Sign in with
Sign up | Sign in
Your question

Measure Volts on the GTX 460 Hawk

Last response: in Graphics & Displays
Share
November 4, 2010 5:15:38 PM

I just measured the volts on the gpu with a multimeter and got these results:
idle 0.880
stock 1.043

then I would OC by +100 mV and press apply, it reads about ~1.152 (last decimals jump a bit)

however when running furmark, it goes up to ~1.260

my question is, how comes furmark adds another +100 mV? is this normal?

btw I'm using Afterburner
a c 592 U Graphics card
November 4, 2010 6:41:21 PM

Are you using the V Check points on the video card itself? These will give the most accurate readings. The MSI Hawk has several power phases, indicated by the red LED lights. I'm assuming that you will get different voltage readings based on the power state. The LED's should give you an indication of the power state, and I would think that the voltage would be the same for each level (i.e. if 4 lights are on versus if 5 lights are lit). So, I'm wondering if Furmark is pushing the card to it's highest state of power usage and if the voltage readings are accurate. If so, that would be normal.
m
0
l
a b U Graphics card
November 4, 2010 6:48:57 PM

With what matt said, the cards/bios downclock and down volt.
Mine according to gpu-z is at .987, but down volts to .887 at idle.
As soon as you do something or open gpu-z it temporarily jumps up back to .987

So I think your card was in a idle mode when you took your first reading. And when you added a amount it was over the stock voltage , not the stock idle (downvolt) voltage.

When things are under load, as with cpu V droop, voltages vary, in attempt to stay steady.
m
0
l
Related resources
a c 592 U Graphics card
November 4, 2010 8:11:02 PM

Another tidbit I read on a review:
"The Hawk uses a 7+1 phase PWM design which increases the maximum current capabilities to 120 amps. That's 30 more amps than the 4+1 phase design found on the reference model, and potentially increases the overclocking potential of the GTX 460 Hawk."
m
0
l
November 4, 2010 9:56:42 PM

No, it's not idle. Well I'll elaborate:

My gpu has 3 levels of clocking. Idle, 2d and 3d.
If we were to go according to gpu z, that would be:

idle: 0.8750 V (50mhz clocked)
2d: 0.9120 V (405mhz clocked)
3d: 1.0250 V (780mhz clocked)

But since gpuz is not accurate for my gpu, I decided to measure the volts through vcheck points. So I used the vcheck point for the gpu and my results were:

idle 0.880 (50mhz clocked)
3d stock 1.043 (780 mhz clocked)

I didn't bother to see what 2d would be since I don't really care about that when I want to go for an OC.

Now, here is the thing. When I press Apply Profile 1 on Afterburner, I get 1.043 V - that would be my stock voltage since that's when Afterburner shows the clock go to 780 mhz.

So that's cool, now I press Apply on Profile 2, which is my +100 mV profile, and I get 1.152 V - that's good I guess, makes sense, about +100 mV.

HOWEVER, I than ran Furmark while keeping the multimeter on the vcheck point and it went up to 1.260 V.

This is what I don't get, why did it just add another +100 mV? I thought load voltage = pressing Apply + X Profile, since for my very first result with Profile 1 I got 1.043 V, so it should be the same when pressing Apply + Profile 2 with an additional +100 mV (which it is because I got 1.152V)

So why was it reading 1.260 V when I ran a Furmark stress?

Sorry for the confusion, hope this is more clear.
m
0
l
a b U Graphics card
November 4, 2010 10:04:44 PM

Why don't you just tell use what the slider and # says in msi afterburner for your voltage.
It does not say +.100 , it says a voltage.
That might make things more clear.
Obviously its says one thing and your reading another.
Its not uncommon. Software sensors are calibrated en mass.
Here mine says I'm applying 1.087, and the only other way I have to verify that is gpu-z, which is reading the same sensor the same way.
m
0
l
November 4, 2010 10:12:21 PM

notty22 said:
Why don't you just tell use what the slider and # says in msi afterburner for your voltage.
It does not say +.100 , it says a voltage.

lol

It does. The slider can only go from -100 mV to +100 mV. What it does not say is the # in volts, hence why there are vcheck points to measure. Google some reviews about the gtx 460 hawk, you'll see what I mean.
m
0
l
a b U Graphics card
November 4, 2010 10:21:06 PM

Here are two ways you can get a software reading of that voltage, one is in msi afterburner and the other is gpuz.
Compare what those sensors are saying to your actual reading.
They are both reading .912 in that picture LOL
m
0
l
November 4, 2010 10:28:18 PM

Erm... here, you'll see what I mean:



1) No # volt slider, it's a -/+100 slider
2) Voltage monitoring is not supported either

This is why there's vcheck points.
m
0
l
a b U Graphics card
November 4, 2010 10:32:04 PM

Voltage monitoring is supported.

Do you think everyone is going to use a multimeter ?

Go in to the monitoring tab and check gpu voltage 1

The contact points are to compare to software sensors, for those that just want to.
m
0
l
November 4, 2010 10:37:57 PM

Try to find me a screenshot showing a hawk having voltage monitoring and get back at me.
m
0
l
a c 592 U Graphics card
November 4, 2010 10:42:59 PM

I still think when you run Furmark, you are going into an even higher PWM state than in your normal 3D mode. The card may be retaining some headroom for extreme 3D that you are only reaching with Furmark, which is definitely extreme versus playing a game. I also may have no idea what I'm talking about.
m
0
l
a b U Graphics card
November 4, 2010 10:50:14 PM

plum said:
Try to find me a screenshot showing a hawk having voltage monitoring and get back at me.

Post a screen shot of your gpu-z monitoring tab.
That will prove to YOU, that you have a software sensor on the card.
I know you do, sorry to inform you.


Here is a thread with gpu-z screenshots the op states at the bottom its of a gtx 460 hawk.
The voltage sensor readings is what the question is about.
http://forums.guru3d.com/showthread.php?p=3752530
m
0
l
a b U Graphics card
November 4, 2010 11:01:22 PM

plum said:
http://forums.guru3d.com/showpost.php?p=3749461&postcou...

Unwinder, MSI Afterburner developper and creator of RivaTuner
http://forums.guru3d.com/member.php?u=30019

GPU Z is wrong and not accurate for this GPU, as I said 5+ posts ago.
and btw GPU Z =/= Afterburner, hence why you'll never find a screenshot with Afterburner showing voltage reading for the hawk.


You still need to tell us what gpu-z is reading for a voltage when you quote a voltage that you read from your multi-meter.
It does us NO GOOD for you to say , I added .1 ?

How do you know your multi-meter is correct or calibrated ?

edit:

What did you expect it to read ? When you got 1.260.
edit: You still need to compare it , to what the software sensor is reporting.
m
0
l
November 4, 2010 11:02:27 PM

Read my second post
m
0
l
November 4, 2010 11:12:07 PM

First, I would really appreciate if you took some time to read my posts instead of rushing at answering the posts. Don't mean to be rude but I asked you to google it earlier. If you did, you would have saved yourself quite some time there.

Secondly, since I read 1.043 V for stock, I was expecting to read about +100 mV on top of that (which I did, I got 1.152 V) but then as I ran Furmark, I saw it increase to 1.260 V and this is the only reason I'm here. To find out why it added an additional +100 mV.

About the sensor, even if I add +50mV or +100 mV, doesn't matter, it will not show higher than 1.0250 V because GPU Z is not accurate as I said before. It's not reading values properly. If you read Unwinder's post, the +100 mV OC comes from an additional voltage controller chip so it's normal it doesn't recognize it.
m
0
l
a b U Graphics card
November 4, 2010 11:19:55 PM

plum said:
First, I would really appreciate if you took some time to read my posts instead of rushing at answering the posts. Don't mean to be rude but I asked you to google it earlier. If you did, you would have saved yourself quite some time there.

Secondly, since I read 1.043 V for stock, I was expecting to read about +100 mV on top of that (which I did, I got 1.152 V) but then as I ran Furmark, I saw it increase to 1.260 V and this is the only reason I'm here. To find out why it added an additional +100 mV.

About the sensor, even if I add +50mV or +100 mV, doesn't matter, it will not show higher than 1.0250 V because GPU Z is not accurate as I said before. It's not reading values properly. If you read Unwinder's post, the +100 mV OC comes from an additional voltage controller chip so it's normal it doesn't recognize it.

Don't mean to be rude, but you don't even know what your asking. :) 
m
0
l
a b U Graphics card
November 4, 2010 11:20:50 PM

What happens if you run a normal game or benchmark, like crysis or 3D Mark? Does that go to 1150 or 1250 mV. Interesting issue btw.
m
0
l
November 4, 2010 11:28:19 PM

notty22 said:
Don't mean to be rude, but you don't even know what your asking. :) 

Or maybe full sentences are difficult to you. Here, I'll make it simple:

MULTIMETER READING AT IDLE: 0.8750 V
MULTIMETER READING AT STOCK: 1.043 V
MULTIMETER READING OC (+100mV): 1.152
MULTIMETER READING OC + FURMARK RUN: 1.260

QUESTION: Why did it add another +100 mV?

If you don't get that one, just give up.

EXT64 said:
What happens if you run a normal game or benchmark, like crysis or 3D Mark? Does that go to 1150 or 1250 mV. Interesting issue btw.

I'll try that later, don't have the multimeter with me at the moment so can't test. Personally I thought overvolting a gpu worked like overvolting a cpu, so there's no reason volts should be exceeding 1.152 but I don't know, I'm not too experienced.
m
0
l
a b U Graphics card
November 4, 2010 11:36:16 PM

I would be concerned that since the author cannot be sure of behavior of the sensor circuit, how can the behavior of adding offset voltage to any degree of certainty be guaranteed.
Meaning the 1.2 you that you read may be accurate, and I would be careful not to over volt your gpu.
We see this behavior from cpu's when overclocking, and running prime 95. Disabling or enabling Vdroop modifies the behavior, but the jury is out, what is really best.
The voltage circuit responds to the load. Do what ext mentioned, run a cyrsis demo benchmark and take a full load reading. One more reason not to run a synthetic like Furmark, imho.

edit: Also even O/C the voltage drops when load is off. That is clear from temp sensor readings.
m
0
l
November 4, 2010 11:44:55 PM

there should be a setting of force constant voltage under settings in msi after burner. If there isn't go into your cfg file located in the afterburner directory and change it from 0 to 1
m
0
l
November 5, 2010 12:07:52 AM

I don't think Force Constant Voltage will do much since that +100 mV isn't changing. But well, can't test now either way.

What I can say for now is that when I run 3DMark Vantage with my OC profile (+100 mV/900mhz core), I get 58 C max. As for when I run Furmark (windowed 1920 x 1080), max it hits is about 71c. In games, never seen it go past 55c I think.

It's not a fair comparaison though since Furmark is much more constant on the GPU when 3D Mark Vantage just throws load at it somewhat randomly.

I'm just trying to understand why it does this, I'll try contacting MSI I guess. This card seems to be in it's own league.
m
0
l
!