$300 CPU Beats $4000 CPU?? - Cores vs clockspeed for video encoding

Linus Tech Tips ·Linus Tech Tips ·2017-05-06 · 1,489 words · ~7 min read
Floatplane YouTube

Transcript

JSON SRT VTT 127
0:00 okay do you remember that project I was working on where for the better part of
0:03 6 months I built up this badass 36 core
0:08 dual xon server machine to handle our video encoding and transcoding tasks
0:12 over the network here well fast forward almost a year and many many hours spent
0:19 on diagnosis not to mention a kick in the right direction from this post over
0:23 on Puget systems I think I finally figured out why we never got quite the
0:29 performance that I expected is it
0:32 possible then that a $4,000 22 core CPU could be outperformed
0:39 by one that costs only a few hundred bucks for video encoding is it possible
0:45 that I made a mistake nothing to hold on
0:50 to fails a lot if they reading the sign I'm definitely getting their attention
0:54 so does one of the recurring themes of these laptop or bus videos
0:59 become line failure
1:02 montages I I mean aside from those
1:06 ones let's find
1:16 out freshbooks is the super simple invoicing solution that lets you get
1:21 organized save time and get paid faster
1:24 click now at the link in the video description to try it for
1:28 free Okay so to open this video up we need to take a closer than usual look at
1:33 my test bench I wanted to eliminate bottlenecks wherever possible so that
1:38 the CPU is the only factor in my
1:41 performance evaluation so for that reason most of the performance testing
1:46 was done on an Intel 750 series 1.2 TB
1:49 ndme SSD a GTX Titan x 128 gigs of ddr4
1:54 quad Channel memory on an x99 Deluxe 2 motherboard and the CPUs tested are as
1:59 follows Intel's top of the server line
2:03 2699 V4 22 coron they top of the
2:07 high-end desktop line 10 core core i7
2:10 extreme 6950x the 8 core and 6 core 6900k and
2:16 6800k and finally I decided to throw in their Flagship mainstream 6700k quad
2:22 core to give us the most complete picture possible at the end of the day
2:28 as for the video tests I apologize in advance if the codec or encoder
2:33 application that you personally prefer wasn't covered but this was done as much
2:37 to optimize the Linus Media Group workflow as it was for the purposes of
2:40 creating a video so I'm looking at four different scenarios that we encounter
2:45 pretty much daily one transcoding a 4K
2:49 mxf off of our Sony fs5 to 1080p copor
2:53 our mezzanine codec of choice for editing two exporting a finished project
2:59 in this case a green screened episode as fast as possible directly to h264 for
3:04 publication to YouTube Three a quick
3:07 export in cfor how we normally export so that a network media encoder machine
3:12 with a watch folder can transcode it to h264 and automatically upload it to the
3:17 channel and four finally the performance
3:21 of that copor to h264 conversion with
3:24 the 1080p to 4K upsampling that we perform for the reasons we covered more
3:29 thoroughly in this video here so I ran
3:32 every test with and without Cuda acceleration enabled in Adobe Media
3:37 encoder and used a second machine to capture the screen output with CPU and
3:42 GPU usage displayed so I could review it later let's begin then with scenario one
3:49 this is what most people probably expect
3:52 from a multicore CPU in a video encoding
3:55 Benchmark traditionally this is one of the easiest workloads to scale AC
3:59 crossmore course and our CPU usage graph
4:02 indicates that all is working beautifully throwing a GPU into the mix
4:07 levels the playing field somewhat but this won't surprise anyone who knows how
4:11 GPU dependent a video Codec cineform is and how that bastard law of diminishing
4:17 returns Works moving on to exporting a
4:20 project directly from our cfor timeline
4:24 in CPU only mode we see nice scaling
4:27 with more cores but maybe not quite the
4:30 dominance we'd expect from a chip with and yes I know it doesn't quite work
4:34 this way like 60 GHz of theoretical
4:37 total performance this is a hint of
4:40 things to come and Bam throwing a GPU
4:44 into the mix paints a much more extreme picture here the Cuda accelerated code
4:50 path not only reaps very little benefit
4:54 from more than six cores it punishes
4:57 CPUs with lower clock speed in a way that I really didn't expect observed GPU
5:04 usage is much lower than any other
5:07 processor in this test for our $4,000
5:10 chip and the CPU usage we see of about
5:14 25% tells us this is not a heavily
5:17 threaded workload oops all right so let's break that down
5:23 then into the individual steps and find out where our heavy multi thousand
5:28 investment in an Uber Zeon falls apart
5:31 exporting the project from a cfor 1080p timeline to a cineform 1080p file
5:37 theoretically Elsewhere on the network but I'm using my NVMe drive as a target
5:41 for these benchmarks for consistency sake is pretty flat across the board and
5:46 curiously this is true with or without Cuda acceleration enabled in media
5:50 encoder GPU usage is 85% regardless of
5:53 which drop down so this is clearly nearly 100% GPU dependent which leads us
6:00 then to the second step in the process converting from cineform 1080 to h264 4K
6:07 in CP mode only we see a similar Trend
6:10 to our initial injest test more horses
6:14 is better but only to a point then in GPU assisted mode there it is we are
6:20 almost entirely Bound by per core performance with a lowly quad core
6:25 costing one/ tenth as much handily beating our Zeon be
6:31 so then did I horribly misconfigure our video encoding injest stations and
6:35 output server our Zeon basically pointless NVIDIA
6:39 work well if you're looking simply at the graphs I just showed you along with
6:44 these charts of approximate CPU and GPU
6:48 usage in all the different scenarios I tested then it's pretty clear that these
6:53 lower clocked many core chips are being
6:56 underutilized and the money though I fortunately didn't pay for them would be
7:00 better invested almost anywhere else but
7:04 as always the real world isn't really that simple and it's going to come down
7:07 to the needs and workflow of each individual or organization
7:12 virtualization can be used to get damn near 100% scaling out of as many cores
7:17 as you please encoding software like Sor and squeeze can process many files at a
7:23 time and on the subject of different software testing any given codec in any
7:28 given soft software could yield very different results from what you're
7:32 looking at here so there's no way around
7:36 testing just make sure that when you do so for yourself you go in without any
7:41 assumptions about what the right tool for the job will end up being so you can
7:46 avoid pulling a Linus speaking of tools for the job it's
7:52 summer apparently something something boarding Planes Trains driving a car
7:57 leave your worries behind okay I don't know what any of that stuff in my notes is but today's sponsor is tunnel bear
8:03 and if today's lack of online privacy brings out your inner grizzly
8:07 bear ra then you can try tunnel bear
8:11 it's simple and it is free to try at the
8:14 link in the video description it's the easy to use VPN that makes it so you can
8:19 browse privately and enjoy a more open internet without all that hassle
8:23 associated with more complex VPN Solutions any you know port forwarding
8:28 or DNS or any nonsense like that you just click the button and boom you can
8:34 tunnel into up to 20 different countries and it will appear to the websites and
8:39 services that you are using as though you are coming from that country and
8:43 tunnel bear has a top rated privacy policy and does not log your activity so
8:48 try it free with 500 megabytes and no credit card required and if you decide
8:53 you like it and you want to get a year of unlimited data you can save 10% by
8:57 going to tunnel bear.com T Linked In the video description so
9:02 thanks for watching guys if this video sucked you know what to do but if it was awesome get subscribed hit that like
9:06 button or even check out the link to where to buy the stuff we featured at
9:10 Amazon in the video description also linked in the description is our merch
9:14 store which has cool t-shirts just like this one and our community Forum which
9:17 you should totally join now that you're done doing all that stuff you're probably wondering what to watch next so
9:21 check out that little button in the top right to check out our latest video over
9:24 on Channel Super Fun