WEBVTT

00:00:00.160 --> 00:00:05.760
hmm what's the perfect CPU for this build

00:00:06.080 --> 00:00:12.080
ooh this looks pretty good now i just need a graphics card to pair

00:00:10.400 --> 00:00:14.960
it with right hmm

00:00:13.840 --> 00:00:20.080
well this looks too pinner this is probably going to bottleneck my CPU

00:00:20.560 --> 00:00:27.119
but then this huh this is probably over the top it might

00:00:25.199 --> 00:00:31.359
like destroy it just to be in close proximity to it

00:00:29.519 --> 00:00:37.840
but i mean how do you keep from bottlenecking your

00:00:35.440 --> 00:00:41.760
CPU with your GPU and then do you have to worry about picking the

00:00:39.120 --> 00:00:45.760
right CPU for the GPU can it work both ways

00:00:43.600 --> 00:00:49.840
what does a balanced build even look like these days

00:00:48.399 --> 00:01:00.160
good thing we've got all the tools to find out

00:01:00.160 --> 00:01:07.119
when i say bottleneck what comes to your mind is it super laggy unplayable frame

00:01:05.760 --> 00:01:10.960
rates and games is it waiting five minutes just to load

00:01:09.200 --> 00:01:15.439
Windows is it the neck of a bottle

00:01:13.760 --> 00:01:19.119
whatever your exact experience with bottlenecks is they all have one thing

00:01:17.920 --> 00:01:23.280
in common like a five lane highway merging before

00:01:21.600 --> 00:01:28.080
a single lane tunnel one aspect of your computer system will

00:01:25.600 --> 00:01:32.159
be holding back another one keeping it from reaching its full

00:01:29.840 --> 00:01:35.360
performance potential so

00:01:33.360 --> 00:01:39.600
bottlenecking has become one of the most common fears among new pc builders i

00:01:38.320 --> 00:01:45.360
mean after all no one likes wasting money so are these

00:01:42.560 --> 00:01:49.680
concerns legit or has bottlenecking

00:01:46.799 --> 00:01:54.799
become some sort of boogeyman for system builders to deter new users from

00:01:52.560 --> 00:01:59.280
configuring their own we were actually inspired to make this video by the

00:01:57.719 --> 00:02:05.200
incredible bottlenecking that we experienced when

00:02:01.119 --> 00:02:07.759
we tried to game on this right here a 64

00:02:05.200 --> 00:02:11.599
core xeon phi system but

00:02:08.879 --> 00:02:18.319
that was obviously an unrealistic scenario i mean this CPU is made up of

00:02:14.640 --> 00:02:21.520
64 like crappy atom processors

00:02:18.319 --> 00:02:24.480
so today's investigation will test just

00:02:21.520 --> 00:02:31.040
how badly we can bottleneck a more normal system built from actual desktop

00:02:27.520 --> 00:02:33.920
components so in front of me here is a

00:02:31.040 --> 00:02:42.000
lower end current gen CPU a popular but pretty obsolete CPU from

00:02:37.280 --> 00:02:44.560
2013 and a high-end current generation

00:02:42.000 --> 00:02:50.879
processor we're going to start then with the most basic of our graphics cards a

00:02:47.360 --> 00:02:52.319
gtx 1030. this will help us establish a

00:02:50.879 --> 00:02:57.760
baseline now i'm actually expecting

00:02:54.800 --> 00:03:03.840
similar performance in games going all the way from a mid-tier processor from

00:03:00.879 --> 00:03:08.879
five years ago all the way up to the brand new one

00:03:05.120 --> 00:03:12.319
that would indicate a GPU bottleneck

00:03:08.879 --> 00:03:13.840
where it's holding back our better CPU

00:03:12.319 --> 00:03:18.159
and success i guess

00:03:16.000 --> 00:03:23.360
because we're basically even across the board surprisingly cs go actually runs

00:03:21.599 --> 00:03:28.800
really well on this card averaging over 120 FPS for all of our benches but

00:03:26.560 --> 00:03:35.040
what's less surprising is the way that in a gaming workload an underpowered

00:03:31.920 --> 00:03:37.680
graphics card will perform its best

00:03:35.040 --> 00:03:40.560
pretty much regardless of which CPU it gets paired with so you heard it here

00:03:39.360 --> 00:03:47.120
first folks a processor upgrade will do nothing for

00:03:43.920 --> 00:03:48.480
your gaming experience if your GPU can't

00:03:47.120 --> 00:03:54.239
keep up with that said 3d creation software

00:03:51.680 --> 00:03:58.080
blender reveals a difference between our platforms demonstrating that whether a

00:03:56.720 --> 00:04:03.280
system is bottlenecked or not is heavily dependent

00:04:01.519 --> 00:04:07.760
on the type of work you plan to do on it

00:04:06.720 --> 00:04:16.400
now then let's retest everything with a gtx 1060

00:04:12.959 --> 00:04:19.759
and see how the situation changes

00:04:16.400 --> 00:04:21.040
right away we're greeted with oh good a

00:04:19.759 --> 00:04:27.120
difference so the core i7 pulls way ahead in

00:04:23.840 --> 00:04:30.080
assassin's creed origins and cs go and

00:04:27.120 --> 00:04:35.040
3dmark confirms this data with a much nicer spread between our platforms which

00:04:32.560 --> 00:04:39.440
means then that our more powerful processors are getting to stretch their

00:04:37.360 --> 00:04:43.360
legs a little more because they're not being held back to the same degree but

00:04:42.320 --> 00:04:50.400
wait it appears as though the gtx 1060 is

00:04:47.280 --> 00:04:51.919
still the bottleneck for deus ex and far

00:04:50.400 --> 00:04:55.680
cry 5. this is where our second lesson about

00:04:54.000 --> 00:05:01.759
bottlenecks comes in it can even be dependent on the

00:04:58.639 --> 00:05:05.360
individual piece of software with some

00:05:01.759 --> 00:05:08.479
games favoring faster cpus and others

00:05:05.360 --> 00:05:11.280
needing more GPU horsepower in order to

00:05:08.479 --> 00:05:15.840
look their best coming back around then to our non-gaming tests the results

00:05:13.360 --> 00:05:20.880
confirmed that productivity still likes a fast CPU to keep the GPU fed

00:05:19.199 --> 00:05:25.039
who knew okay then so that's probably quite

00:05:22.880 --> 00:05:30.600
enough of these uh blue collar gpus where the devil is the titan v

00:05:27.600 --> 00:05:30.600
james

00:05:31.360 --> 00:05:38.720
james is running something on the titan v right now so we're going to use this

00:05:35.520 --> 00:05:41.120
titan v box this gtx 1070 and this gold

00:05:38.720 --> 00:05:44.800
sharpie marker to create a reasonable facsimile of a titan v

00:05:46.160 --> 00:05:51.759
here we're expecting some pretty different results and

00:05:50.800 --> 00:05:59.280
yes we have finally broken the GPU

00:05:54.720 --> 00:06:02.320
bottleneck on deus ex and our ryzen 3 is

00:05:59.280 --> 00:06:05.520
definitely beating out the fx 6350 for

00:06:02.320 --> 00:06:08.479
the first time in far cry 5 showing us

00:06:05.520 --> 00:06:12.560
what an optimized title can do unfortunately blender still doesn't run

00:06:10.960 --> 00:06:19.680
on the titan v so we're just left with v-ray which doesn't scale

00:06:15.680 --> 00:06:22.720
bringing us to our last test scenario

00:06:19.680 --> 00:06:26.639
dropping our monitor's resolution

00:06:22.720 --> 00:06:28.880
down to 1280 times 1024

00:06:26.639 --> 00:06:32.720
on that very same titan v

00:06:31.120 --> 00:06:37.440
test platform and oh what's this

00:06:36.319 --> 00:06:43.440
interesting even with a titan v

00:06:40.000 --> 00:06:45.440
we're back to CPU bottlenecking

00:06:43.440 --> 00:06:52.560
ah yes the final lesson is that bottlenecks

00:06:48.960 --> 00:06:54.880
will also appear and disappear depending

00:06:52.560 --> 00:07:00.400
on the settings that you're using in your program

00:06:56.800 --> 00:07:03.599
lower resolution gaming is easier on the

00:07:00.400 --> 00:07:07.280
graphics card but actually doesn't ease

00:07:03.599 --> 00:07:10.400
the CPU's job to nearly the same degree

00:07:07.280 --> 00:07:13.039
and higher resolution gaming is harder

00:07:10.400 --> 00:07:18.960
on the graphics card but since that drives the frame rates lower the CPU

00:07:16.319 --> 00:07:25.280
doesn't end up having to work as hard so if your gaming system was severely

00:07:22.400 --> 00:07:29.599
CPU bottlenecked you actually might be able to upgrade your monitor and play at

00:07:28.000 --> 00:07:34.880
a higher resolution without losing any FPS

00:07:32.720 --> 00:07:41.360
go figure so really the main takeaway here

00:07:37.759 --> 00:07:44.639
is that there is no way

00:07:41.360 --> 00:07:45.919
to avoid bottlenecks all together

00:07:44.639 --> 00:07:50.639
think about it a theoretical system with no bottlenecks

00:07:49.599 --> 00:07:56.080
would have unlimited performance in every workload

00:07:54.319 --> 00:08:02.160
because nothing would be a limiting factor obviously that's ridiculous

00:07:59.440 --> 00:08:07.360
something is always going to hold you back and you can't get

00:08:03.759 --> 00:08:09.840
paralyzed by the bottleneck boogeyman

00:08:07.360 --> 00:08:14.080
as you've seen here as long as you apply some common sense to your config and

00:08:12.560 --> 00:08:19.440
avoid pairing thirty dollar cpus with

00:08:16.639 --> 00:08:24.400
three thousand dollar graphics cards you should be just fine and you can always

00:08:22.160 --> 00:08:29.840
ask for advice like maybe over on the Linus tech tips forum about where the

00:08:27.280 --> 00:08:34.159
bottleneck might be in your system and how severe it might be

00:08:32.640 --> 00:08:39.680
just make sure to provide some information about what you expect to use

00:08:36.640 --> 00:08:41.440
your pc for and what kind of settings

00:08:39.680 --> 00:08:46.959
you expect to run at or as you can see it's going to be pretty

00:08:45.040 --> 00:08:50.560
hard to help you so thanks for watching guys if this

00:08:48.399 --> 00:08:54.000
video sucked you know what to do but if it was awesome get subscribed hit that

00:08:52.720 --> 00:08:58.399
like button or check out the link to where to buy the stuff we featured in the video description also down there is

00:08:57.040 --> 00:09:02.800
our merch store which has cool shirts like this one and our community forum

00:09:00.800 --> 00:09:06.600
which you should definitely join so you can get bottleneck advice