WEBVTT

00:00:00.040 --> 00:00:07.600
so this video got its inspiration from a demo that I saw at the Intel booth at

00:00:04.839 --> 00:00:14.599
Pax they were editing 4K video footage from a Panasonic GH4 camera with truly

00:00:11.320 --> 00:00:16.240
astonishing performance on an 8 Core

00:00:14.599 --> 00:00:21.359
Extreme Edition processor with quad Channel ddr4 memory Etc so a setup

00:00:18.160 --> 00:00:23.560
that's not attainable to most people and

00:00:21.359 --> 00:00:27.640
the demo would have been really impressive to me and i' have gone wow

00:00:25.039 --> 00:00:32.040
the power of the8 core if I hadn't recently with a lot of help from edzel

00:00:29.800 --> 00:00:39.200
who also did the science behind this video discovered that by transcoding 4K

00:00:35.680 --> 00:00:41.800
video to gopro's copor video Codec

00:00:39.200 --> 00:00:46.399
thanks to its GPU acceleration we could get that same performance that we saw in

00:00:44.120 --> 00:00:52.120
the demo with no significant loss and quality on our video editing desktops

00:00:49.120 --> 00:00:53.680
here at the office mere six Core Extreme

00:00:52.120 --> 00:01:00.280
editions you can watch the whole video about all of that here but then I was

00:00:56.440 --> 00:01:04.439
thinking to myself this isn't some wimpy

00:01:00.280 --> 00:01:08.479
Channel this is minus

00:01:04.439 --> 00:01:10.880
t so we decided to amp it up a notch we

00:01:08.479 --> 00:01:16.479
approached ASUS to sponsor a deeper investigation and answer this question

00:01:14.240 --> 00:01:24.759
once and for all although admittedly it's one no one has ever asked me can

00:01:19.360 --> 00:01:27.360
you edit 4K video on an ultra portable

00:01:24.759 --> 00:01:35.759
laptop let's find out

00:01:35.759 --> 00:01:41.520
okay so let's start with a look at the machine that we had as Su send us for

00:01:39.360 --> 00:01:46.200
science the Zenbook ux303u there were a couple of key things

00:01:43.840 --> 00:01:51.159
we were looking for one while not necessarily an Ultra Book by Intel's

00:01:48.799 --> 00:01:55.360
definition it had to be ultra portable it wouldn't impress anyone if we pulled

00:01:52.719 --> 00:02:00.159
a full-fledged gaming laptop or portable workstation like a g751 out of a

00:01:57.600 --> 00:02:07.079
backpack and started editing video on it number two we needed a discret GPU that

00:02:04.119 --> 00:02:11.840
is to say a standalone one not quiet and hard to notice one Intel's onboard

00:02:09.360 --> 00:02:16.440
Graphics have made great strides but a dedicated graphics card in this case one

00:02:14.040 --> 00:02:21.760
with Cuda support is going to do more for most GPU accelerated workloads than

00:02:18.760 --> 00:02:23.360
onboard can and we need as much power as

00:02:21.760 --> 00:02:30.239
we can get if we're going to have any hope of editing 4K and finally three we

00:02:27.000 --> 00:02:32.519
needed enough CPU horsepower and memory

00:02:30.239 --> 00:02:39.239
to support video editing at all so we went for a config with an Intel Skylake

00:02:34.720 --> 00:02:42.360
core i7 6500 U with 12 gigs of RAM but

00:02:39.239 --> 00:02:44.920
is that even enough well we started with

00:02:42.360 --> 00:02:50.760
a baseline by taking 4K footage right off our Sony fs5 dumping it into a

00:02:47.560 --> 00:02:53.760
timeline and trying to edit it

00:02:50.760 --> 00:02:55.840
abominable I mean CPU usage was pegged

00:02:53.760 --> 00:03:01.680
at 100% right away to the point where we were measuring performance in seconds

00:02:58.000 --> 00:03:03.840
per frame not frames per second a7s2

00:03:01.680 --> 00:03:09.480
footage didn't Fair much better with better CPU usage but up to a 4 second

00:03:07.159 --> 00:03:15.319
delay when moving the playhead to a new spot on the timeline yuck but that's

00:03:12.680 --> 00:03:22.000
what we expected so we had a desktop machine transcode 4K footage to cfor 4K

00:03:19.400 --> 00:03:29.439
using Adobe Media encoder and copied it over the network to the local dis on the

00:03:24.200 --> 00:03:32.560
notebook and whoa friends CPU and GPU

00:03:29.439 --> 00:03:36.439
use at full 4K resolution in the preview

00:03:32.560 --> 00:03:39.319
window was sitting at 70 and 90% or more

00:03:36.439 --> 00:03:45.360
respectively with choppy but at least consistent playback not bad dialing the

00:03:43.439 --> 00:03:50.400
preview window down to half resolution yields a significant reduction in CPU

00:03:47.760 --> 00:03:55.920
and GPU use with one quarter resolution running it anywhere from 27 to 30 FPS

00:03:53.680 --> 00:04:01.439
with responsive timeline scrubbing and 1/ 18 able to run at 30 FPS solid now

00:03:59.799 --> 00:04:05.879
now it should be noted that frame rate dips were observed across all preview

00:04:04.280 --> 00:04:10.959
resolutions when additional layers were added so the experience is not

00:04:08.159 --> 00:04:17.120
absolutely perfect but when I popped the big question to Ed could you edit 4K

00:04:14.120 --> 00:04:19.919
comfortably on this setup the answer was

00:04:17.120 --> 00:04:24.160
yeah but hold the phone lonus you're ignoring one of the biggest bottlenecks

00:04:21.560 --> 00:04:28.360
in the video editing process the export of the finished file actually I'm not

00:04:26.720 --> 00:04:32.120
ignoring it at all so there's a couple of different ways you can tackle it with

00:04:30.240 --> 00:04:36.880
a watch folder configuration on your desktop you can finish the file export

00:04:34.520 --> 00:04:40.479
it in cfor drop it in there and have that desktop deal with it but what we

00:04:38.720 --> 00:04:46.440
also discovered is that that didn't actually end up much faster if at all

00:04:43.880 --> 00:04:50.840
than just exporting an h264 anyway thanks again to GPU acceleration so you

00:04:48.960 --> 00:04:56.120
can see I'm actually running an export right now this is a two-minute file

00:04:52.639 --> 00:04:58.600
that's going to finish in about 17 to 20

00:04:56.120 --> 00:05:06.120
minutes and with GPU usage pegged at 100% it is using the GPU wo but copying

00:05:04.080 --> 00:05:11.320
all the project files to a portable device to work on them and then copying

00:05:08.479 --> 00:05:16.160
them back or whatever isn't going to appear in an advertising brochure for a

00:05:13.880 --> 00:05:21.080
thin and light anytime soon with wireless connectivity everywhere and the

00:05:18.759 --> 00:05:25.120
cloud people are getting used to the idea of just working off of network data

00:05:23.479 --> 00:05:29.840
and not having to store anything locally on their machines and while you won't be

00:05:27.280 --> 00:05:34.919
editing 4K video over the internet anytime soon I wanted to investigate

00:05:32.240 --> 00:05:40.000
further to see if we could at least edit directly off of the drive of the more

00:05:37.280 --> 00:05:46.240
powerful machine or a storage machine somewhere else using a home wireless

00:05:42.600 --> 00:05:47.680
router so step one was to uh enable

00:05:46.240 --> 00:05:53.840
maximum performance in the power profiles of the machine and step two was

00:05:50.400 --> 00:05:55.759
to plug in a wired Network USB adapter

00:05:53.840 --> 00:06:02.120
so I used the 100 megabit one that was in the box and the experience was subpar

00:05:59.800 --> 00:06:07.280
we're talking five frames per second regardless of the playback resolution

00:06:04.800 --> 00:06:12.520
running off the lowspeed Wi-Fi in our office yeah that was a similar story but

00:06:09.639 --> 00:06:17.639
our xi3 access points are not designed for blistering speed they're for range

00:06:14.759 --> 00:06:23.360
and consistency so we did make some useful observations here though we got

00:06:19.880 --> 00:06:26.440
low CPU usage on our 4K footage and

00:06:23.360 --> 00:06:28.599
higher CPU usage on the 1080p footage

00:06:26.440 --> 00:06:35.599
that was mixed into our timeline indicating ample processing power and a

00:06:32.880 --> 00:06:41.000
bottleneck somewhere else because this is one of the challenges of copor the

00:06:38.160 --> 00:06:45.880
file sizes are actually larger than the footage directly off the camera in many

00:06:42.840 --> 00:06:48.240
cases so we were easily exceeding our

00:06:45.880 --> 00:06:56.039
connection speed with a 100 megabit wired connection and our office Wi-Fi so

00:06:52.280 --> 00:06:59.160
I grabbed a thirdparty gigabit USB to

00:06:56.039 --> 00:07:01.080
Ethernet adapter and boom timeline

00:06:59.160 --> 00:07:05.599
performance performance got responsive and we were back up to Performance that

00:07:03.160 --> 00:07:11.639
was pretty much indistinguishable from editing locally

00:07:07.680 --> 00:07:14.440
progress but that still involves a wire

00:07:11.639 --> 00:07:21.479
that's not really the point could we do it with Wi-Fi so I dusted off a wireless

00:07:18.240 --> 00:07:23.960
AC 1300 megabit access point these

00:07:21.479 --> 00:07:28.319
typically run around 150 to 200 bucks for a good one like the tplink Google

00:07:25.879 --> 00:07:36.039
onhub one that we checked out recently and took another crack at it there we go

00:07:32.240 --> 00:07:38.319
boo freaking y 300 megabit throughput

00:07:36.039 --> 00:07:42.400
while scrubbing footage about the same as on our wired connection and again

00:07:40.800 --> 00:07:48.520
pretty much indistinguishable performance on Wi-Fi compared to editing

00:07:45.879 --> 00:07:51.919
locally and while there are some caveats you're not going to be able to get this

00:07:49.840 --> 00:07:57.520
kind of performance if you're 200 ft away from your access point what we have

00:07:54.680 --> 00:08:02.720
managed to do here is validate that it's possible to edit 4K video on a thin and

00:08:00.680 --> 00:08:09.080
light at your dining room table with a mere dual core core i7 if you pick the

00:08:06.120 --> 00:08:12.639
right configuration namely one with strong Wi-Fi we tried this same

00:08:11.120 --> 00:08:17.960
experiment on the surface book and it was not nearly as successful and a

00:08:15.000 --> 00:08:22.360
dedicated GPU that part's important because even at one quarter resolution

00:08:20.120 --> 00:08:28.280
in our preview window GPU usage on our 940m was sitting at 30 to 40% and we

00:08:26.039 --> 00:08:34.240
were pegging it when we were exporting finished files

00:08:29.960 --> 00:08:37.360
speaking of graphics cards and video

00:08:34.240 --> 00:08:39.800
things what about graphics and other

00:08:37.360 --> 00:08:44.320
cool video things like video blocks video blocks provides affordable premium

00:08:42.760 --> 00:08:49.279
stock video and they've been doing it since 2011 they operate on a

00:08:46.399 --> 00:08:54.000
subscription-based unlimited library and they add new footage to the library

00:08:51.120 --> 00:09:00.320
twice per month they've got over and Brace yourselves here $10 million worth

00:08:57.480 --> 00:09:05.720
of footage After Effects templat and motion backgrounds and everything in

00:09:02.480 --> 00:09:07.959
video Block's unlimited library is 100%

00:09:05.720 --> 00:09:12.560
royaltyfree and yours to use for personal and Commercial projects they

00:09:10.680 --> 00:09:16.480
recently launched a new members only video Marketplace as an added benefit

00:09:14.720 --> 00:09:20.839
for subscribers and they've got clips from contributors around the world that

00:09:18.839 --> 00:09:25.760
are only available to video block subscribers so contributors on that

00:09:22.920 --> 00:09:30.440
Marketplace by the way keep 100% of the sales as commission video blocks takes

00:09:28.160 --> 00:09:35.640
no cut since the marketplace place is members only so they're already

00:09:33.200 --> 00:09:40.800
supported by their subscription Revenue they've already got 1,500 plus artists

00:09:38.279 --> 00:09:43.920
with more than 200,000 new clips in the few months since the marketplace

00:09:42.160 --> 00:09:48.200
launched and if all that sounds pretty darn good you can get access to the

00:09:46.120 --> 00:09:52.399
unlimited library and the marketplace for only

00:09:49.360 --> 00:09:53.360
$99 per year but if you use our link

00:09:52.399 --> 00:09:59.600
which you can find in the video description box below it's down there and if you sign up during the month of

00:09:57.279 --> 00:10:05.560
December clock's a ticking so should do it quickly you will get onee video

00:10:01.399 --> 00:10:09.120
blocks for only $49 a savings of well

00:10:05.560 --> 00:10:10.800
$50 to put that in context one clip of

00:10:09.120 --> 00:10:15.480
stock video footage similar to those found on video blocks can be bought for

00:10:13.800 --> 00:10:19.000
as much as 50 bucks or actually more depending on what you're using it for so

00:10:17.480 --> 00:10:23.560
if you use the service once your subscription pays for itself go check it

00:10:21.079 --> 00:10:27.959
out today all right thanks for watching guys if this video sucked you know what

00:10:25.519 --> 00:10:31.720
to do but if it was awesome get subscribed leave a comment letting me

00:10:29.920 --> 00:10:36.800
know if you thought this was going to work and if we surprised you here hit

00:10:34.160 --> 00:10:41.120
that like button if you liked the video and if you also like the video maybe

00:10:38.800 --> 00:10:44.000
consider supporting us by using our affiliate code to shop at Amazon

00:10:42.600 --> 00:10:47.240
instructions are up there it costs you nothing by buying a cool shirt like this

00:10:45.760 --> 00:10:50.920
one which cost you something but it's a really cool shirt or by giving us a

00:10:49.720 --> 00:10:54.959
direct monthly contribution through our community Forum which gives you a little contributor badge now that we're done

00:10:53.920 --> 00:10:58.040
doing all that stuff you're probably wondering what to watch next so click

00:10:56.920 --> 00:11:04.839
that little button in the top right corner to check out our update on our

00:11:01.200 --> 00:11:09.240
130 terte storage server it's pretty

00:11:04.839 --> 00:11:09.240
cool seriously it's cool