, :) No AMD support... But Nvidia does have the vast majority market share.
But I'm staying with AMD graphics. Nvidia has ethics issues. Can not use Nvidia products.

jmc
 
#4
In v6 we're going to have an option for NVEnc too. If you have an Nvidia card it can encode at 40% faster then even QuickSync. I've done test encodes where I got over 700fps. It's incredibly fast, and to my eyes looks fine.
Yes nVidia is so dominant now with NVEnc, and when i built my sons new X1700 Ryzen editing pc a few months ago, we put in a GTX-1080 8gb graphics card to utilize its processing power, and in his case, using Premier and DaVince Resolve both are being developed to support the new and faster nVidia cards.

My Dell Precision M4700 will be my temporary editing machine for a few months, but it has the Intel HD4000 graphics on the i7 3740qm cpu (can use Quicksync) plus it has a dedicated Quadro K2000M card as well, so i will be on the lookout for a bigger 17" laptop for editing, and i will be making sure it has a decent nVidia graphics card in it, in anticipation for the arrival of VRD version 6 and NVEnc.

Hmm, maybe it is core count... I find VRD's X264 to be more then twice the speed of VRD's h264 encoder.
Thinking... that h264 is core limited. Think I remember someone here talking about that. Probably Dan203.
Are you sure the x264 in VRD is twice as fast as the h264 encoder ? because i personally can't see how it could be, even the VRD implimentation of Intel Quicksync is barely twice as fast as the other encoders, as you can see by my results.

Can't see how it could be so different, even given the differences in our editing systems.
 
Last edited:

jmc

Active member
#5
Are you sure the x264 in VRD is twice as fast as the h264 encoder ? because i personally can't see how it could be, even the VRD implimentation of Intel Quicksync is barely twice as fast as the other encoders, as you can see by my results.

Can't see how it could be so different, even given the differences in our editing systems.
I've never been sure why my 3930 does X264 so well.
Is it the large cache or some other advanced bit of hardware in the consumer level "xeon" cpu.
I even cut it down to a quad core in the bios and still it was so much faster then quad core
the Devs tested...
----------------------------------
DanR-"3770 (4 core, 8 HT). x.264 was faster by about 25% at best."

My cpu with 4 cores (2 cores turned off) running "default mp4 profile" did 86% better.
----------------------------------


(I have other threads with a lot of 3930 6 core x264/h264 testing)
http://www.videoredo.net/msgBoard/showthread.php?35948-FYI-X264-seems-to-be-TWICE-as-fast-as-H264!

Now the L3 Preset makes VRD twice as slow as the default preset(L2)
And I believe makes more use of the cores.

-----------------------------------HD -NTSC
X264-17 FPS-----HD-60sec.m2ts 21.7 Mbps, L3Preset
H264--8.4 FPS---6 core CPU 4.3 Ghz

------------------------------------SD -PAL
X264-107.02 FPS-----DVD mpg 6.6Mbps, L3Preset
H264--45.78 FPS-----6 core CPU 4.3 Ghz

-----------------------------------------------------------------------------------

But the 8 core Ryzen is also almost twice the speed with X264 vs H264.


http://www.videoredo.net/msgBoard/showthread.php?36244-FYI-8core-Ryzen-results-H264-vs-X264


This is the main info with the 8 core Ryzen.
-----------------------------------------------------------------------------------------------------
Was interesting to see H264 ONLY using the actual cpu cores,
Just every other column was used in taskmanager.
That does not happen on my 6/12 core. Stuff jumps around in it.

Now using a standard dvd 5 minute MPG segment and the Default .MP4 profile
I got the following...


W10/64 H264------------------------------------------X264-----------------Ryzen cores 8/16
-------191 to 195 fps ----------------------------391fps (3.1Ghz)

W7/64 204 to 209 fps ------------------------391fps (3.1Ghz)
-------------------------------------------------------------------------------------


-------------------------------------------------------------------------------------cores 6/12 Sandybridge (2012)
My old 6/12 core beat the 8 cores with H264 because of the very
poor use of the 8cores. But then the 6core is overclocked to 4.3 Ghz

W7/64 -----H264----------------------------X264
------------225 fps ------------------------301fps (4.3 Ghz)

The 6 core overclock won with H264 but the 8/16 cores won with X264.
---------------------------------------------------------------------------------------------
 
#6
Ok the tests i did with those results above, i only had the M4700 laptop 1 week, so i only had a normal 500gb Hdd in the main Hdd bay with Win 10 installed, and i was storing the 4k source file on that, and exporting the 1080p output file back to that same drive, not the way you do it.

I now have a new Hdd caddy installed in the Dvd drive bay, and i have installed a brand new 2tb Hdd (seagate 7200rpm) in that to store my personal files, and to use for rendered exports.

I also have a 250gb Samsung 850 EVO SSD coming next week to replace the current 500gb Hdd that has windows 10 on it.

When i get that set up, i will store my source files on the SSD and export the renders to the 2tb Hdd and see if it makes any difference.

I also have an mSATA slot as well, which i can install a 32gb mSATA card in for my cache drive, but i really can't be bothered with that, and won't make that much difference anyway.

I will post any changes to my results when i get the chance.
 
Top