WEBVTT

00:00:00.080 --> 00:00:06.480
It used to be that two monitors was only for business users. [music]

00:00:03.840 --> 00:00:11.679
But these days, whether you're a worker, a gamer, a streamer, multiple monitors

00:00:09.280 --> 00:00:16.400
have gone from being a luxury to being almost essential. I mean, you know who

00:00:14.400 --> 00:00:20.960
you are with [music] the stream playing during your Zoom meeting or the latest

00:00:18.720 --> 00:00:25.439
Mr. B shenanigans next to that boring tutorial that you're really watching.

00:00:23.600 --> 00:00:29.530
But what if I told you that there's a dark side to your harmless little habit?

00:00:27.510 --> 00:00:34.000
[laughter] That's right. Your extra monitors are

00:00:31.679 --> 00:00:38.559
costing you more than just desk space and power consumption. I am talking

00:00:36.480 --> 00:00:44.320
about performance. How much? Well, the general consensus is

00:00:41.600 --> 00:00:48.320
that that's not really enough to matter. But considering the lengths that gamers

00:00:46.239 --> 00:00:53.760
will go for another three frames per second, I disagree. So, we set the Labs

00:00:51.280 --> 00:01:00.079
team loose to settle this debate once and for all. [music] Is this why you're

00:00:56.879 --> 00:01:03.039
stuck at Silver 2? Is this why you can't

00:01:00.079 --> 00:01:07.600
get sponsored by Origin PC? This August, get up to $1,000 off Origin PC's 30

00:01:05.439 --> 00:01:11.520
series laptops, $200 off their desktops, and 100 bucks off their 40 series

00:01:09.520 --> 00:01:15.040
laptops. Learn more using the link below. That's a lot of discount. To

00:01:13.280 --> 00:01:20.560
cover all our bases, the Labs team needed a lot of gear. A high-end test

00:01:18.000 --> 00:01:27.200
bench, a variety of popular games, and four displays that we then ran at either

00:01:23.520 --> 00:01:29.759
1080p or 4K. They then set out to ensure

00:01:27.200 --> 00:01:35.200
that the physical displays were the only variable, which is how we ended up with

00:01:32.240 --> 00:01:40.320
a pretty unusual starting config. Three Edge browser Windows, each running the

00:01:37.680 --> 00:01:45.119
same 4K YouTube video in the background with our game out front. We recorded

00:01:42.960 --> 00:01:50.000
performance, shut down the system, plugged in a second monitor, dragged one

00:01:48.079 --> 00:01:55.119
of our background videos over to the secondary monitor, and reran the test.

00:01:52.720 --> 00:02:00.799
Then for you mega taskers out there, we did the same thing but with four

00:01:57.280 --> 00:02:04.719
displays. And right out of the gate, the

00:02:00.799 --> 00:02:07.920
results were pretty shocking. At 1080p,

00:02:04.719 --> 00:02:11.920
even with all four monitors in use, this

00:02:07.920 --> 00:02:15.599
only cost us 1% performance in Red Dead

00:02:11.920 --> 00:02:17.680
Redemption 2, which is fantastic. Case

00:02:15.599 --> 00:02:21.200
closed. Uh you should keep all your monitors. Maybe buy a few more. We're

00:02:19.920 --> 00:02:25.760
going to have some nice ones linked in the video description. But wait a

00:02:23.120 --> 00:02:31.840
second, that's a lot of bar left down there. Ah, okay. That's cuz the story is

00:02:29.040 --> 00:02:37.760
far from over. In that same scenario, we lost nearly 7% of our performance in

00:02:34.640 --> 00:02:40.879
Total Warhammer 3 with Cyberpunk 2077

00:02:37.760 --> 00:02:43.360
landing in the middle at just over 3%.

00:02:40.879 --> 00:02:48.560
Obviously, these aren't earthshattering numbers, and the impact is lower if you

00:02:46.239 --> 00:02:53.920
only run two displays rather than four. But I was still surprised by how

00:02:51.120 --> 00:02:58.640
significant the performance impact was. And this is on a top-of-the-line

00:02:55.599 --> 00:03:01.840
machine. And it gets worse. I mean,

00:02:58.640 --> 00:03:04.640
given that 4K is literally four times

00:03:01.840 --> 00:03:08.640
the pixel count of a 1080p display, it probably won't surprise you much that

00:03:06.239 --> 00:03:13.882
our performance loss was even more substantial. Total War Warhammer 3

00:03:11.040 --> 00:03:20.000
actually improved going down to 5%. [music] But Red Dead and Cyberpunk shot

00:03:16.720 --> 00:03:22.480
up to around 7% if we had video playback

00:03:20.000 --> 00:03:26.800
running on our three extra screens. The real story though is that this doesn't

00:03:24.800 --> 00:03:32.319
just apply to your average [music] frame rates where realistically in most of

00:03:29.519 --> 00:03:40.239
these games you can probably spare a few FPS. It also applies to your 1% lows,

00:03:36.560 --> 00:03:42.640
the situations where 8 to 9% difference

00:03:40.239 --> 00:03:47.360
could be noticeable as hitching or additional stutter in your gameplay.

00:03:44.667 --> 00:03:51.840
[music] Now, in fairness, this is an extreme example, and most people aren't

00:03:49.760 --> 00:03:56.640
going to be running three full screen video instances while they game. But on

00:03:55.040 --> 00:04:02.080
the other hand, I don't think it's entirely unrealistic either. A streamer,

00:03:59.840 --> 00:04:06.799
for example, could easily have their game running here, their OBS preview

00:04:04.560 --> 00:04:11.280
over here, their live Twitch feed over here, and maybe a let's play or a

00:04:09.120 --> 00:04:16.107
walkthrough running on a fourth display. And the lower the performance of your

00:04:13.519 --> 00:04:22.000
PC, the more you're going to feel [music] this difference. I mean, even a

00:04:18.400 --> 00:04:27.440
pro might struggle to notice 200 FPS

00:04:22.000 --> 00:04:29.360
versus 220 FPS. But 30 versus 33 when

00:04:27.440 --> 00:04:33.840
you're feeling those dips, that's something just about anyone could feel,

00:04:31.440 --> 00:04:40.000
even if they're not sure what's off about it, which is all interesting, but

00:04:37.280 --> 00:04:45.360
doesn't answer why this is happening. To find that out, we reran our tests in 4K,

00:04:43.040 --> 00:04:51.600
but this time without the video playback. And you might think, well,

00:04:48.400 --> 00:04:53.600
obviously outputting those 8.3 million

00:04:51.600 --> 00:04:59.600
pixels per display is going to cost us something. But as it turns out, that's

00:04:56.560 --> 00:05:02.000
just not the case. Our game benchmarks

00:04:59.600 --> 00:05:07.520
actually came back either the same as if we were on a single monitor or about 1

00:05:04.800 --> 00:05:13.280
FPS faster, which [music] is weird, but low enough to chalk it up to runto-run

00:05:09.600 --> 00:05:15.199
variance. So, what's the catch then?

00:05:13.280 --> 00:05:22.160
Pushing more pixels means our computer is working harder, doesn't it?

00:05:18.560 --> 00:05:24.160
Yes, but also no. When you have a static

00:05:22.160 --> 00:05:29.120
image that's just sitting on the screen in Windows, it gets stored in system

00:05:26.400 --> 00:05:34.639
memory. When a window gets dragged to a new position, it gets redrawn as needed.

00:05:32.000 --> 00:05:39.680
But if the contents of that screen need to be changing, all that optimization

00:05:37.280 --> 00:05:44.880
goes out the window. Watch this. I'm going to move around the signup sheet

00:05:42.479 --> 00:05:47.919
for our stubby screwdriver launch here. By the way, be the [music] first to

00:05:46.240 --> 00:05:52.000
know. Sign up and don't miss it. Got that linked in the video description. I

00:05:49.840 --> 00:05:59.919
shake it around and we see an immediate spike in 3D GPU usage. But then I stop

00:05:56.639 --> 00:06:01.759
and boop, it goes away. [music] Then if

00:05:59.919 --> 00:06:07.199
I want to kick things up a notch, rendering animations or decoding a video

00:06:04.400 --> 00:06:11.199
will put additional load on the GPU. Oh, like this. the LTT desk pad configurator

00:06:09.840 --> 00:06:16.479
where you can put in the size of your desk and what kind of deskpad. Oh, look

00:06:13.039 --> 00:06:18.479
at that usage. Yikes. Though again, in

00:06:16.479 --> 00:06:22.400
fairness, that's a little unrealistic. So, why don't we just go with something

00:06:20.319 --> 00:06:29.840
that's closer to our test and play back a video where 3D usage not so bad, but

00:06:27.280 --> 00:06:35.520
video decode starts to pop off. You're going to see this depending on the load,

00:06:32.319 --> 00:06:38.240
there will be some usage. It's not that

00:06:35.520 --> 00:06:45.520
much and it's mostly not on the parts of the GPU that your 3D games rely on, but

00:06:41.360 --> 00:06:47.120
it obviously is something. Of course,

00:06:45.520 --> 00:06:53.520
this isn't the first time we've learned this lesson. Years ago, if you wanted to

00:06:49.840 --> 00:06:56.319
stream to TW, excuse me, kick, the best

00:06:53.520 --> 00:07:00.479
option was to use a second computer to capture and encode your stream. But that

00:06:58.479 --> 00:07:06.639
all changed when NVIDIA introduced Envank and NVEC which offload encoding

00:07:03.759 --> 00:07:12.479
and decoding tasks to dedicated hardware on their GPUs. It was a gamecher [music]

00:07:10.560 --> 00:07:18.080
and it has enabled more than just streaming from a single PC. Remote play

00:07:15.199 --> 00:07:22.080
for example relies heavily on Envank and Nv. [music]

00:07:19.199 --> 00:07:27.120
However, their promise that with decoding and encoding offloaded, the

00:07:24.479 --> 00:07:34.000
graphics engine and the CPU are free for other operations wasn't entirely true.

00:07:31.280 --> 00:07:38.160
Envank has always come with a measurable hit to in-game performance. [music]

00:07:35.840 --> 00:07:44.400
And as we've seen here today, this is also true for decoding and for running

00:07:40.720 --> 00:07:46.319
extra displays. Even if it's as small as

00:07:44.400 --> 00:07:51.199
reallocating the power budget that's needed to fire up those other functional

00:07:48.080 --> 00:07:54.319
units on the GPU, it will cost you

00:07:51.199 --> 00:07:57.280
something. However, even as a one

00:07:54.319 --> 00:08:02.080
monitor guy, both at home and at work, I am not going to tell you to throw away

00:07:59.360 --> 00:08:06.639
your secondary display. For all but the most competitive gamers out there, the

00:08:04.160 --> 00:08:10.560
benefits of multimonitor probably far outweigh the drawbacks.

00:08:08.800 --> 00:08:14.560
However, as a one monitor guy, even if you're not

00:08:13.280 --> 00:08:18.240
worried about the performance loss you've seen here today, there are some

00:08:16.560 --> 00:08:22.319
reasons that I do prefer a single screen. There's less to troubleshoot

00:08:20.160 --> 00:08:27.599
when video issues arise. It requires less power and less cable management.

00:08:24.879 --> 00:08:31.919
And most importantly for me, for gaming, I find it more immersive. Not only can

00:08:30.080 --> 00:08:35.519
you dive right in without a second monitor stealing you away through your

00:08:33.599 --> 00:08:40.719
peripheral vision, but as someone who really needs a break when I take time to

00:08:37.440 --> 00:08:42.959
game, I appreciate that I never see

00:08:40.719 --> 00:08:47.360
notifications popping up on that other screen. Of course, none of that makes a

00:08:45.680 --> 00:08:51.200
difference for you if you need to have Twitch chat up or whatever else. [music]

00:08:49.680 --> 00:08:55.200
So, for those of you who are looking for a second monitor, it is true, by the

00:08:53.360 --> 00:08:58.320
way, we do have some displays linked down below. And I want to especially

00:08:56.720 --> 00:09:02.080
shout out Sony for providing the displays that the lab used for our

00:08:59.760 --> 00:09:05.279
testing. It's a super nice one. We did a full ShortCircuit video and we're going

00:09:03.440 --> 00:09:09.120
to have that one linked down below along with some other affordable and more

00:09:07.586 --> 00:09:12.480
[music] premium options. If you enjoyed this video, make sure to check out the

00:09:10.800 --> 00:09:18.000
one where we explored motherboard performance with 10 different

00:09:14.320 --> 00:09:21.680
motherboards and one CPU. Once again, I

00:09:18.000 --> 00:09:21.680
think the results [music] might surprise