WEBVTT

00:00:06.640 --> 00:00:12.440
The Cooler Master glacer 240l CPU Cooler delivers the convenience of an

00:00:10.559 --> 00:00:18.560
all-in-one and the performance of a custom water cooler click now to learn

00:00:14.679 --> 00:00:21.160
more latency or lag is the delay between

00:00:18.560 --> 00:00:25.119
an action and the resulting reaction such as in gaming the pressing of a

00:00:23.439 --> 00:00:29.439
button on a controller and the corresponding animation of a character

00:00:26.840 --> 00:00:34.320
on the screen this latency is an enemy of game it makes controls feel less

00:00:32.320 --> 00:00:39.640
responsive and makes it more difficult to among other things time jumps and

00:00:36.760 --> 00:00:43.719
land head shot there are many smaller contributors that make up the total

00:00:41.600 --> 00:00:47.320
latency in the example I just gave ranging from ones small enough to not

00:00:45.800 --> 00:00:51.239
matter such as the delay between pressing a button on your wired keyboard

00:00:49.280 --> 00:00:56.520
and that action being registered by your computer all the way up to image

00:00:53.239 --> 00:00:59.079
processing delays as high as 68

00:00:56.520 --> 00:01:04.479
milliseconds or 3 to four frames of action on a 60 HZ TV today's

00:01:02.160 --> 00:01:08.880
investigation is not a detailed look into the contribution of every single

00:01:06.439 --> 00:01:13.799
one of these leg factors but rather a look at the current state of local game

00:01:11.439 --> 00:01:18.759
streaming technology and whether it's a feasible alternative to plugging

00:01:15.439 --> 00:01:21.320
directly into a PC or game console now

00:01:18.759 --> 00:01:26.079
when I say local game streaming I am not referring to subscription-based cloud

00:01:23.400 --> 00:01:30.280
services like on live that's internet game streaming and has a whole different

00:01:28.320 --> 00:01:34.320
set of challenges because it relies on your internet connection that means

00:01:32.399 --> 00:01:38.360
latency is higher and beyond your control and it also means that like

00:01:36.560 --> 00:01:43.240
Netflix it will use your monthly bandwidth quota if your internet service

00:01:40.439 --> 00:01:49.079
provider has one local game streaming by contrast uses only your own networking

00:01:46.520 --> 00:01:53.040
equipment which can be as cheap or as fancy as you want does not require

00:01:51.280 --> 00:01:57.680
subscription since your games belong to you already and does not use your

00:01:55.799 --> 00:02:04.280
internet connection because both the game serving PC and the client PC or

00:02:01.079 --> 00:02:06.479
device are within your own network so

00:02:04.280 --> 00:02:11.160
here's our setup our streaming computer for all the tests is using a core i7

00:02:08.440 --> 00:02:15.720
3770k CPU with 8 gigs of RAM and a GTX 780ti it's a very high-end machine but

00:02:14.040 --> 00:02:19.760
in order to reduce the variables in our test it was important to use a gaming

00:02:17.840 --> 00:02:23.400
computer that is easily capable of powering these inhome streaming

00:02:21.360 --> 00:02:28.480
solutions for networking we're using an ASUS rtn 66u which is the best consumer

00:02:26.959 --> 00:02:33.480
grade router that I've personally used for game stream and of course uh it has

00:02:31.720 --> 00:02:38.200
gigabit connectivity for our wired examples for our clients we're using a

00:02:35.599 --> 00:02:43.519
Steam OS machine as well as an Invidia Shield to show off those two different

00:02:40.640 --> 00:02:50.360
streaming Solutions so to get started we're going to plug both a CRT Monitor

00:02:46.040 --> 00:02:52.959
and an ASUS pb278q into a PC in cloned

00:02:50.360 --> 00:02:56.760
display mode for a local gaming session we're using Counterstrike sources in

00:02:54.760 --> 00:03:01.440
game Benchmark to simulate our local gaming session because it's really quick

00:02:58.800 --> 00:03:06.440
to launch and it has an explosion that changes the color of the whole frame

00:03:03.599 --> 00:03:13.360
very suddenly and predictably we were able to consistently measure a 10 to 11

00:03:10.200 --> 00:03:15.200
frame difference when filming at 480

00:03:13.360 --> 00:03:20.200
frames per second between the first bright flash of fire on the CRT and the

00:03:17.920 --> 00:03:26.920
first appearance of it on the LCD that gives us about a 22 millisecond delay

00:03:23.959 --> 00:03:30.280
this methodology unfortunately doesn't give us the exact input leg of the

00:03:29.400 --> 00:03:35.760
monitor but since most people won't notice the

00:03:33.000 --> 00:03:40.120
difference between a CRT and an LCD it does allow us to make relative

00:03:37.640 --> 00:03:46.120
comparisons between the contenders while giving us some context for the following

00:03:42.879 --> 00:03:48.560
numbers first up we're going to test

00:03:46.120 --> 00:03:52.640
steam's inhome streaming you can find more details about inhome streaming in

00:03:50.319 --> 00:03:57.560
the link in the video description and to do that we're going to take our LCD and

00:03:55.120 --> 00:04:01.760
run it off of our steam machine a normal PC that is connected via a wired Gab

00:03:59.840 --> 00:04:07.159
Ethernet connection to the same network as our gaming PC the gaming PC keeps the

00:04:04.599 --> 00:04:12.959
CRT so that we can measure the additional delay that is caused by the

00:04:10.439 --> 00:04:18.799
h264 video encoding being done by the server PC the network transfer and then

00:04:16.000 --> 00:04:24.560
the decoding being done by the client PC it looks like it adds about 74

00:04:21.639 --> 00:04:29.720
milliseconds compared to a CRT monitor which is about equivalent to seeing your

00:04:26.720 --> 00:04:32.320
game with a 4 and 1/2 frame delay on a

00:04:29.720 --> 00:04:37.400
60 HZ monitor not bad when you consider that there are TVs that can add as much

00:04:34.160 --> 00:04:39.919
as 68 milliseconds or about 3 to 4

00:04:37.400 --> 00:04:45.199
frames but you also need to remember guys a leggy display will further

00:04:42.520 --> 00:04:49.520
increase the perceived latency so while it's playable in this situation there's

00:04:47.880 --> 00:04:52.680
still a lot of room for improvement and one to two frames would be ideal just in

00:04:51.440 --> 00:04:58.919
case you hook it up to one of those extremely laggy TVs in scenario 2 we're

00:04:56.440 --> 00:05:02.360
going to use NVIDIA's proprietary gam stream technology

00:05:00.240 --> 00:05:08.000
with a shield handheld Android gaming system to see how close it comes to

00:05:04.919 --> 00:05:10.639
valve's more open solution NVIDIA's

00:05:08.000 --> 00:05:14.880
Advantage is that due to its closed ecosystem they can keep tight control

00:05:12.919 --> 00:05:19.319
over the data chain and fully leverage their own Technologies such as the

00:05:16.440 --> 00:05:23.120
built-in Hardware h264 video encoders and decoders on their Graphics

00:05:20.759 --> 00:05:28.039
processors however NVIDIA's disadvantage here is that they're running wirelessly

00:05:25.360 --> 00:05:32.080
that's how they advertise game stream uh NVIDIA also has the benefit of tightly

00:05:30.520 --> 00:05:39.360
controlling the settings at which games can run and they allow only a maximum of

00:05:34.720 --> 00:05:41.039
720p 60 FPS compared to the 1080p 60fps

00:05:39.360 --> 00:05:47.520
that we were able to run with a wired connection on a steam inhome streaming

00:05:43.880 --> 00:05:50.000
setup so gam stream ads about well about

00:05:47.520 --> 00:05:55.280
the same 74 milliseconds compared to our CRT which seems pretty good considering

00:05:52.960 --> 00:06:00.800
that it's wireless because it's wireless though it is subject to interference and

00:05:57.520 --> 00:06:03.199
you won't get that 74 Mill second delay

00:06:00.800 --> 00:06:08.120
consistently since you won't always be right next to the router the way that we

00:06:05.160 --> 00:06:13.280
are for our controlled testing scenario 3 addresses an imbalance in the previous

00:06:10.840 --> 00:06:19.199
two examples by taking our steam machine and turning it down to 720p 60fps than

00:06:16.639 --> 00:06:23.360
running it wirelessly so it's more equivalent to the shield the ASUS 87i

00:06:21.680 --> 00:06:28.440
Deluxe motherboard that's running in here has onboard dualband AC Wi-Fi which

00:06:26.120 --> 00:06:33.319
should be basically equivalent to the Dual Band n Wii on the shield because

00:06:30.520 --> 00:06:36.880
we're connecting to an N access point going Wireless with steam in home

00:06:34.800 --> 00:06:42.479
streaming adds about there's that number again 74 milliseconds compared to our

00:06:39.479 --> 00:06:45.319
CRT so performance was pretty much the

00:06:42.479 --> 00:06:51.560
same as the shield in terms of lagginess and very playable but not something that

00:06:49.039 --> 00:06:56.680
I would actually recommend doing Shield runs on a small built-in display we only

00:06:54.400 --> 00:07:00.560
used the HDMI output to go to our bigger one to control the variable of the

00:06:58.759 --> 00:07:05.840
displays input legs so that we got comparable results so compressed 720p

00:07:03.720 --> 00:07:09.560
images will normally still look really good and really sharp on that tiny

00:07:07.280 --> 00:07:13.440
little display but for something bigger like you'd be running on a steam machine

00:07:11.039 --> 00:07:18.960
like a TV or a monitor I would really recommend a wired connection so you can

00:07:15.680 --> 00:07:21.039
game at 1080p scenario 4 is an

00:07:18.960 --> 00:07:25.360
open-source streaming app called Limelight that keeps getting brought up

00:07:23.280 --> 00:07:30.360
to me whenever I talk about inhome streaming

00:07:26.400 --> 00:07:32.680
Solutions it works yes but I hope this

00:07:30.360 --> 00:07:37.800
video serves as a demonstration of why I don't talk about it much yet and why I

00:07:35.000 --> 00:07:41.919
don't care about it much yet it works on pretty much any reasonably powerful

00:07:39.599 --> 00:07:46.840
Android device which is really cool but it's just not even in the same ballpark

00:07:44.599 --> 00:07:51.680
as the others and in my experience does not provide a playable experience on

00:07:49.479 --> 00:07:58.000
either of the supported devices I own an NVIDIA shield and an HTC1 our measured

00:07:54.560 --> 00:08:00.000
latency was about 98 milliseconds but

00:07:58.000 --> 00:08:04.319
unfortunately the image quality was lower than any of the other

00:08:01.960 --> 00:08:10.720
Solutions with more frequent unplayable stutters in my testing so in summary the

00:08:08.280 --> 00:08:15.479
best streaming Solutions we have are tripling the delay that you experienced

00:08:12.879 --> 00:08:21.280
when you upgraded from a CRT to a good quality LCD but in that time they're

00:08:19.199 --> 00:08:27.879
encoding an image sending it over a network then decoding it 74 milliseconds

00:08:25.120 --> 00:08:32.320
of display lag is noticeable but it's definitely playable for most genres of

00:08:30.240 --> 00:08:37.320
games I actually beat the first couple worlds of Super Meat Boy streaming on my

00:08:34.479 --> 00:08:42.599
shield I'll be it not as easily as I can when hooked up directly with a wired

00:08:39.320 --> 00:08:45.120
controller but I was able to do it 100

00:08:42.599 --> 00:08:49.800
milliseconds becomes too much to deliver a satisfactory game experience for

00:08:47.320 --> 00:08:55.519
anything but slow paced perhaps turn-based gaming for me and while your

00:08:52.680 --> 00:09:00.560
sensitivity may vary the takeaway today is that gaming on the machine you're

00:08:57.880 --> 00:09:06.880
directly attached to is still better we're talking 22 milliseconds compared

00:09:03.040 --> 00:09:09.800
to about 75 so we're talking about a 3X

00:09:06.880 --> 00:09:15.120
betterness Improvement even with an LCD monitor compared to a CRT versus

00:09:12.120 --> 00:09:17.160
streaming it's still the best but

00:09:15.120 --> 00:09:21.839
quality local game streaming Solutions are adding about another 50 milliseconds

00:09:19.040 --> 00:09:25.720
of leg which means that we can achieve very playable results today if we

00:09:24.000 --> 00:09:30.279
control the additional latency introduced by other parts of the chain

00:09:27.640 --> 00:09:35.920
such as wireless gaming controllers or leggy TVs and in a generation or two if

00:09:33.839 --> 00:09:40.760
encoder and decoder Hardware can be improved we could be able to get pretty

00:09:38.519 --> 00:09:45.920
close to the performance of sitting in front of the device which is extremely

00:09:43.200 --> 00:09:50.720
exciting the future is going to be awesome folks I hope you enjoyed this

00:09:47.959 --> 00:09:55.240
video like it if you liked it dislike it if you disliked it leave a comment and

00:09:52.959 --> 00:09:59.160
let me know what's your sensitivity like did you really notice when you went from

00:09:56.880 --> 00:10:03.959
CRT to LCD do you notice the difference from one LCD to another cuz I certainly

00:10:01.519 --> 00:10:07.760
do and have you tried in home streaming or inidia shield with its gam stream

00:10:05.880 --> 00:10:12.880
technology and did you find it very playable would love to hear from you

00:10:09.399 --> 00:10:14.240
guys and uh I'm sure I have a standard

00:10:12.880 --> 00:10:29.179
intro but I've completely forgotten it now don't forget to subscribe peace

00:10:30.720 --> 00:10:41.010
n