WEBVTT

00:00:00.080 --> 00:00:06.799
obviously over the next couple of years we're going to be building a ton of sick

00:00:04.000 --> 00:00:12.480
water cooled gaming rigs using rtx 3000 series graphics cards like this one

00:00:09.840 --> 00:00:19.279
but while in the past we were able to use the old rule of thumb which is about

00:00:15.679 --> 00:00:21.840
one 120 millimeter worth of radiator per

00:00:19.279 --> 00:00:28.160
heat generating component in the system i don't necessarily think that the rules

00:00:24.720 --> 00:00:31.840
of it should work necessarily apply to

00:00:28.160 --> 00:00:35.280
the rtx 3000s particularly the power

00:00:31.840 --> 00:00:37.440
hungry and enormous rtx 3090. so today's

00:00:35.280 --> 00:00:42.000
video is going to be us taking you along for the journey of some work that we

00:00:39.360 --> 00:00:48.640
would have had to do anyway figuring out exactly how much radiator we need for a

00:00:45.520 --> 00:00:51.840
water cooled rtx 3090 we're going to

00:00:48.640 --> 00:00:53.600
start with a slim single which probably

00:00:51.840 --> 00:00:58.960
ain't going to work and work our way all the way up to well realistically

00:00:56.079 --> 00:01:02.160
whatever it takes just like i'll do whatever it takes to transition to our

00:01:01.120 --> 00:01:06.240
sponsor ridge wallet wants to redefine the

00:01:04.239 --> 00:01:09.760
wallet with its compact frame and rfid blocking plates check out how they can

00:01:08.000 --> 00:01:16.400
keep your wallet bulge down and use offer code Linus to save 10 and get free

00:01:12.320 --> 00:01:16.400
worldwide shipping at the link below

00:01:23.920 --> 00:01:30.320
before we can water cool it though we need to figure out how well the system performs when it's just air cooled so

00:01:29.200 --> 00:01:34.960
for the testing we're going to be using this ASUS strix rtx 3090 you might

00:01:32.720 --> 00:01:38.799
remember it from our is sli actually dead video it

00:01:36.479 --> 00:01:43.040
basically is so yeah let's just chuck it in here and find out how well she goes

00:01:40.640 --> 00:01:47.280
our fan config in this case is a bit overkill if i'm honest uh we have three

00:01:45.600 --> 00:01:50.560
fans in the base that are just blowing straight up into it and three fans up

00:01:48.799 --> 00:01:55.280
here and three they're blowing up there nine in total it's a bit ridiculous

00:01:53.840 --> 00:01:59.200
the same time like you know you want best case scenario for your card so

00:01:57.200 --> 00:02:03.600
we've got our test set up here just a very simple time spy extreme we're going

00:02:01.360 --> 00:02:09.200
to enable looping and let it go for 30 minutes

00:02:06.079 --> 00:02:10.250
it's been 30 minutes just 32 at this

00:02:09.200 --> 00:02:15.120
point and what oh my god you piece of crap why is

00:02:13.520 --> 00:02:20.160
there no monitoring data yeah i have to do it again well see you tomorrow

00:02:18.160 --> 00:02:24.160
okay testing's done it looks like we don't have any issues with thermal

00:02:22.000 --> 00:02:29.120
throttling whatsoever so hovering right around 1800 megahertz and

00:02:27.520 --> 00:02:31.520
let's see here but come on

00:02:31.680 --> 00:02:37.040
what a beautiful card i haven't been this excited to take something apart for

00:02:34.959 --> 00:02:40.480
a really long time this is my first look at the inside of an ampere card and what

00:02:39.360 --> 00:02:45.920
a beauty i'm going to be using the ek quantum

00:02:42.959 --> 00:02:49.720
vector today to water cool it here's the back plate

00:02:51.760 --> 00:02:58.480
oh that's actually really nice damn look at that

00:02:56.080 --> 00:03:03.519
oh that's just so nicely machined the real star of the show here yes water

00:03:01.040 --> 00:03:07.920
blocks 175 that's enough to buy yourself

00:03:05.280 --> 00:03:12.159
a pretty decent GPU but you know you've already got 30 90. so who cares

00:03:10.080 --> 00:03:15.920
everything should be cnc machined i was hoping that we'd get a see-through one

00:03:13.760 --> 00:03:19.120
this is kind of a pre-production block so we don't have you know

00:03:17.599 --> 00:03:23.760
the things you might expect it to include like instructions

00:03:21.599 --> 00:03:28.800
this is all that i've got it's a little pdf that includes where the thermal pads

00:03:26.640 --> 00:03:33.840
go and that's it that's all we get probably

00:03:30.720 --> 00:03:33.840
going to be fine

00:03:34.000 --> 00:03:39.040
all right all the thermal pads are on

00:03:37.040 --> 00:03:46.519
finally get to just apply some thermal paste

00:03:41.440 --> 00:03:46.519
get it all over here oh wow

00:03:52.640 --> 00:03:59.519
huh all of these mounting holes don't quite line up on the block because of

00:03:57.120 --> 00:04:05.120
these two capacitors right here yeah this is bad so should we

00:04:02.879 --> 00:04:07.439
yes oh

00:04:09.120 --> 00:04:15.280
oh garlic okay we should

00:04:13.519 --> 00:04:19.280
we should have the clearance now but hopefully by the time you know actual

00:04:17.359 --> 00:04:23.919
real people get this they don't have to do that even with a

00:04:21.600 --> 00:04:28.400
water block on it the thing is still huge yeah massive shout out to ek by the

00:04:26.880 --> 00:04:34.320
way for sending us over one of their first vector strix blocks for the rtx

00:04:31.199 --> 00:04:37.280
3090 from ASUS um

00:04:34.320 --> 00:04:40.639
it did come with some growing pains in terms of installation

00:04:39.199 --> 00:04:45.600
they actually hadn't accounted for a couple capacitors here so we had to make

00:04:43.520 --> 00:04:50.080
our own clearance for them but now that it's on theoretically i think we should

00:04:47.360 --> 00:04:54.400
be okay so this is as simple as let's fire it up and get going yeah power this

00:04:52.000 --> 00:04:56.639
puppy up we're gonna need not one not two

00:04:55.280 --> 00:05:00.400
but three eight pin pci express connectors

00:04:59.199 --> 00:05:04.479
and then i've gone ahead and installed the plugs down here and we're gonna throw some fittings on here and build

00:05:03.280 --> 00:05:08.320
ourselves the dirtiest water cooling loop that you

00:05:07.120 --> 00:05:14.160
ever did see i mean not literally dirty it's going to have clean water in it but just

00:05:12.560 --> 00:05:18.400
kind of guts spilling all over the table style obviously NVIDIA's last few

00:05:16.560 --> 00:05:23.520
generations of graphics cards have thermal throttle that about 80 degrees

00:05:20.800 --> 00:05:28.320
so effectively that you can't really use temperatures to determine if a cooling

00:05:25.840 --> 00:05:32.560
solution is adequate nowadays you have to just monitor the boost clocks that

00:05:30.240 --> 00:05:38.000
the card runs at to see if it's reaching its full potential so this air cooled

00:05:35.520 --> 00:05:42.560
rtx 3090 from ASUS represents the numbers that we actually already ran on

00:05:40.720 --> 00:05:46.960
this exact card because there are variances from one to the other and

00:05:44.400 --> 00:05:51.680
that'll tell us how much radiator reaches the threshold of the air cooler

00:05:49.120 --> 00:05:55.440
and then how much radiator we need to if there is any more headroom exceed it

00:05:53.759 --> 00:05:59.919
for convenience we're using one of these pump reservoir combo units these really

00:05:57.520 --> 00:06:03.680
are the fastest way to fill up a loop and get all the bubbles out of it

00:06:01.280 --> 00:06:07.680
because the water goes directly into the inlet of the pump and you've got this

00:06:05.440 --> 00:06:12.160
little anti-vortex thing down here in the bottom that helps the bubbles make

00:06:10.160 --> 00:06:17.360
their way out of the liquid one thing to watch out for is modern GPU blocks like

00:06:15.039 --> 00:06:21.840
CPU blocks can be directional so this particular one has the inlet on the left

00:06:19.440 --> 00:06:25.600
here so the water flows down into the cold plate and then it actually splits

00:06:23.919 --> 00:06:29.919
and flows along two different paths to help with flow restriction boy when we

00:06:27.360 --> 00:06:35.120
build an ugly loop we build an ugly loop that's not that bad Linus this sag is

00:06:32.160 --> 00:06:38.880
something else that's pretty bad card looks curved

00:06:36.880 --> 00:06:43.440
that's okay though because you know you can make yourself feel better with a

00:06:41.039 --> 00:06:46.960
refreshing drink from lttstore.com hey there it is all right to keep our

00:06:45.039 --> 00:06:50.960
results consistent the pump is going to be run at full speed all the time and

00:06:49.039 --> 00:06:55.280
then our fan for the radiator or radiators is going to go to the system

00:06:53.840 --> 00:06:58.720
fan header and we're going to configure it in the BIOS so that they're all

00:06:56.960 --> 00:07:03.199
running at about i don't know what's reasonable like 1200 RPM i think is a

00:07:00.800 --> 00:07:06.400
good balance between noise and cooling so we'll go with that look at all those

00:07:04.639 --> 00:07:11.360
bubbles in there you got to get them out see now it's grabbing that water oh this is

00:07:09.599 --> 00:07:16.639
great we were fighting with smart fan mode with all these different temperatures which by the way is very

00:07:14.160 --> 00:07:20.880
cool and you know different fan rpms but you can just turn it off that's not a

00:07:18.639 --> 00:07:26.240
drop down it's a check mark and you can just set it 50 which gives

00:07:23.120 --> 00:07:29.199
us around 1100 RPM on this fan which is

00:07:26.240 --> 00:07:33.440
pretty quiet like but then the main reason that you go water cooling these

00:07:30.880 --> 00:07:37.120
days is of course a for looks and b for quieter operation if you don't want it

00:07:34.880 --> 00:07:40.000
quiet don't water cool i mean it's gotta where is it look at the cooler on this

00:07:39.120 --> 00:07:45.680
thing like so yeah that's what we're gonna settle

00:07:42.960 --> 00:07:51.440
on it's pretty crazy to see an rtx 39d even at idle running at 32 degrees

00:07:48.560 --> 00:07:54.400
celsius damn dang and that's even with room temperature

00:07:53.120 --> 00:08:00.319
water that we put in here because we didn't want to wait forever for it to heat up and reach equilibrium

00:07:57.759 --> 00:08:03.840
good luck little radiator okay i'm rooting for you buddy i'm gonna take

00:08:02.319 --> 00:08:09.680
lunch we returned from lunch to find our system black screened and unresponsive

00:08:07.840 --> 00:08:14.400
thankfully our data logging still worked and we discovered that our GPU shot up

00:08:11.599 --> 00:08:18.800
to 87 degrees immediately after the test started which is obviously not right

00:08:16.639 --> 00:08:23.520
even with an inadequate radiator the water would still need time to heat up

00:08:22.720 --> 00:08:29.039
so upon further investigation we discovered

00:08:26.080 --> 00:08:34.240
that in ek's mad rush presumably to get us one of these blocks ahead of retail

00:08:30.800 --> 00:08:36.399
availability they the block is great but

00:08:34.240 --> 00:08:41.120
they appear to have made a small boo-boo and sent us two millimeter rather than

00:08:38.880 --> 00:08:46.720
one millimeter thermal pads so here's the one they sent us against a one mil

00:08:44.240 --> 00:08:50.959
what that did was actually prevented the GPU from making proper contact you can

00:08:49.120 --> 00:08:53.839
see we've done a test fit now and everything's doing great with the one

00:08:52.320 --> 00:08:57.519
mils so we just need to remount this and take another crack at it ow

00:08:57.839 --> 00:09:00.880
it's really hot

00:09:00.959 --> 00:09:08.080
that's not quite as bad but the tanks on the sides of the radiator

00:09:05.680 --> 00:09:11.920
i'm surprised we changed our test methodology long story but we're using

00:09:10.480 --> 00:09:14.959
fur mark and letting it reach equilibrium then logging for five

00:09:13.680 --> 00:09:20.760
minutes and it's like turboing completely to two gigahertz and

00:09:19.279 --> 00:09:25.839
maxing out it 66 degrees

00:09:23.600 --> 00:09:30.240
in fact what's weird is like the back of the GPU we put some tape on it to deal

00:09:27.839 --> 00:09:34.640
with the reflectiveness of the of the back plate there

00:09:32.000 --> 00:09:40.480
so the GPU itself is actually running at around the same temperature

00:09:36.480 --> 00:09:42.399
as the back of the like back plate crazy

00:09:40.480 --> 00:09:46.959
that RAM on the back of the card must get freaking toasty maybe we shouldn't

00:09:44.959 --> 00:09:54.000
be that surprised though i mean we've seen a 1080 ti and what an 8600k on a

00:09:51.040 --> 00:09:56.560
single 8700k on a single 120 millimeter radiator

00:09:55.200 --> 00:10:02.480
that's not to say though that we think this is ideal let's see how much better

00:09:59.120 --> 00:10:02.480
we can do with a better rad

00:10:03.519 --> 00:10:09.600
to say our results from the thick single were a disappointment would be an

00:10:07.120 --> 00:10:14.160
understatement we got basically the exact same temps even though we amped up

00:10:12.320 --> 00:10:18.560
the thickness of our radiator although if you think about it it does kind of

00:10:16.240 --> 00:10:23.279
make sense because we are adding some more surface area but we're also adding

00:10:21.120 --> 00:10:28.000
more restriction to the airflow so you tend to need higher RPM or higher static

00:10:26.000 --> 00:10:31.440
pressure fans to get the most out of thicker rads actually here's another

00:10:29.760 --> 00:10:35.120
thing i didn't notice until i went to do that b-roll shot showing the relative

00:10:33.680 --> 00:10:38.560
thickness the high-performance thicker one

00:10:36.959 --> 00:10:42.800
actually has kind of like a built-in shroud so you can see there's more space

00:10:40.959 --> 00:10:45.360
between where it screws in where the fin stack is

00:10:44.160 --> 00:10:51.040
so the difference in thickness of the actual cooling element is not as much

00:10:49.760 --> 00:10:55.040
you can see it also has a much more dense fin arrangement it's possible then

00:10:53.120 --> 00:10:59.120
that a more powerful fan could have improved our results with the thicker

00:10:56.560 --> 00:11:02.160
radiator but not to the degree that we'd be getting temperatures that i'd be

00:11:00.880 --> 00:11:07.279
happy with on water cooling so let's move on to a

00:11:04.800 --> 00:11:12.880
dual 120 millimeter radiator this is also of the dense fin thicker variety

00:11:10.240 --> 00:11:17.680
come on pump you can do it half an hour into our dual rad i gotta

00:11:15.120 --> 00:11:23.440
say i am blown away like i knew it was gonna perform better two fans is

00:11:20.000 --> 00:11:28.560
definitely better than one but wow we

00:11:23.440 --> 00:11:32.320
went from 66 degrees down to 48

00:11:28.560 --> 00:11:34.560
degrees that's almost a 20 degree drop

00:11:32.320 --> 00:11:39.040
just by adding a little bit more radiator surface area so without testing

00:11:36.959 --> 00:11:43.519
the triple i can already say that this is probably going to be the sweet spot

00:11:40.959 --> 00:11:46.240
for performance and noise just because i'm not expecting a big

00:11:44.880 --> 00:11:52.160
difference in performance this time around though doesn't mean i can skip out on testing the triple what you get

00:11:49.839 --> 00:11:56.800
with an overkill-sized radiator in addition to potentially lower coolant

00:11:54.000 --> 00:12:00.160
temperatures is more headroom for optimizing your system for example if

00:11:59.120 --> 00:12:04.480
you've got restrictive grills on your case a larger

00:12:02.640 --> 00:12:08.240
radiator can help compensate for less airflow in the same way you could just

00:12:06.480 --> 00:12:12.800
turn your fan speeds down if you wanted your system to operate more quietly

00:12:10.240 --> 00:12:16.079
finally of course there's the f word future proofing

00:12:14.240 --> 00:12:20.000
if you've got a larger radiator and you do decide to add another component to

00:12:18.000 --> 00:12:24.320
your loop down the road it's more likely that you'll be able to do so without

00:12:22.079 --> 00:12:28.480
changing out your rad as expected we only dropped about four degrees this

00:12:26.000 --> 00:12:33.200
time telling us that our sweet spot really is a double radiator but if we

00:12:31.200 --> 00:12:35.839
wanted some room to grow or we knew that we were going to be giving up a bunch of

00:12:34.560 --> 00:12:41.120
our airflow to something like a restrictive fan filter then a triple

00:12:38.240 --> 00:12:45.360
would be a great bet for an rtx 3090. this is crazy because

00:12:43.200 --> 00:12:50.839
that means that doing small form factor machines with a high performance CPU

00:12:47.920 --> 00:12:56.639
and an rtx 3090 is basically going to require either a dual and a single or

00:12:54.160 --> 00:13:01.120
two triples and remember too that the scenario we're testing here assumes that

00:12:59.200 --> 00:13:06.639
our radiators are going to have access to fresh air sometimes they're going to

00:13:03.519 --> 00:13:09.680
have a mixture of fresh air and case air

00:13:06.639 --> 00:13:11.839
in fact like you see in a case like this

00:13:09.680 --> 00:13:16.320
one where there's kind of like a gap between the back panel and the fans

00:13:14.000 --> 00:13:20.639
where it can kind of mix what comes in through the grill and what's already

00:13:18.399 --> 00:13:23.519
present in the case so

00:13:21.600 --> 00:13:26.880
now we know just like we all knew that i was going

00:13:25.200 --> 00:13:32.399
to use this opportunity to tell you about our sponsor micro center promises

00:13:29.839 --> 00:13:36.560
you the best prices and selection across any of their 25 locations across the

00:13:34.800 --> 00:13:40.959
united states you can join the new microcenter community to give and

00:13:38.560 --> 00:13:45.279
receive advice on pc part selection and hardware and you can check out their

00:13:42.639 --> 00:13:48.720
custom pc builder configurator just check out the link in the video

00:13:46.560 --> 00:13:54.639
description for a free 32 gig flash drive and 32 gig micro sd card valid in

00:13:51.680 --> 00:13:58.480
store only no purchase necessary if you guys enjoyed this video you might

00:13:56.399 --> 00:14:02.320
enjoy other water cooling experiments like oh shoot i don't know if this one's

00:14:00.240 --> 00:14:07.199
going to be up yet by the time you guys watch this but uh

00:14:04.959 --> 00:14:11.199
what what in the heck is this thing ml360 sub-zero