WEBVTT

00:00:07.160 --> 00:00:16.600
the Dark Art of overclocking so the GTX 680 is to my understanding fairly over

00:00:13.839 --> 00:00:21.160
over overclockable but I haven't seen any numbers yet so I'm going to be using

00:00:19.000 --> 00:00:26.679
a retail sample so this is probably going to be the only review that's up on

00:00:23.279 --> 00:00:29.359
launch day of the GTX 680 using a retail

00:00:26.679 --> 00:00:34.559
non handpicked board and uh we're going to see how far it goes and how well it

00:00:31.880 --> 00:00:40.160
can compare in terms of performance to the highly overclockable

00:00:37.120 --> 00:00:43.039
hd7970 from AMD so I've already got a

00:00:40.160 --> 00:00:49.760
lot of my numbers prepared here uh I have my stock speed 680 results I have

00:00:46.480 --> 00:00:52.399
my doc speed 7970 results then I have my

00:00:49.760 --> 00:00:57.640
680 OC results still to fill in because I have not yet begun overclocking and

00:00:54.760 --> 00:01:04.239
then I've got my 7970 OC results which are based on uh basically maxing out the

00:01:02.079 --> 00:01:11.560
overclocking within the Catalyst Control Center so we were running at uh oh shoot

00:01:07.720 --> 00:01:15.680
what was it just right 1.125 GHz on the

00:01:11.560 --> 00:01:18.880
processor and then uh 6.3 GHz on the

00:01:15.680 --> 00:01:22.360
memory so this is one freaking fast card

00:01:18.880 --> 00:01:24.960
and I am very very excited to see what

00:01:22.360 --> 00:01:30.439
we can get out of its main competitor the GTX

00:01:27.880 --> 00:01:35.960
680 here you oh yeah I'll be having a right so guys I'll be having a look at

00:01:33.399 --> 00:01:39.280
not only the performance but also the difference in power consumption from

00:01:37.520 --> 00:01:45.280
overclocking these cards so you can see there was uh definitely an increase an

00:01:42.439 --> 00:01:49.119
increase increase uh from the stock power consumption to the overclock power

00:01:47.000 --> 00:01:52.920
consumption I'll also be looking at how temperatures are affected because I've

00:01:50.960 --> 00:01:57.360
got all of my Baseline readings done so stay tuned guys we are going to be using

00:01:55.040 --> 00:02:00.560
EVGA Precision which at launch is the only overclocking application to support

00:01:59.439 --> 00:02:04.640
the 6 I'm more of an MSI After Burner guy

00:02:02.560 --> 00:02:09.239
myself I love their oh yeah see it's not going to work um I love their simple

00:02:06.719 --> 00:02:12.560
interface and I'm just very used to it I like the way they do their graphs but

00:02:11.039 --> 00:02:17.480
since Precision is what we're using Precision is what I will have to figure

00:02:14.640 --> 00:02:20.640
out how to how to work so stay tuned guys let's see how this goes well I

00:02:19.080 --> 00:02:25.160
guess we better do a basic rundown of precision X first so number one is that

00:02:23.000 --> 00:02:31.400
it has integrated voltage adjustments for the GPU so you can actually slide

00:02:28.360 --> 00:02:32.840
these to where you want them to go

00:02:31.400 --> 00:02:37.319
you can press set I'm not going to be changing any voltages I did not change

00:02:34.640 --> 00:02:41.519
voltages on the 7970 so I will not be adjusting voltages on the GTX 680 for

00:02:39.840 --> 00:02:45.640
the sake of this basic overclocking guide you guys want to turn up voltages

00:02:43.200 --> 00:02:49.400
on your CPUs be careful in terms of your cooling be careful not to fry your GPU

00:02:47.720 --> 00:02:54.360
because it is a very expensive mistake to make and while yes turning up

00:02:51.920 --> 00:02:57.519
voltages may not affect it in any way this is a new graphics card on a new

00:02:56.000 --> 00:03:00.400
manufacturing process there are no guarantees that it won't break your card

00:02:59.200 --> 00:03:06.000
whether it's um today or tomorrow or 2 months from

00:03:03.760 --> 00:03:09.680
now so maybe let someone else be the guinea pig Okay so we've got a

00:03:07.640 --> 00:03:13.640
performance log frame rate target which is this kind of neat thing right here we

00:03:11.159 --> 00:03:19.159
can enable this to set a Target FPS which is interesting because it seems to

00:03:16.040 --> 00:03:21.959
be related to the fact that the GTX 680

00:03:19.159 --> 00:03:29.159
dynamically scales its power consumption which is I don't know probably here

00:03:25.239 --> 00:03:30.760
somewhere okay don't see it uh but it

00:03:29.159 --> 00:03:36.720
scales it power consumption with its GPU clocks in order

00:03:33.920 --> 00:03:41.239
to achieve the optimal heat output to well to stay within the uh the power

00:03:39.159 --> 00:03:44.760
Target and heat Target that it was designed for so in order to get any

00:03:43.400 --> 00:03:49.040
serious overclocking done you're probably going to want to adjust your power Target so I'm going to go ahead

00:03:47.200 --> 00:03:53.640
and tell it pretty much I don't care what kind of power Target you reach as

00:03:51.400 --> 00:03:56.959
long as we're not turning up any voltage okay so that's one thing we've also got

00:03:55.360 --> 00:04:03.239
monitoring of temperature monitoring of fan speed this is cool you can use this

00:04:00.200 --> 00:04:04.959
software to set up a fan curve I love

00:04:03.239 --> 00:04:08.879
fan curves I think they're amazing I think giving users the control to say

00:04:06.760 --> 00:04:13.439
okay yeah I'm comfortable you know with really low fan speeds up to 70 degre 70°

00:04:11.920 --> 00:04:17.079
and then I want it to ramp right up to 100% get it cooled back down if you want

00:04:15.480 --> 00:04:23.840
it to work that way you can totally make it work that way so I'm going to turn that off adjusting voltages aha GPU

00:04:21.919 --> 00:04:28.400
clock offset so let's talk about GPU boost so

00:04:25.400 --> 00:04:31.720
the basic clock here so that's at about

00:04:28.400 --> 00:04:34.280
1 GHz 1,00 6 MHz or something like that

00:04:31.720 --> 00:04:39.960
and then the Boost clock at 1056 or something along those lines those are

00:04:36.720 --> 00:04:44.039
both clock speeds of your GPU

00:04:39.960 --> 00:04:46.880
so you can you can set an offset so what

00:04:44.039 --> 00:04:53.919
that means is the offset until it detects that it has extra TDP room extra

00:04:49.919 --> 00:04:56.919
power room in order to uh boost itself

00:04:53.919 --> 00:04:59.880
up from the offset that you've set so

00:04:56.919 --> 00:05:03.880
for example if I set the offset at 100 MHz and I know this GPU is probably

00:05:02.120 --> 00:05:09.240
going to be good for 100 MHz otherwise I wouldn't set that you should probably only go up in increments of like you

00:05:07.000 --> 00:05:14.720
know 15 to 25 go ahead and apply that you can see those dials are hopefully

00:05:12.080 --> 00:05:20.720
going to move in some way or unless that's just showing the uh the base

00:05:16.680 --> 00:05:22.840
default one okay cool

00:05:20.720 --> 00:05:26.560
interesting maybe we'll just get started and see how this

00:05:24.520 --> 00:05:32.319
goes okay but either way that's how offsets are supposed to work so as it

00:05:28.759 --> 00:05:35.520
needs more power it can boost itself up

00:05:32.319 --> 00:05:37.440
and uh you're just setting up what the

00:05:35.520 --> 00:05:41.600
base GPU clock will be and it'll Boost from there that's sort of the point okay

00:05:39.880 --> 00:05:45.440
guys so I see how this works now while I don't recommend using furmark for any

00:05:43.440 --> 00:05:50.960
kind of extended period of time it's been handy for me for this because it's

00:05:47.160 --> 00:05:53.000
allowed me to uh examine how the here

00:05:50.960 --> 00:05:57.000
with the performance log I can see the percentage of power that the board is

00:05:54.479 --> 00:06:01.080
using please Focus there we go um compared to see we've hit 132 which is

00:05:59.520 --> 00:06:06.080
the the maximum I was able to set for the power Target compared to what the

00:06:03.520 --> 00:06:12.240
GPU clock is actually running at so 11:45 is what we seem to be peaking at

00:06:08.720 --> 00:06:14.599
which is actually ah 1172 is our 100%

00:06:12.240 --> 00:06:22.840
Peak so that's uh we should be somewhere in the neighborhood of 106 + 153 so what

00:06:17.599 --> 00:06:25.000
is that uh or 1,6 + 153 so

00:06:22.840 --> 00:06:28.560
1161 something like that so yeah you can see we're we're popping up there once in

00:06:27.000 --> 00:06:33.840
a while but not staying there consistently so I'll have to look at how

00:06:31.039 --> 00:06:37.840
performance is actually affected but it looks like the way the overclocking

00:06:35.400 --> 00:06:43.599
works at least with Precision is that we are pretty much just overclocking the

00:06:40.160 --> 00:06:46.039
Boost clock and I don't see a way to

00:06:43.599 --> 00:06:49.960
adjust the base clock although because yeah obviously I can't find uh any

00:06:48.280 --> 00:06:54.039
manual or instructions for this right now but if I it turns out that I'm wrong

00:06:52.199 --> 00:06:58.080
then I apologize and guys please leave a comment and let me know but there we

00:06:56.759 --> 00:07:04.080
have it so let's see what kind of a performance Improvement we get by

00:07:00.000 --> 00:07:07.479
increasing the turbo GPU boost clock

00:07:04.080 --> 00:07:09.560
offset on the GTX 680 well I can see

00:07:07.479 --> 00:07:14.120
this causing some confusion so due to the way GPU boosts uh Works where it

00:07:12.000 --> 00:07:20.639
turns up the GPU clock if it has the available power headro um if you turn

00:07:18.000 --> 00:07:25.039
the overclocking the offset up past where it's capable of throttling to it

00:07:22.280 --> 00:07:30.039
looks like performance does not continue to scale so you can see we have a score

00:07:27.919 --> 00:07:35.560
of 3167 so I'm I'm going to go back to my spreadsheet now when I was at 153

00:07:32.560 --> 00:07:38.599
offset we were able to score

00:07:35.560 --> 00:07:41.000
3488 on the extreme preset in 3D Mark 11

00:07:38.599 --> 00:07:46.919
however what you guys will see is 3172 so this is almost exactly the same

00:07:44.720 --> 00:07:52.080
as what I got with my stock configuration on the GTX 680 so it looks

00:07:49.199 --> 00:07:56.960
like once we turn it up past the threshold it doesn't necessarily scale

00:07:54.919 --> 00:08:01.599
100% correctly now this is an early beta version of precision X and like I said I

00:08:00.000 --> 00:08:04.240
have no instructions so I'm not even 100% sure if I'm doing it right but I'm

00:08:03.039 --> 00:08:08.199
going to put it back at the setting I had before and run it one more time to

00:08:06.080 --> 00:08:15.720
see if I can get my overclocked setting dialed in here so now we see an even

00:08:10.680 --> 00:08:17.599
higher score so this is at uh plus 153

00:08:15.720 --> 00:08:23.240
so you can see we're getting about 1250 MHz when the benchmarks are actually

00:08:19.759 --> 00:08:27.400
running and we're seeing about a 100ish

00:08:23.240 --> 00:08:29.520
102 119 126 so we're seeing pretty close

00:08:27.400 --> 00:08:33.039
to the power threshold that uh that we've said over here in some strenuous

00:08:31.599 --> 00:08:39.440
situations so let's do some real world benchmarking and find out how this

00:08:36.159 --> 00:08:40.800
um GPU boost overclocking actually works

00:08:39.440 --> 00:08:46.000
now I would really prefer if they made some tweaks to the to the interface here

00:08:43.800 --> 00:08:49.080
allowing you to boost up this base clock rather than only the Boost clock so that

00:08:47.720 --> 00:08:53.240
you have at least some idea of a baseline that you're running at uh

00:08:51.040 --> 00:08:56.480
without it doing any like funky scaling back and forth but maybe that's

00:08:54.920 --> 00:09:01.160
something we'll see in a future revision hard to say let's check out the real

00:08:58.839 --> 00:09:05.519
world performance perance uh so this is weird check this out guys so with the

00:09:02.839 --> 00:09:09.440
overclocked settings running on the GTX 680 I'm not getting that weird

00:09:07.560 --> 00:09:14.360
flickering that I was getting in Skyrim with it at stock speed so

00:09:11.800 --> 00:09:17.600
um yeah I don't know what to say it seems to have kind of healed itself

00:09:15.920 --> 00:09:23.200
anyway I'm just finishing up my overclocked settings benchmarking and

00:09:20.440 --> 00:09:28.600
I'll be done shortly so in conclusion even after overclocking both of the

00:09:25.560 --> 00:09:30.880
cards the GTX 680 is still the winner

00:09:28.600 --> 00:09:33.959
and once I get the hang of overclocking it it might actually be a winner by a

00:09:32.680 --> 00:09:38.040
more substantial margin so I have a game running in the background so this shows

00:09:36.040 --> 00:09:41.560
what my final overclock ended up being 145 this was stable in all of my gains

00:09:40.079 --> 00:09:45.800
you can see the fluctuations in the clock speed as well as the fluctuations

00:09:43.800 --> 00:09:52.120
in the peak power consumption according to the uh Power Target up here so um I

00:09:50.399 --> 00:09:55.560
mean yeah I think there's probably still some refinement to be done to the

00:09:53.519 --> 00:09:59.680
Precision X tool or whatever other tools end up becoming popular with these cards

00:09:57.959 --> 00:10:07.320
but let's talk results for a little bit here so idle power is up about 10 watts

00:10:03.600 --> 00:10:10.040
versus the stock clocked card and then

00:10:07.320 --> 00:10:17.800
um idle temps are around the same 3D Mark 11 uh we saw up to

00:10:13.600 --> 00:10:20.880
3572 3D marks um although at the same

00:10:17.800 --> 00:10:23.200
clock speeds we saw scores as about 100

00:10:20.880 --> 00:10:27.079
marks lower even so there seems to be some inconsistency in the scores

00:10:24.880 --> 00:10:30.920
although we saw fairly consistent improvements across the board at

00:10:29.360 --> 00:10:37.519
actually what I'm going to do is I'm going to um hide all this other stuff so

00:10:35.560 --> 00:10:45.959
we saw fairly consistent improvements across the board so here is the aha so

00:10:41.680 --> 00:10:50.360
we saw about a 7 8% Improvement in

00:10:45.959 --> 00:10:52.920
Battlefield 3 then in crisis 2 we saw

00:10:50.360 --> 00:10:57.480
yeah very little there so about a 4% Improvement 3 and 1/2% Improvement

00:10:54.839 --> 00:11:02.600
something along those lines Witcher 2 we see about a you know % Improvement

00:11:00.519 --> 00:11:06.800
something along those lines or like six or whatever that works out to uh power

00:11:05.200 --> 00:11:09.959
consumption is a little bit higher across the board using the same

00:11:08.279 --> 00:11:14.760
methodology I was using in my original performance review to measure that um

00:11:12.720 --> 00:11:19.120
see this is weird Dirt 3 actually went down a little bit in terms of average

00:11:17.440 --> 00:11:23.760
frame rate although power consumption went up substantially Batman Arkham City

00:11:21.920 --> 00:11:27.480
is up significantly I'm probably going to have to rerun these numbers and I'll

00:11:25.120 --> 00:11:31.560
do an annotation on my uh original review video if there's anything foul

00:11:30.160 --> 00:11:36.360
about them because you can see that power consumption number is low that could also have to do with GPU boost

00:11:34.680 --> 00:11:43.639
scaling back the power and scaling back the performance Skyrim we see uh NVIDIA

00:11:40.440 --> 00:11:45.880
is definitely the king here and we see a

00:11:43.639 --> 00:11:49.200
very moderate performance Improvement in Skyrim that is what I would consider to

00:11:47.600 --> 00:11:53.079
be within the margin of error for this test so we didn't see much of a

00:11:51.000 --> 00:11:59.600
performance boost there however it is still good enough to beat the radon

00:11:56.079 --> 00:12:02.160
7970 overclocked so there you have it it

00:11:59.600 --> 00:12:05.519
guys my GTX 680 let's call this preliminary overclocking video because I

00:12:04.160 --> 00:12:08.800
don't think we can call these results final but this is what I've got ready

00:12:07.480 --> 00:12:12.200
for launch and hopefully I'll be able to do a followup at some point let you guys

00:12:10.720 --> 00:12:15.720
know if there are any changes to my overclocking methodology with this card

00:12:14.160 --> 00:12:19.959
don't forget to subscribe to L Tech tips for more unboxings reviews and other

00:12:17.639 --> 00:12:23.440
computer computer videos and don't forget how dedicated I am to you guys

00:12:22.120 --> 00:12:28.120
here here I am finishing up my benchmarks and I'm probably not done for

00:12:25.320 --> 00:12:34.079
the night yet so enjoy subscribe for more sleep deprived nights with me