WEBVTT

00:00:00.000 --> 00:00:06.660
I always thought that when it comes to water cooling more is automatically

00:00:04.470 --> 00:00:10.610
better that is to say if one radiator is not enough you can just add another one

00:00:08.519 --> 00:00:17.730
or another two to get better performance well it seems that I've been water

00:00:13.830 --> 00:00:19.830
cooling incorrectly for years or at

00:00:17.730 --> 00:00:25.500
least that's what Coursera leads me to believe like what how could that

00:00:22.980 --> 00:00:31.080
possibly be that adding more radiators wouldn't help I just can't believe that

00:00:28.340 --> 00:00:34.980
to find out exactly what they said to me you're gonna have to sit through the

00:00:32.399 --> 00:00:39.750
segue to xsplit makes powerful streaming PC apps for streamers vloggers and more

00:00:37.440 --> 00:00:45.590
check out their V cam software to change your background without a green screen

00:00:41.780 --> 00:00:45.590
learn more at the end of the video

00:00:52.060 --> 00:01:00.320
when we showed our concept cooling system for our hack Pro 2019 with a

00:00:57.620 --> 00:01:05.840
front intake radiator and a rear exhaust radiator Corsair emailed me to say that

00:01:03.350 --> 00:01:10.370
according to their simulation data that second radiator would be at best

00:01:07.520 --> 00:01:15.110
pointless and could even make our cooling worse because it adds more

00:01:12.590 --> 00:01:20.960
airflow resistance and because I mean think about it the second radiator can't

00:01:17.479 --> 00:01:23.869
cool water with hot air from inside the

00:01:20.960 --> 00:01:27.890
case I ignored them and then when we showed off Jake's custom minecraft

00:01:25.640 --> 00:01:33.259
server that used not one not two but three thick alpha cool radiators stacked

00:01:31.310 --> 00:01:38.690
against each other with layers of Noctua eighty millimeter fans in between course

00:01:35.360 --> 00:01:41.420
I reached out again to admonish me for

00:01:38.690 --> 00:01:47.360
water cooling wrong the problem is like guys my experience tells me that

00:01:43.720 --> 00:01:50.270
stacking radiators works great or at

00:01:47.360 --> 00:01:54.080
least I thought it did I've never done a side-by-side comparison it makes sense

00:01:52.610 --> 00:02:00.170
though like you would think that doubling the surface area radiating heat

00:01:56.990 --> 00:02:02.150
away from the system must lead to if not

00:02:00.170 --> 00:02:07.310
double than at least significantly better performance but guys turns out

00:02:05.180 --> 00:02:11.510
it's not that simple in order to dissipate heat in a water cooling system

00:02:09.170 --> 00:02:16.160
there needs to be a difference in temperature across the materials in your

00:02:13.910 --> 00:02:20.750
heat exchanger in this case that's the ambient air which is cooled and the

00:02:18.410 --> 00:02:24.489
water which is warm from your components and then this is really important the

00:02:22.880 --> 00:02:30.950
greater that difference in temperature the more heat can be exchanged and the

00:02:27.920 --> 00:02:33.350
thing is because modern radiators are so

00:02:30.950 --> 00:02:37.850
efficient as Corsair if the water temperature is let's say 40 degrees

00:02:35.660 --> 00:02:42.530
under load we can bet that the air coming out of that rad is going to be

00:02:40.070 --> 00:02:47.180
just below that meaning that by the time it picks up some extra heat from air

00:02:44.540 --> 00:02:52.640
cooled components like motherboard VRMs it's not gonna have any heat capacity

00:02:49.640 --> 00:02:54.950
left to pull heat from a second radiator

00:02:52.640 --> 00:03:00.620
that it's going through that kind of makes sense but I wasn't ready to give

00:02:57.380 --> 00:03:03.140
up yet so to determine once and for all

00:03:00.620 --> 00:03:07.360
if adding a second rad is useless we decided to build a test rig

00:03:05.240 --> 00:03:15.020
call and got to work and piece together there it is that little tower sporting

00:03:10.550 --> 00:03:17.210
an RTX 2080 TI and Intel's 9900 KS or as

00:03:15.020 --> 00:03:21.770
I like to call it my little furnace together this pair can draw a little

00:03:19.400 --> 00:03:27.530
over 400 watts of power under a synthetic load and all that heat is

00:03:24.200 --> 00:03:29.180
going right into our water bt-dubs guys

00:03:27.530 --> 00:03:31.880
if you like water loops get subscribed we're planning to water cool a gaming

00:03:30.590 --> 00:03:38.090
chair and you're not gonna want to miss that if you get warm in the summer so our first test is to establish a

00:03:35.450 --> 00:03:41.690
baseline with a single 240 millimeter rad upfront pulling air in to be

00:03:40.190 --> 00:03:46.400
exhausted out of the top of the case with a second pair of matching fans the

00:03:44.270 --> 00:03:51.440
system stays cool enough not to thermal throttle but it's also obvious that our

00:03:48.980 --> 00:03:56.720
radiator is up against its limit so according to my common-sense PC building

00:03:54.410 --> 00:04:00.920
wisdom this system would have looked like a perfect candidate for just

00:03:58.790 --> 00:04:05.180
throwing another radiator at the problem with that out of the way then it's time

00:04:03.140 --> 00:04:09.860
to plumb up our second rad in series and put it in the top of the case with more

00:04:07.340 --> 00:04:14.030
or less the same fan configuration now we did end up having some RAM clearance

00:04:12.080 --> 00:04:19.250
issues requiring us to put the fans on top of the case but that should not

00:04:16.760 --> 00:04:23.810
negatively impact our data the important thing is that the radiator is still

00:04:21.350 --> 00:04:28.490
inside the case exposed to all the other heat sources that are in there now for

00:04:26.390 --> 00:04:32.660
the moment of truth our stress test was designed to hit both the CPU and

00:04:30.320 --> 00:04:36.170
graphics card with fir mark while logging all the pertinent data using

00:04:34.550 --> 00:04:41.540
hardware info that means not just temperatures but also power consumption

00:04:38.540 --> 00:04:43.160
and clock speeds in each scenario the

00:04:41.540 --> 00:04:47.690
system was allowed to hit a steady-state and then we stopped the loads and let

00:04:45.500 --> 00:04:52.040
the system cool back down to our idle temperatures let's look at the numbers

00:04:49.550 --> 00:04:56.090
popping right into the GPU temps the redline is with the single rad and the

00:04:54.170 --> 00:05:01.040
green line is with our dual rad setup not only did our dual rad setup take way

00:04:58.610 --> 00:05:08.000
longer to reach its max temperature it also ran way cooler about 20 degrees

00:05:04.670 --> 00:05:11.300
cooler in fact so that's it Corsair your

00:05:08.000 --> 00:05:13.040
myth is busted and a video I'm out just

00:05:11.300 --> 00:05:18.110
kidding we did our due diligence and check the CPU as well to make sure that

00:05:15.890 --> 00:05:24.419
both loops were actually experiencing the same load and they were both

00:05:21.750 --> 00:05:29.550
scenarios had the CPU clock between 4.5 and 4.6 nothing out of the ordinary

00:05:26.430 --> 00:05:33.270
there and our power usage looked nearly

00:05:29.550 --> 00:05:35.610
identical as well so then why didn't our

00:05:33.270 --> 00:05:40.260
results match Corsairs claims to be clear guys I'm not saying that they're

00:05:37.560 --> 00:05:45.690
CFD analysis or their real-world testing is wrong or that what they said doesn't

00:05:42.450 --> 00:05:48.210
make sense and let's remember as well

00:05:45.690 --> 00:05:53.280
they didn't actually critique this exact build all I'm trying to say is that this

00:05:50.850 --> 00:05:57.870
kind of simulation rarely translates perfectly to the real world and also

00:05:55.950 --> 00:06:01.979
that you should go delta-t sir calm ok no that was an aside so our working

00:05:59.880 --> 00:06:07.200
theory is that the openings in the case at both the back and the bottom allowed

00:06:04.380 --> 00:06:12.510
for some fresh ambient air to come in and mix with the hot air before being

00:06:09.930 --> 00:06:16.770
pulled up through the second read now if we had sealed off every potential air

00:06:14.880 --> 00:06:21.750
leak it's possible we would have seen something closer to Corsair simulation

00:06:18.990 --> 00:06:25.140
but as far as we can tell that's just not a true representation of the real

00:06:23.640 --> 00:06:30.120
world I mean nobody would do that another thing to consider is that every

00:06:27.720 --> 00:06:34.950
system is a little different fan orientations fan speeds radiator

00:06:32.700 --> 00:06:39.990
placement pump flow rates case designs all these things will affect how a

00:06:37.680 --> 00:06:43.740
cooling system works as we've seen with the minecraft server and the hack probe

00:06:41.970 --> 00:06:47.190
sometimes the truth of the matter is you've just got to work with the space

00:06:45.479 --> 00:06:51.990
you've got and we also did another sanity check we had Jake remove one of

00:06:50.220 --> 00:06:56.729
his stacked radiators which ended up resulting in a 5 degree increase in

00:06:53.970 --> 00:07:01.229
temps under load that's another proverbial nail in this coffin now I'm

00:06:59.250 --> 00:07:06.870
sure that in some rare cases it could actually be a negligible or a bad thing

00:07:04.050 --> 00:07:12.300
to add a second radiator but a blanket statement saying that stacking rads is

00:07:08.700 --> 00:07:13.470
bad just isn't even nearly true and you

00:07:12.300 --> 00:07:18.360
should really look at your particular use case and evaluate for yourself if

00:07:16.470 --> 00:07:22.470
you can so the moral of the story here is take everything with a grain of salt

00:07:20.580 --> 00:07:25.710
including what I'm saying right now do your own research and do a little

00:07:24.000 --> 00:07:31.110
benchmarking on your phone on your phone on your own it's easy free fun and can

00:07:29.430 --> 00:07:34.570
tell you loads about how to make your system work best for you so that's it if

00:07:33.640 --> 00:07:40.180
you guys want to check out some more benchmark related videos why not take a look at the video we did last year about

00:07:37.960 --> 00:07:44.380
finding out if your gaming rig is bottlenecked we're gonna have that

00:07:41.440 --> 00:07:48.160
linked in the end screen for you thanks to xsplit for sponsoring today's video

00:07:46.000 --> 00:07:52.600
with xsplit v-cam you can remove replace and blur your background without a green

00:07:50.650 --> 00:07:57.010
screen it works with any webcam and there's no need for complicated lighting

00:07:54.370 --> 00:07:59.860
setups it uses AI to automatically tell the difference between you and the

00:07:58.450 --> 00:08:04.030
background and you can customize the background to be anything you want DCAM

00:08:02.110 --> 00:08:09.220
works with other xsplit broadcast tools but also slack hangout Skype and more

00:08:06.970 --> 00:08:12.360
and Cleave got 24/7 support and lifetime updates and you can check it out at the

00:08:10.870 --> 00:08:15.900
link in the video description wow I already kind of through to another

00:08:14.470 --> 00:08:19.580
video so