NVIDIA GeForce GTX 670 Temperature Acoustics and Power Consumption Linus Tech Tips
Linus Tech Tips
·Linus Tech Tips
·2013-05-07
·
1,252 words · ~6 min read
0:07
now while you may see something like this on my test bench and assume that
0:10
this is going to be a video about three-way SLI scaling on the GTX 670
0:16
that is not actually the topic of this video that will be a separate video
0:19
where I'm going to do GTX 670 3-way
0:22
versus Radon hd6970 3way because the
0:26
performance of these cards is pretty close so I am very interested to see who
0:30
scales better at a large at a high resolution like 2560 x600 but no this
0:37
video is about temperatures power
0:40
consumption and noise yes those three things of the
0:46
GTX 670 compared to its immediate competition that is the GTX 680 the 7970
0:52
the 7950 the 7870 and the card that it replaces the
0:57
GTX 570 so one of the points I just
1:00
wanted to make to open this one up that's why I have three cards on the
1:03
bench right now is that even with these three
1:07
cards we are drawing less than 100 Watts
1:11
from the wall at idle this is with an overclocked 4.4 GHz
1:16
um 3930 3630
1:21
K 3930 sorry guys it's like 4 in the
1:25
morning and I'm really tired so it's a 3930k at 4.4 GHz with all the power
1:29
saving features turned off 16 gigs of RAM SSD boot Drive big old XL ATX
1:35
motherboard three GTX
1:38
670s and this is running off of a 1250
1:41
wat 80 plus gold power supply so what that means is because this is a 1250 W
1:45
power supply is extremely inefficient at only 10% load so in terms of actual
1:50
power consumption this system is probably in the neighborhood of about 80
1:53
Watts at idle which means these cards are pulling next to nothing at idle this
1:58
GTX 670 is very power effici I'm going to show you some graphs and charts as
2:02
well as some audio clips comparing its cooler and its uh cooling
2:06
characteristics as well as its power draw to its competition now I showed you
2:11
guys in the introduction that the system is drawing less than 100 Watts at the
2:14
wall so what that means is that when we have a single card drawing 70 Watts from
2:18
the wall those two additional cards running in three-way SLI when they're
2:22
idling are basically almost completely shut down which is very very cool it
2:27
makes for a much more efficient Computing experience because you're not
2:32
always going to be loading up all three cards with the latest 3D game sometimes
2:35
you have to check your email and it's great if you're not sucking a ton of
2:38
power back and paying an immense electricity bill on account of your uh
2:43
your massive gaming rig and this is this is totally labeled incorrectly so I
2:48
apologize for that this is uh
2:53
uh Power in watts and and then less is
3:00
better in case you guys didn't know that okay so what else we got here so you can
3:04
see that the GTX 670 is very competitive in terms of its performance per watt
3:09
because it does outperform the hd7950 and it performs a little bit less
3:13
than the 7970 so it falls right into where it's supposed to be um the idle
3:19
power is very very very exciting as well and you can see it draws just a little
3:24
bit less power than its older brother the GTX 680 which stands to reason when
3:28
you compare it to the previous generation product it smokes it not only
3:31
in terms of performance but also in terms of both gaming load power as well
3:36
as Idle power so how did I measure gaming load power using this guy right
3:40
here and then I would stand next to a particular car in Battlefield 3 in the
3:45
uh in the underground parade where I was getting very consistent power readings
3:49
from all the cards that were pretty near the peak that I was seeing during gaming
3:53
just glancing at my power meter the whole time so that's how I'm measuring
3:56
the maximum power no it's not 100% scientific but does give you a pretty
4:00
good idea of what you're going to be able to expect from any of these given
4:05
cards and the way I was measuring idle power was just with my spreadsheet here
4:09
open and nothing else happening so all of these new fangled cards are so
4:13
efficient at idle um that we only really start to see some separation once they
4:17
really get loaded up now let's talk temperatures now these are going to be a
4:21
little bit tough to compare because the 7950 data especially is not going to be
4:26
very similar to the other data because this is not using a reference cooler
4:30
this is using a sapphire dual fan cooler that does a as you can clearly see
4:35
significantly better job than the than the reference cooler that's on the 7970
4:40
right there even accounting for the fact that the 7950 does not run as hot so
4:44
bear that in mind these numbers are not directly comparable between the 670 and
4:49
the 7950 so the 670 runs pretty much the
4:54
same temps as its older brother I measured these using pretty much the
4:57
same method so I'm taking the idle temperature once it's idled at the desktop for you know 5 minutes or
5:02
whatever and it's pretty much settled into a very flat line and load temps are
5:06
taken when I'm looking at that same car so that's after I've been benchmarking for about 15 minutes already with the
5:11
games beforehand and then I get onto Battlefield 3 and I'm seeing at that
5:15
point as well a very very flat line so that's when the temperatures have
5:19
completely equalized so this gives you some idea what to expect from a
5:22
reference card however even at launch for GTX 670 there are non- reference
5:28
cards available already and I'm having trouble finding it right now but I have
5:32
one lying around here somewhere there you go galaxy has a pretty sweet
5:36
non-reference card at launch that has like a glowing GeForce GTX 670 logo
5:43
thing going on and dual 92 mm cooling
5:46
fans and it looks just like beastly and everything so so bear that in mind guys
5:50
this is just based on reference cards based on a non- overclocked card and I
5:55
guess that's pretty much all there is to it they run about what you would expect
5:59
compared to similar competition like the hd7970 a little bit cooler at idle a
6:04
little bit hotter under load now guys you're going to see some
6:08
clips of just you know close-ups of the video cards like
6:11
this these are once again not scientific these are not meant to evaluate the
6:17
exact you know deciel level noise level of these cards they are meant to give
6:21
you an idea of the idle and
6:25
load um signature of the noise so
6:28
whether it's a low growl or whether it's kind of a high pitch squeal um that is
6:33
what you guys can evaluate with these clips it should give you some idea of
6:37
the of the loudness of the cards but it's not you know it's not scientific so
6:41
just bear that in mind
7:07
oh