WEBVTT

00:00:00.160 --> 00:00:06.400
does this idling at 220 watts yes jeebus

00:00:03.919 --> 00:00:12.800
now at the end of our pc pro video we told you guys that yeah it's awesome

00:00:09.760 --> 00:00:15.920
but it lacks focus like it's not really

00:00:12.800 --> 00:00:17.840
designed for a single task it's a middle

00:00:15.920 --> 00:00:21.840
of the road performer for everything

00:00:19.199 --> 00:00:28.240
not so with the creature that we are showing off today this is the camino rm

00:00:25.359 --> 00:00:34.640
and it is designed for one thing and one thing only absolutely chewing

00:00:31.519 --> 00:00:38.160
through video rendering it's got four

00:00:34.640 --> 00:00:41.040
rtx 2080 tis a threadripper CPU and a

00:00:38.160 --> 00:00:45.200
design like nothing else i have ever seen before

00:00:43.200 --> 00:00:49.760
just like this segway is like nothing you've ever seen before this video is

00:00:47.120 --> 00:00:53.680
brought to you by glasswire instantly see your current and past network

00:00:51.200 --> 00:00:57.520
activity detect malware and block badly behaving apps with glasswire check it

00:00:55.840 --> 00:01:01.359
out at the link in the video description to get 25

00:00:59.039 --> 00:01:01.359
off

00:01:08.320 --> 00:01:17.920
to be clear i'm not saying this is a slow machine when we put our 32 000 pc

00:01:14.000 --> 00:01:20.320
pro up against a 33 000 mac pro it was

00:01:17.920 --> 00:01:25.759
twice as fast in our mark bench benchmark which is a project

00:01:23.680 --> 00:01:31.920
that we created identically in both premiere pro and final cut to see how

00:01:29.040 --> 00:01:34.320
fast a pc versus a mac could encode a video

00:01:33.119 --> 00:01:40.720
however when we take a more specialized machine

00:01:37.280 --> 00:01:43.119
can we beat the mac pro buy more

00:01:40.720 --> 00:01:46.240
there's only one way to find out here we go time remaining oh wow it's

00:01:44.799 --> 00:01:51.680
actually not a head by that much but it is ahead so far

00:01:49.280 --> 00:01:56.880
man this new GPU encoding in adobe premiere really hits it doesn't it we're

00:01:53.520 --> 00:01:59.520
seeing GPU usage as high as 50 70.

00:01:56.880 --> 00:02:05.680
i saw 80 for a second there and that's all four of the rtx 2080 ti's what's our

00:02:02.399 --> 00:02:07.759
power draw on that oh a cool 760 watts

00:02:05.680 --> 00:02:11.760
really that's all that's it huh yeah it's efficient

00:02:09.599 --> 00:02:16.160
i think a big part of it is just that it's running a threadripper rather than

00:02:13.520 --> 00:02:21.440
an epic because instead of running it three gigahertz and change boost this

00:02:18.959 --> 00:02:26.879
puppy is boosting to four four point four point two five gigahertz it's a 32

00:02:24.000 --> 00:02:32.160
core processor it's nuts oh and it's done oh and it's done

00:02:29.520 --> 00:02:38.800
let's check the times and coding time one minute 45 for the camino

00:02:34.959 --> 00:02:40.080
151 for the pc pro we've got a victory

00:02:38.800 --> 00:02:44.959
now it's still pretty expensive at around fifteen

00:02:42.400 --> 00:02:50.720
thousand dollars but this kind of custom chassis work and the water cooling

00:02:47.200 --> 00:02:54.160
that's inside it does not come cheap and

00:02:50.720 --> 00:02:55.760
it still costs about half as much as the

00:02:54.160 --> 00:03:01.360
pc pro let's go on wow okay i was not expecting

00:02:59.120 --> 00:03:06.239
this uh we only have one license for solidworks visualize but apparently it's

00:03:05.200 --> 00:03:13.040
given us a little bit of leeway on that right now so the purpose of this program is to

00:03:10.720 --> 00:03:18.080
take a project that you've created and render it in like a super cool photo

00:03:15.519 --> 00:03:21.599
realistic way for promotional imagery so we're going to go ahead and take this

00:03:19.519 --> 00:03:27.360
camaro which is included with the program and we are going to output it

00:03:24.800 --> 00:03:30.799
and click right there right so that just brings up the thing and make sure the

00:03:28.879 --> 00:03:36.239
passes yeah make those each a thousand a hundred thousand passes oh my lord

00:03:33.360 --> 00:03:39.200
enable denoiser ouch you ready are we gonna blow breaker alex i don't

00:03:38.080 --> 00:03:44.159
know are they hooked up to different breakers i'm not entirely sure

00:03:41.440 --> 00:03:50.159
oh snap okay well the cpus are going the gpus haven't kicked in we're at holy

00:03:47.440 --> 00:03:54.319
we're at 300 watts how is it even doing that it's not even hitting those gpus or

00:03:52.239 --> 00:03:59.560
so it thinks it must be yeah task manager has to be confused

00:03:56.560 --> 00:03:59.560
whoa

00:04:00.000 --> 00:04:05.599
hey i mean why buy a space heater for your bedroom

00:04:03.439 --> 00:04:10.000
for the winter just pick up one of these puppies

00:04:06.879 --> 00:04:13.200
holy crap this is only halfway through

00:04:10.000 --> 00:04:15.519
its passes and this one is done

00:04:13.200 --> 00:04:22.800
it goes to show you that if your workload just relies on raw GPU muscle

00:04:20.079 --> 00:04:28.160
a quadro is not going to do anything for you that a g-force can to put this in

00:04:25.440 --> 00:04:34.320
perspective we are twice the performance of our pc pro which was 36 times the

00:04:31.840 --> 00:04:37.040
speed of the mac pro and

00:04:35.280 --> 00:04:40.800
we're half the price while we wait around for the pc pro to finish let's

00:04:38.639 --> 00:04:46.320
set it to a hundred thousand passes let's start up the render again and have

00:04:43.120 --> 00:04:48.960
a look at the incredible cooling system

00:04:46.320 --> 00:04:52.560
in the camino rm here when i filmed my initial unboxing of

00:04:50.880 --> 00:04:57.040
this thing on our ShortCircuit channel go subscribe by the way

00:04:54.800 --> 00:05:02.639
all i could really say was what am i even looking at here because

00:04:59.680 --> 00:05:07.360
you got your CPU with this crazy custom block they've designed on it you've got

00:05:04.720 --> 00:05:11.600
a loop whose order kind of makes no sense like it splits out and then half

00:05:09.919 --> 00:05:16.880
of the water goes to the CPU and the other half goes to the four gpus which

00:05:14.720 --> 00:05:19.919
are all in like these little like GPU sandwiches

00:05:18.160 --> 00:05:26.160
under the motherboard and then you've got these incredible shielded pci

00:05:22.560 --> 00:05:27.919
express risers that actually run around

00:05:26.160 --> 00:05:33.360
to the underside of the motherboard tray leaving the top of the system looking

00:05:30.479 --> 00:05:37.440
really clean well as it turns out we've been in touch with them and they've

00:05:34.800 --> 00:05:41.440
explained every funny thing that we noticed about the design and it was all

00:05:39.840 --> 00:05:46.800
with the intention of creating a rendering and ai

00:05:44.000 --> 00:05:51.520
just brood of a machine that is still standalone so it doesn't require like a

00:05:49.600 --> 00:05:56.479
chilled air cooling system in a data center or anything like that and

00:05:53.840 --> 00:06:02.880
reasonably quiet so the reason that they went with these 140

00:05:59.280 --> 00:06:05.120
millimeter Noctua ppc 3000 RPM fans is

00:06:02.880 --> 00:06:10.560
that they've got a great combination of both airflow moving capacity and static

00:06:08.639 --> 00:06:15.360
pressure so they've got this seal here on the back so that the triple 120

00:06:12.880 --> 00:06:18.960
millimeter radiator is able to get all of the air moving through it instead of

00:06:17.120 --> 00:06:26.479
spilling out around it the reasoning behind the triple sfx 750 watt power

00:06:22.880 --> 00:06:28.400
supplies basically came down to avoiding

00:06:26.479 --> 00:06:32.080
atx power supplies because they would have interfered with them being able to

00:06:29.840 --> 00:06:37.600
fit the triple radiator and also avoiding noisy redundant server style

00:06:35.039 --> 00:06:40.160
power supplies which are well noisy

00:06:41.759 --> 00:06:49.919
whoa hey okay now we ran this benchmark in a room

00:06:47.840 --> 00:06:55.199
temperature room so about 22 degrees celsius 23 degrees celsius for an hour

00:06:53.280 --> 00:07:00.160
and a half to give the system all the time it needed to reach equilibrium and

00:06:57.120 --> 00:07:03.599
they managed 77 degrees at the highest

00:07:00.160 --> 00:07:07.360
for our CPU and a maximum of just 66

00:07:03.599 --> 00:07:09.840
degrees for our hottest GPU so as crazy

00:07:07.360 --> 00:07:16.800
as it sounds initially to split up your coolant flow

00:07:11.360 --> 00:07:19.120
uh half to the CPU and half to four gpus

00:07:16.800 --> 00:07:24.479
it looks like that was the right move the benefits of a water cooling system

00:07:21.039 --> 00:07:27.039
like this are two-fold one is that each

00:07:24.479 --> 00:07:31.199
of these four gpus it's crazy it only looks like two gpus doesn't it because

00:07:29.360 --> 00:07:37.120
they're like they're sandwiched together each of them was turboing to 1800

00:07:34.319 --> 00:07:41.759
megahertz during that long intensive test and

00:07:38.960 --> 00:07:46.240
according to kamino anyway they can be quickly and easily swapped out for

00:07:44.479 --> 00:07:49.440
maintenance without even draining the water cooling loop which i'm kind of

00:07:47.759 --> 00:07:53.599
looking at going um really

00:07:51.440 --> 00:07:57.520
cause what am i even looking at here uh the gpus are all kind of on a tray that

00:07:55.759 --> 00:08:02.879
comes out oh there's the security torx there okay well i'm a little worried that nope

00:08:01.440 --> 00:08:06.240
that's gonna happen come on baby

00:08:04.560 --> 00:08:10.160
come on you can do it don't be shy oh there we go what's it

00:08:08.240 --> 00:08:13.120
stuck on you know maybe you just need to just force it

00:08:13.440 --> 00:08:19.599
okay i see their point but i'm also not a

00:08:17.599 --> 00:08:24.840
hundred percent sure that i agree with it yes you could strip the cooler off of

00:08:21.840 --> 00:08:29.360
another 2080 ti take that bare

00:08:24.840 --> 00:08:32.080
pcb slop it on there and uh in practice

00:08:29.360 --> 00:08:37.200
i think that opportunities to change out graphics cards without draining the loop

00:08:33.839 --> 00:08:38.399
would be few and far between because you

00:08:37.200 --> 00:08:42.959
know if you wanted to upgrade for example you would obviously need to replace the block um do you do should we

00:08:41.680 --> 00:08:46.959
pull a card off and just have a look at the block i'm sure can i do it

00:08:45.279 --> 00:08:51.600
right now it's held on by only thermal compound

00:08:48.640 --> 00:08:51.600
which seems fine

00:08:52.240 --> 00:08:56.000
fine alex just chill

00:08:56.160 --> 00:09:05.920
oh interesting that is a very unusual approach normally

00:09:03.720 --> 00:09:09.920
manufacturers don't try to get the tolerances so tight on their copper

00:09:08.240 --> 00:09:14.560
blocks and they use a pad for memory instead of

00:09:12.160 --> 00:09:17.920
thermal compound and actually you can see the wisdom of that so there's some

00:09:16.399 --> 00:09:22.160
there's some learnings in here for camino you can see that the contact

00:09:20.160 --> 00:09:25.680
between these memory chips and the block is not actually amazing one of the

00:09:24.000 --> 00:09:31.360
reasons for that is that when you're mounting uh surface mount bga components

00:09:28.399 --> 00:09:34.399
like these gpus the tolerances can be a little bit off and it's far more

00:09:32.959 --> 00:09:38.480
important you have good contact on the GPU because these memory chips don't

00:09:36.480 --> 00:09:42.720
actually output that much heat you can see their GPU contact is excellent so

00:09:41.200 --> 00:09:48.320
what they'd probably be better off with in a future revision of this block is

00:09:45.200 --> 00:09:50.399
bringing this part down just fraction of

00:09:48.320 --> 00:09:54.000
a millimeter and then using pads for these just like they do with the voltage

00:09:52.240 --> 00:09:59.200
regulators and mosfets and all that kind of stuff we can take a moment now to

00:09:56.160 --> 00:10:02.240
appreciate camino's pci express daughter

00:09:59.200 --> 00:10:04.720
board riser card thing so what's really

00:10:02.240 --> 00:10:08.959
neat about these is they allow them with just these skinny

00:10:06.399 --> 00:10:15.360
uh rounded connectors so instead of the like big flat pci express extenders to

00:10:12.320 --> 00:10:17.040
take a full 16x slot worth of bandwidth

00:10:15.360 --> 00:10:21.760
around the motherboard around to the back of the motherboard tray these right

00:10:19.040 --> 00:10:25.279
here each handle 8x worth of bandwidth so if you hook up just one of them you

00:10:23.440 --> 00:10:29.279
would have an 8x slot and if you hook up the second one you've got the full 16x

00:10:27.680 --> 00:10:33.279
bandwidth to your card which might not be that important for gaming but for

00:10:31.440 --> 00:10:36.000
compute applications which is exactly what they're doing here that's

00:10:34.480 --> 00:10:40.240
absolutely critical you want as much bandwidth to the slot as you can get i

00:10:37.920 --> 00:10:45.519
figured it out alex there's angled screws down here angled screws angled

00:10:42.959 --> 00:10:51.680
they're hiding ah oh lordy

00:10:47.440 --> 00:10:53.519
oh whoa hey hi ho hey how's it going

00:10:51.680 --> 00:10:57.440
is how the power supplies come out while i'm not sure i quite buy it on the

00:10:55.279 --> 00:11:00.880
graphics cards power supply swaps do appear to be

00:10:58.720 --> 00:11:06.079
pretty much as simple as pippity poppity modular interface screw

00:11:03.360 --> 00:11:09.920
in a new one and you're ready to rock we are back up and running so we're going

00:11:07.279 --> 00:11:15.120
to do our last benchmark against our 32 000 pc this is octane bench this is a

00:11:12.959 --> 00:11:18.959
GPU rendering benchmark let's go ahead and have a look at our

00:11:17.600 --> 00:11:23.360
usage here okay they're both fast

00:11:21.120 --> 00:11:30.560
so that's nice and a big surprise to absolutely no one

00:11:27.200 --> 00:11:33.680
the camino rm crushinated the pc pro

00:11:30.560 --> 00:11:35.760
with that said as we said about the pc

00:11:33.680 --> 00:11:40.079
pro in our original video it's an all-rounder so if we were to run into a

00:11:38.240 --> 00:11:45.839
situation where we were working on a project or a data set that was so big

00:11:43.279 --> 00:11:50.000
that we need more than 11 gigabytes of video memory

00:11:47.839 --> 00:11:54.560
the pc pro would absolutely pull out ahead so this is more of a lesson in

00:11:52.720 --> 00:12:00.800
buying the right tool for the job than it is necessarily just a sales pitch for

00:11:57.680 --> 00:12:02.560
the camino rm it's a super cool machine

00:12:00.800 --> 00:12:06.320
but unless you absolutely know you need one of these i probably wouldn't call

00:12:04.320 --> 00:12:11.519
them up and uh you know send them a purchase order that said i would love to

00:12:08.959 --> 00:12:15.839
use this as our video rendering server it's freaking fast just like the speed

00:12:14.480 --> 00:12:21.040
at which i'm going to transition to our sponsor blinkist takes the best insights and

00:12:19.519 --> 00:12:24.399
most important information from thousands of non-fiction books and

00:12:22.959 --> 00:12:29.200
condenses them into 15 minutes you can either read or listen

00:12:27.440 --> 00:12:33.600
from the app and you can learn about self-development and success business

00:12:31.279 --> 00:12:38.399
and finance parenting psychology self-help all kinds of amazing topics

00:12:36.560 --> 00:12:42.240
and it'll help you discover different perspectives and views to incorporate

00:12:40.480 --> 00:12:46.880
into your everyday life you would be amazed how much you can learn in 15

00:12:44.959 --> 00:12:50.480
minutes they've got offline downloads which makes it perfect for commutes and

00:12:48.959 --> 00:12:54.160
some of their most popular titles include what to do when machines do

00:12:52.399 --> 00:12:58.079
everything which explains how automation can be incorporated into current

00:12:55.519 --> 00:13:02.399
business models and streaming sharing stealing which outlines successful

00:13:00.160 --> 00:13:05.920
companies like apple netflix and amazon on how they use data to understand their

00:13:04.399 --> 00:13:09.519
consumers in the ever-changing entertainment industry it's available on

00:13:07.680 --> 00:13:13.760
the google play store and apple's app store and the first 100 people to go to

00:13:12.000 --> 00:13:16.880
blinkist.com Linus tech tips we're going to have the

00:13:15.040 --> 00:13:21.120
link below are going to get unlimited access for one week to try it out also

00:13:19.680 --> 00:13:25.200
they have full length audio books now with premium subscribers getting special

00:13:22.800 --> 00:13:29.519
pricing up to 65 off retail price so go check them out down below if you guys

00:13:27.519 --> 00:13:32.800
enjoyed this video maybe check out the ShortCircuit where we did our initial

00:13:31.279 --> 00:13:37.600
unboxing and impressions of it actually we kind of covered everything that we said in that one but in more depth today

00:13:35.760 --> 00:13:43.519
so maybe go check out the pc pro where we built the 32 core

00:13:40.240 --> 00:13:45.600
quadro rtx 8000 monster that you see

00:13:43.519 --> 00:13:49.410
before you this case alone is like over a thousand

00:13:47.360 --> 00:13:50.800
dollars

00:13:50.800 --> 00:13:56.200
and look it doesn't even have as much front io as this one