WEBVTT

00:00:00.400 --> 00:00:06.480
sure most gaming pcs don't spit out plumes of

00:00:04.480 --> 00:00:12.240
smoke while they're running but all that power that your thousand watt

00:00:09.679 --> 00:00:17.039
power supply is sucking up has to come from somewhere and unless you're on

00:00:14.240 --> 00:00:21.039
solar wind or hydro chances are it's fossil fuels so then

00:00:19.359 --> 00:00:24.560
what's an environmentally conscious gamer to do

00:00:22.560 --> 00:00:29.439
well thanks to csonic sponsoring this video we're going to be taking a step

00:00:26.720 --> 00:00:34.960
back today from raw performance and seeing exactly how low we can take our

00:00:32.719 --> 00:00:42.719
power consumption while keeping our games running great

00:00:45.760 --> 00:00:52.960
before we can optimize for power efficiency we need to determine how much

00:00:50.079 --> 00:00:56.480
performance we feel like we need for gaming

00:00:53.840 --> 00:01:00.800
so basically to prevent ourselves from running quake 3 on a raspi and declaring

00:00:58.719 --> 00:01:06.159
victory we're saying that our rig has to run a modern

00:01:02.480 --> 00:01:08.400
aaa title at max ish settings 1080p

00:01:06.159 --> 00:01:13.360
60fps so we're going to use shadow of the tomb raider as our bar

00:01:10.640 --> 00:01:17.920
now strictly speaking most games don't need a six core CPU right now but we're

00:01:15.840 --> 00:01:23.520
setting the bar there for two reasons one even if a game doesn't utilize every

00:01:20.479 --> 00:01:25.759
core background processes still can and

00:01:23.520 --> 00:01:30.479
having extra cores to throw those onto makes a big difference if you're already

00:01:27.439 --> 00:01:32.880
at peak capacity and two we are

00:01:30.479 --> 00:01:35.040
definitely going to be at peak capacity with the clocks we're going to be

00:01:33.840 --> 00:01:40.479
running so then we've really only got a couple

00:01:38.159 --> 00:01:46.159
of choices here representing AMD is either the ryzen 5 3600 or 3600x and

00:01:44.560 --> 00:01:49.759
we're going to opt for the x version here in hopes that its higher silicon

00:01:47.840 --> 00:01:53.360
quality will allow us to undervolt it to a greater degree saving us a little bit

00:01:52.159 --> 00:01:59.520
of power then on the Intel side the obvious

00:01:55.600 --> 00:02:02.079
equivalent is the core i5 9600 kf a

00:01:59.520 --> 00:02:07.439
variant of the 9600k that has its GPU fused off and at least in theory

00:02:04.719 --> 00:02:12.239
unpowered as the kf is a binned CPU compared to the lower end 9400 we should

00:02:10.560 --> 00:02:16.000
be able to get more stable performance at lower voltages on average although

00:02:14.560 --> 00:02:21.040
again it comes down to your luck in the silicon lottery you might think we'd go

00:02:18.560 --> 00:02:24.160
for the equivalent t-series CPU but we couldn't find any we asked Intel about

00:02:22.959 --> 00:02:28.319
this and they said subscribe to pewdiepie we informed them that that's a

00:02:26.160 --> 00:02:31.760
completely dead meme at this point and we ended up getting a much better answer

00:02:29.840 --> 00:02:37.040
second time around as it turns out there's nothing stopping retailers from

00:02:33.840 --> 00:02:39.280
selling their low power t-series chips

00:02:37.040 --> 00:02:43.360
but they're only available in oem tray packaging which means large minimum

00:02:41.280 --> 00:02:46.959
order quantities and reduced warranty support so most apparently haven't

00:02:46.000 --> 00:02:50.879
bothered now we'll want to grab motherboards with

00:02:49.120 --> 00:02:55.360
as few extra bells and whistles as possible while still retaining decent

00:02:53.200 --> 00:02:58.640
power delivery so the mid-range b series offering for both teams fits the bill

00:02:57.360 --> 00:03:02.400
here and since ryzen relies so heavily on

00:03:00.879 --> 00:03:08.480
memory speed for performance we're going to grab a 200 megahertz trident z memory

00:03:05.040 --> 00:03:11.840
kit based on samsung b-dye chips

00:03:08.480 --> 00:03:13.760
why such crazy fancy memory well we're

00:03:11.840 --> 00:03:17.760
going to under-bolt the snot out of it just like our CPU in the interest of

00:03:15.599 --> 00:03:22.080
saving a little bit of power so now we just take what we have drop the clocks

00:03:19.760 --> 00:03:26.239
and voltages and see what happens i guess we should probably assemble it

00:03:23.760 --> 00:03:30.239
first well yes we can't really drop the clocks and voltages unless we have yeah

00:03:28.319 --> 00:03:33.840
right now there's zero one moment please it took a long time to dial in our

00:03:31.840 --> 00:03:39.360
clocks but in the end we were left with both cpus running at 2.3 gigahertz with

00:03:36.640 --> 00:03:43.440
voltages well under 1 volt our memory is running with timings as fast as we can

00:03:41.040 --> 00:03:47.840
manage with as low a voltage as possible now as an aside our AMD system actually

00:03:46.159 --> 00:03:51.680
managed a higher stable memory speed than our Intel system did then for good

00:03:49.760 --> 00:03:56.000
measure we dropped our pci express link speed back to gen 2 for our GPU and gen

00:03:54.480 --> 00:04:00.239
1 for everything else in order to reduce the energy required for those devices as

00:03:57.840 --> 00:04:04.239
well any non-critical devices SATA controllers we aren't using wi-fi

00:04:02.080 --> 00:04:08.080
adapters things like that were disabled and secondary voltages were tweaked

00:04:05.840 --> 00:04:14.239
where possible okay so how are we doing here well my drink is still cold

00:04:12.040 --> 00:04:19.120
lttstore.com but actually what i meant was the CPU testing oh in terms of raw

00:04:17.040 --> 00:04:22.560
numbers our cpus are about even under synthetic load with AMD pulling a couple

00:04:21.040 --> 00:04:27.360
of extra watts but Intel definitively idles lower

00:04:25.120 --> 00:04:30.800
neither CPU idles as low as it could since i ended up needing to set a static

00:04:29.120 --> 00:04:35.840
voltage for stability but overall it seemed like a worthwhile

00:04:33.120 --> 00:04:39.040
trade-off right because we're gaming not using our pc as a glorified digital

00:04:37.840 --> 00:04:44.720
picture frame so then AMD's more watts so we're going

00:04:42.160 --> 00:04:47.600
Intel then right not quite see

00:04:45.520 --> 00:04:52.320
our Intel bench here is having a little bit of trouble with frame pacing

00:04:49.919 --> 00:04:55.440
even with vsync enabled to cap the frame rate

00:04:53.199 --> 00:04:59.440
our AMD bench didn't do that and since they're both so similar in power draw at

00:04:57.440 --> 00:05:04.720
these frequencies already bumping up the core i5's core clocks to fix that makes

00:05:01.840 --> 00:05:08.400
it actually draw more power under load in spite of its instructions per clock

00:05:06.320 --> 00:05:13.039
advantage over ryzen i even tried comparing to the standard 9600k and it's

00:05:10.720 --> 00:05:17.600
the same story though the kf does add a lot lower if that matters at all

00:05:15.280 --> 00:05:21.120
interesting so i guess that means we're going AMD then

00:05:19.360 --> 00:05:26.479
also i couldn't help noticing that you are not using a fan on our cooler here

00:05:23.680 --> 00:05:30.800
wow this thing is barely even warm yeah on our ryzen bench it actually

00:05:28.240 --> 00:05:34.720
didn't even break 50 degrees on open air and i think that's probably thanks to

00:05:32.639 --> 00:05:38.160
its bigger soldered ihs oh right actually that reminds me

00:05:36.320 --> 00:05:42.880
we're doing a video about a really neat new cooler called a thermosyphon

00:05:40.880 --> 00:05:47.919
so make sure you guys are subscribed uh if you want to see how far we can go

00:05:44.639 --> 00:05:49.039
with that um anyway so we've got our CPU

00:05:47.919 --> 00:05:55.039
figured out what about our graphics card surely we're not going to stick with a 2080 ti

00:05:52.160 --> 00:05:58.320
here well i started with an rtx 2080 ti just to eliminate any bottlenecks while

00:05:56.639 --> 00:06:03.360
i was testing the cpus and i've got it under vaulted as far as

00:06:00.080 --> 00:06:04.800
it'll go but i suspect an rtx 2060 might

00:06:03.360 --> 00:06:09.600
actually be better for this with its comparatively simple core layout or

00:06:07.280 --> 00:06:13.039
maybe even rx 5700 well there's only one way for us to find

00:06:11.440 --> 00:06:16.560
out for sure dialing in the GPU was both easier and

00:06:15.120 --> 00:06:20.400
more difficult than you might at first expect yes the tools are easy enough to

00:06:19.039 --> 00:06:26.000
use but there's no clean way to adjust the voltage for a

00:06:23.120 --> 00:06:30.160
frequency just the other way around but using this to our advantage we

00:06:28.000 --> 00:06:34.160
managed to trick our cards into capping out at a given frequency and from there

00:06:32.560 --> 00:06:39.600
it was a matter of figuring out which voltage was both stable and most

00:06:36.400 --> 00:06:41.280
efficient to use i guess team red still

00:06:39.600 --> 00:06:45.199
has some work to do here power consumption was good but

00:06:43.280 --> 00:06:49.600
performance and thermals definitely were not so

00:06:46.400 --> 00:06:51.039
out goes the rx 5700 ouch

00:06:49.600 --> 00:06:54.400
don't worry i'll keep you safe from the NVIDIA shell

00:06:52.639 --> 00:06:59.599
put you over here oh speaking of which um

00:06:56.720 --> 00:07:04.880
i see that you actually turned off the card's lighting and the fans are off

00:07:02.000 --> 00:07:08.400
presumably that's to save some power but how did you stop the fans founders

00:07:06.319 --> 00:07:12.960
editions don't have a zero RPM mode i'm glad you asked i flashed the BIOS that

00:07:10.960 --> 00:07:15.759
was scripted by the way that was scripted by the way

00:07:15.759 --> 00:07:22.400
flashed the BIOS on our card with an evga one so it could be sent to run at

00:07:19.840 --> 00:07:26.720
zero RPM i set the software fan curve to cut in around the 80 degree mark and it

00:07:24.240 --> 00:07:30.319
only really ramps up at around 85 which it never hits on our opener bench

00:07:28.960 --> 00:07:38.000
we're going to have a link by the way to how you can do that in the video description yeah and as for the rtx 2080

00:07:35.039 --> 00:07:42.400
versus 2060 versus 2080 ti no matter what i did i found that around

00:07:39.759 --> 00:07:45.919
700 millivolts was the floor so with that in mind i think the results are

00:07:43.680 --> 00:07:50.240
pretty clear given that the 2060 hits our 60fps target we can't justify the

00:07:48.319 --> 00:07:54.720
bigger GPU unless we're willing to spend 40 watts of our power budget on rtx or

00:07:52.560 --> 00:07:57.840
real-time ray tracing so

00:07:55.680 --> 00:08:01.759
now we're ready to actually build this thing and see how it runs

00:07:59.759 --> 00:08:05.919
do we even need to do that well we don't necessarily need to i just

00:08:04.560 --> 00:08:10.240
figured it would be fun okay yeah let's do that while putting our build together

00:08:08.160 --> 00:08:14.960
we didn't connect our case leds in order to save on that little bit of energy

00:08:12.400 --> 00:08:19.280
that they require and we also decided to only connect the exhaust fan since

00:08:17.280 --> 00:08:23.680
honestly we didn't need much if any additional airflow we chose our power

00:08:21.520 --> 00:08:27.599
supply for a number of reasons and not just because seasonic is sponsoring this

00:08:25.440 --> 00:08:32.320
video the biggest one is that as a titanium unit the efficiency of this

00:08:29.840 --> 00:08:36.640
power supply never drops below 90 under any circumstances period that

00:08:34.800 --> 00:08:41.760
means no matter how low we go we're only looking at 10 wastage at worst and at

00:08:39.360 --> 00:08:47.680
best here on 115 volt mains in north america it reaches nearly 95 efficiency

00:08:45.440 --> 00:08:52.000
just as importantly we need a power supply capable of a rock stable power

00:08:49.600 --> 00:08:56.000
delivery for how low our voltages are any appreciable amount of ripple or poor

00:08:53.760 --> 00:09:00.000
regulation even if it's in spec could cause momentary voltage droops that

00:08:57.680 --> 00:09:03.680
could render our machine unstable or otherwise require us to bump the

00:09:01.519 --> 00:09:07.920
voltages up to compensate similarly to how we chose a bend CPU and GPU we need

00:09:05.920 --> 00:09:11.279
a high-end psu to make things reliable in the interests of efficiency at any

00:09:09.440 --> 00:09:15.200
cost we could have shortened the psu cables as far as they'd go but the

00:09:13.360 --> 00:09:20.240
energy savings is to put it gently quite difficult to

00:09:17.519 --> 00:09:24.320
quantify so this is cool one side effect of having an extremely power efficient

00:09:21.920 --> 00:09:28.959
pc is that you also end up with an

00:09:26.000 --> 00:09:32.000
extremely quiet pc because the entire thing

00:09:29.920 --> 00:09:36.640
is not consuming much energy so there's not a lot of waste heat to get rid of

00:09:34.000 --> 00:09:43.360
so this is great we are right around 30 watts of idle sometimes it jumps up to

00:09:38.880 --> 00:09:46.000
31 sometimes it dips down to 29 28 but

00:09:43.360 --> 00:09:50.160
that is really freaking impressive to put that in perspective i mean compared

00:09:48.000 --> 00:09:55.920
to the incandescent light bulbs that we used to use in our homes that is less

00:09:52.959 --> 00:10:01.440
than a pretty dim 50 watt bulb while the computer is sitting not doing anything

00:09:58.320 --> 00:10:03.920
but of course our real test here is

00:10:01.440 --> 00:10:07.839
while gaming so let's fire up shadow of the tomb raider and see how we're doing

00:10:06.000 --> 00:10:12.080
one thing you might notice despite the fact that this is a 4k monitor and it's

00:10:09.920 --> 00:10:17.760
capable of 144 hertz we're currently running at 4k60

00:10:14.640 --> 00:10:20.240
and it's also capable of HDR that's off

00:10:17.760 --> 00:10:24.320
enabling high refresh rate or HDR increases our power drop by about five

00:10:22.480 --> 00:10:27.760
watts each we don't have the monitor plugged into

00:10:25.839 --> 00:10:32.560
the no but it increases the power draw on the GPU

00:10:30.079 --> 00:10:36.160
i did not know that yeah i didn't either until i looked it up well i tried fresh

00:10:34.640 --> 00:10:41.360
rate obviously because you're pumping out more frames but i didn't know that HDR had such a processing drop yeah me

00:10:39.760 --> 00:10:46.399
neither fascinating yeah you wouldn't think so so to give our system the opportunity to

00:10:44.240 --> 00:10:51.600
draw as much power as possible we've gone ahead and run it for about 10

00:10:49.040 --> 00:10:57.120
minutes in game here to allow our GPU fans to ramp up so what we're gonna see

00:10:54.079 --> 00:11:01.920
then is we've got a peak gaming power

00:10:57.120 --> 00:11:03.519
draw of about 106 to 108

00:11:01.920 --> 00:11:09.279
watts which again to put this in the context of the technology of yesteryear

00:11:07.040 --> 00:11:12.800
when you bought a bright bulb for a large room it was typically like a 100

00:11:11.600 --> 00:11:17.519
watt bulb so our entire gaming system here is

00:11:15.120 --> 00:11:21.680
drawing about as much power as a light bulb from when i was a kid so that's a

00:11:19.519 --> 00:11:25.839
fun fact but it doesn't give us any context for how this solution compares

00:11:23.760 --> 00:11:31.200
to other gaming systems so we're going to go ahead and fire up an xbox one x

00:11:29.279 --> 00:11:36.720
also playing shadow of the tomb raider at 60 FPS and see how that compares

00:11:35.279 --> 00:11:40.000
so we can't guarantee our game is running at exactly the same graphical

00:11:38.480 --> 00:11:44.720
fidelity here because we don't have the same fine control on a console but what

00:11:42.640 --> 00:11:48.720
we can set is the frame rate and the resolution so we're running at 60fps

00:11:47.120 --> 00:11:53.200
1080p high frame rate

00:11:50.800 --> 00:11:57.880
just say 60fps so by comparison our modern console is

00:11:55.600 --> 00:12:01.760
anywhere from 130 to 145 watts depending on what's on screen

00:12:00.959 --> 00:12:06.480
now to be clear we're not exactly declaring

00:12:04.240 --> 00:12:09.519
victory for the pc master race here that's

00:12:07.839 --> 00:12:13.040
actually really impressive but that's of course one of the benefits of having

00:12:10.959 --> 00:12:17.279
something that is fully integrated and there is a bit of a cost advantage on

00:12:14.639 --> 00:12:20.160
the console side as well we're just wanting to provide you guys with the

00:12:18.720 --> 00:12:25.120
context if you're willing to spend the money and spend the time you can get a

00:12:22.399 --> 00:12:31.360
more efficient gaming experience on pc than you can even on a fully boxed in

00:12:28.800 --> 00:12:35.040
custom designed game console so thanks for watching guys even if you don't care

00:12:33.040 --> 00:12:38.399
about having the most efficient gaming pc hopefully you picked up a couple tips

00:12:37.040 --> 00:12:42.399
and tricks here because aside from saving power tuning your pc and choosing

00:12:40.880 --> 00:12:46.399
the right components the way that we did today can also help you save some heat

00:12:44.560 --> 00:12:50.160
output during the summer which can be pretty freaking nice and some noise

00:12:48.720 --> 00:12:55.200
because the more efficiently your system is running the less fan noise you have to deal with massive thanks to sea sonic

00:12:53.839 --> 00:12:59.440
for sponsoring this video we've actually wanted to do it for quite some time and

00:12:58.240 --> 00:13:06.000
if you guys are looking for another video to watch maybe watch the polar opposite of this one yeah yeah where we

00:13:03.040 --> 00:13:10.880
overclock an 11 year old pc that's basically kind of like the muscle car

00:13:08.399 --> 00:13:16.639
of uh of computing from back over a decade ago that was not power efficient

00:13:12.880 --> 00:13:16.639
at all let that link below