WEBVTT

00:00:00.080 --> 00:00:07.200
SLI in a nutshell is a way of using more than one graphics card at a time for a

00:00:05.440 --> 00:00:11.120
richer gaming experience with higher frame rates and more in-game eye candy

00:00:09.280 --> 00:00:17.119
turned on you can learn more about it in this video here on Techquickie now

00:00:13.599 --> 00:00:19.400
NVIDIA's website is full of certain

00:00:17.119 --> 00:00:23.240
information about SLI technology which graphics cards support it where to buy

00:00:21.359 --> 00:00:28.560
the other Hardware you need guides on how to set it up but I noticed that

00:00:25.439 --> 00:00:31.519
Beyond one specific 2-way SLI

00:00:28.560 --> 00:00:36.280
configuration at one specific resolution they don't seem to even want to talk

00:00:33.280 --> 00:00:39.600
about the value proposition of adding in

00:00:36.280 --> 00:00:42.480
more cards or how performance scales in

00:00:39.600 --> 00:00:48.039
relation to the cost as you put more and more graphics cards in your gaming PC so

00:00:46.000 --> 00:00:52.960
when a unique opportunity fell into my lap to test two-way three-way and

00:00:50.559 --> 00:01:02.800
four-way SLI with the brand spanking new gtx980 at 4K I jumped on it

00:01:02.800 --> 00:01:10.680
Corsair gaming RGB keyboards feature Precision Cherry MX RGB key switches for

00:01:07.960 --> 00:01:15.200
16.8 million color per key backlighting for virtually unlimited customization

00:01:13.200 --> 00:01:19.280
click now to learn more I'm going to jump right into numbers here because

00:01:17.200 --> 00:01:23.880
they actually only tell part of the story I tested four games with details

00:01:21.840 --> 00:01:30.720
cranked at 4K resolution on a Gigabyte x99 gaming 5 with a core I 749 30k 6

00:01:26.920 --> 00:01:33.200
core at 4 GHz and 16 gigs of 24 100 MHz

00:01:30.720 --> 00:01:37.880
ddr4 memory and then I threw a Corsair ax 1500i power supply on the bench

00:01:36.000 --> 00:01:42.159
because I was concerned about power consumption and overwhelming the

00:01:39.159 --> 00:01:44.000
existing one but more on that later so

00:01:42.159 --> 00:01:49.240
let's start with what should be the poster child for multi-way SLI Tomb

00:01:46.360 --> 00:01:54.520
Raider 2013 if all you do is hang around and play Tomb Raider go buy 4 GTX 980s

00:01:52.520 --> 00:01:59.240
hell buy an extra one to keep on your desk to remind you of how great it is

00:01:56.799 --> 00:02:05.280
you get 30 FPS average with one card then 20 3 to 24 more frames per second

00:02:02.680 --> 00:02:10.039
with each additional card almost the same scaling going from 1 to two 2 to 3

00:02:07.399 --> 00:02:15.760
and 3 to four cards and on top of that the animations are butter smooth with no

00:02:12.680 --> 00:02:17.360
perceptible micro stutter outstanding

00:02:15.760 --> 00:02:21.200
it's just too bad not every game is like this if Tomb Raider is the poster child

00:02:19.400 --> 00:02:27.239
Watch Dogs is the one in the corner with the duns hat scaling from one GPU to two

00:02:24.480 --> 00:02:31.840
seems okay at first glance if we only look at average frame rates but minimum

00:02:29.920 --> 00:02:36.599
frame rates tell a bit more of the story they dropped almost in half and this

00:02:34.720 --> 00:02:41.120
hitching and stuttering was very noticeable in game it actually gets

00:02:38.720 --> 00:02:46.640
worse the more graphics cards you add yet another issue for the PC version of

00:02:43.080 --> 00:02:49.239
Watchdog next up is Far Cry 3 something

00:02:46.640 --> 00:02:54.400
unusual you might notice about this game is that the minimum frame rates are very

00:02:51.840 --> 00:02:58.560
close to the average frame rates in some cases this is because our run for this

00:02:56.959 --> 00:03:02.640
game involves driving around a track running over dudes and getting shot at

00:03:00.400 --> 00:03:06.120
and looks very consistent having minimum and average frame rates close to each

00:03:04.080 --> 00:03:11.519
other makes animations appear very smooth on screen at least with one or

00:03:09.120 --> 00:03:16.000
two gpus installed in the system even though average frame rates scale very

00:03:13.640 --> 00:03:21.760
well with more gpus you can see that minimums don't and this manifests as

00:03:19.159 --> 00:03:27.680
uneven inconsistent movement of objects on screen with three or four graphics

00:03:24.440 --> 00:03:30.120
cards our last game is Crisis 3 just

00:03:27.680 --> 00:03:35.200
like Far Cry 3 it scaled well from 1 to 2 gpus but we see the same issue with

00:03:33.080 --> 00:03:39.640
great average frame rate scaling and poor minimum frame rate scaling that

00:03:37.040 --> 00:03:44.400
translates to uneven lurchy animations in the case of Crisis 3 it was actually

00:03:42.159 --> 00:03:49.599
noticeably worse than Far Cry 3 with three-way and four-way SL so much so

00:03:47.239 --> 00:03:54.480
that as a gamer I would much rather play with two graphics cards at lower detail

00:03:52.239 --> 00:03:58.360
settings than with more graphics cards and game settings cranked speaking of

00:03:56.760 --> 00:04:01.799
in-game settings if you want to see how I was running everything check out the

00:03:59.959 --> 00:04:05.599
download link in the video description GPU boost clock speed was another

00:04:03.680 --> 00:04:11.120
interesting part of the story with two gpus we saw slightly lower turbo speeds

00:04:08.720 --> 00:04:15.319
because we hit our thermal limit sooner on the top card than the bottom one

00:04:13.040 --> 00:04:19.320
because it has less access to Fresh Air surprisingly though we reached the same

00:04:17.199 --> 00:04:22.720
GPU speed even in the other configurations pretty impressive when

00:04:21.120 --> 00:04:27.199
they're all packed in like sardines in a four-way config speaking of impressive

00:04:25.240 --> 00:04:33.240
power consumption I mean I mentioned before I threw A500 watt 80 plus

00:04:30.039 --> 00:04:35.440
titanium PSU in this rig to ensure that

00:04:33.240 --> 00:04:41.080
I wouldn't be limited the fan on it barely even turned on our four-way SLI

00:04:38.160 --> 00:04:46.039
configuration Drew 620 Watts from the wall in our stare at the fire test in

00:04:43.360 --> 00:04:51.320
Far Cry 3 that means that even accounting for another 50% capacity for

00:04:48.400 --> 00:04:54.960
future expansion 1,000 watt power supply would have been plenty for this rig so

00:04:53.320 --> 00:04:58.440
it's back on the bench all right now back up a second line is you run in here

00:04:56.720 --> 00:05:02.800
throw a bunch of graphs in our faces for five minutes and then basically voice

00:05:01.400 --> 00:05:09.639
over telling us to ignore the performance numbers because they don't reflect the gameplay experience what

00:05:06.400 --> 00:05:11.440
gives great question normally when we do

00:05:09.639 --> 00:05:15.479
benchmarks on this channel it's okay to use fraps to take average frame rates

00:05:13.520 --> 00:05:20.720
and compare graphics cards relative performance because in our game

00:05:18.280 --> 00:05:26.800
showdowns and our graphics card reviews we only test single GPU configurations

00:05:24.360 --> 00:05:30.400
with multi-GPU configurations particularly Beyond 2-way as I've been

00:05:28.720 --> 00:05:35.639
discussing in this video different measurements need to be taken

00:05:32.240 --> 00:05:37.240
to evaluate the evenness of the pace

00:05:35.639 --> 00:05:41.400
with which new frames are delivered to the gamer here's an example the video

00:05:39.759 --> 00:05:46.280
you're watching right now is playing at 30 frames per second

00:05:43.520 --> 00:05:51.720
smooth hold on a second what was that we only dropped the frame rate 3% to 29

00:05:49.759 --> 00:05:56.919
frames per second but because we're doing this by dramatically increasing

00:05:54.280 --> 00:06:00.919
the delivery time of just a few frames periodically you're seeing very

00:05:59.000 --> 00:06:05.479
noticeable hitches and stutters aren't you that's what I'm talking about now

00:06:03.720 --> 00:06:09.360
unfortunately we don't have the equipment to objectively measure the

00:06:07.639 --> 00:06:13.479
perceived smoothness of a game like you'll see on some of the hardcore Tech

00:06:10.960 --> 00:06:17.280
sites like PC per and tech report I recommend reading up on those sites by

00:06:15.039 --> 00:06:21.000
the way uh search for frame pacing if you want to learn more about this but

00:06:18.960 --> 00:06:26.120
the good news is that in the case of our multi-way SLI video today at least with

00:06:23.960 --> 00:06:30.319
this new and relatively immature driver for the GTX 980 it was easy to pick up

00:06:28.800 --> 00:06:34.160
on this stuff with the naked eye so being able to measure it wouldn't have

00:06:31.919 --> 00:06:39.759
changed the conclusion much and here it is what we've been saying for years is

00:06:36.520 --> 00:06:42.240
still true one of the fastest GPU you

00:06:39.759 --> 00:06:47.080
can afford is best in most cases because you'll always get the performance you

00:06:44.000 --> 00:06:49.479
paid for even in Watchdogs until you get

00:06:47.080 --> 00:06:53.759
up to the top tier cards at which point two of them might make sense if you

00:06:51.560 --> 00:06:57.639
really are hungry for more performance but beyond that things really come off

00:06:55.599 --> 00:07:02.560
the rails in a hurry and I guess I can see why NVIDIA offers you the capability

00:07:00.400 --> 00:07:06.879
to run more cards if they want to but didn't even have any three-way or

00:07:04.960 --> 00:07:12.319
four-way configs running even in their own demo systems at the gtx980 launch

00:07:10.039 --> 00:07:14.759
event so that pretty much wraps it up guys like this video if you liked it

00:07:13.479 --> 00:07:18.560
dislike it if you disliked it leave a comment letting me know if your feelings are more complicated than this check the

00:07:17.360 --> 00:07:22.599
link in the video description if you want to support us you can give us a monthly contribution bu a cool t-shirt

00:07:21.360 --> 00:07:27.800
oh I'm not wearing one of our t-shirts today well that's awkward buy a cool line is Tech tips t-shirt like oh why

00:07:26.680 --> 00:07:34.479
thank you like this one yeah I'm standing on a

00:07:30.360 --> 00:07:36.319
box he tripped on it and also uh change

00:07:34.479 --> 00:07:39.879
your link to Amazon to one with our affiliate code so we get a small

00:07:37.599 --> 00:07:43.400
Kickback whenever you buy I don't know licorice or whatever it is you buy on

00:07:41.440 --> 00:07:56.799
Amazon thanks for watching guys and as always don't forget to subscribe