PDA

View Full Version : Interesting hardware issue.



wizard
04-19-2008, 02:58 AM
Now I don't know if this is exactly the best place to ask this, but given the number of developers here I figured that someone here might have encountered this problem also.

Currently I have two distinct render paths both of which use the same materials and are based on the same principles (and using the same method to do primitive rendering). The problem is that when rendering with the other method the graphics card starts making an odd noise. The noise comes only during rendering and if I do something to stop the rendering (in my case grab the window title bar, for instance) the noise stops at that instance only to continue once rendering resumes. I suppose the graphics card circuitry might be making the noise, since the fan speed remains unchanged as does the CPU fan speed (and there is no way that a fan would react so quickly).

So what do you think? My dev graphics card going nuts or a software issue? Can anyone think of anything that a piece of software could do that would result in this happening?

yooyo
04-19-2008, 04:03 AM
Nice effect... if you investigate more about that and learn how to control it, you can make mp3 player on graphics card. :)

It could be anything.. Sometimes I can hear some buzz from machine when I move mouse pointer around. There is no way to circuits produce that noise.

Zengar
04-19-2008, 04:39 AM
It is the power converter. My old card did the same. My LCD does it too :)

remdul
04-19-2008, 07:31 AM
Aha, I noticed the same on the NV7300GT, which I thought was odd, as it didn't have a fan at all. :)

V-man
04-19-2008, 08:51 AM
I've had the same issue with my graphics card and it depends on what is being rendered. Perhaps if there isn't much being rendered, the FPS is way too high and perhaps somehow it caused the GPU to consume more power.

Also, I've had the same effect with the motherboard. Under Linux, there is this high pitch sound. On WinXP, I can only hear the high pitch sound during boot up.

wizard
04-19-2008, 08:59 AM
I seem to have worked around it by rewriting the render path... It's odd because the rendering code was also very slow when the noise was still present. Originally I was also able to lessen the noise by removing pretty much everything from the lighting pass shader (which I didn't really regard as a solution :)).

The power converter... Might be. It's just odd that I don't get the noise from any other heavy duty 3D app that I tested... I even tried Crysis "to no avail".

Jan
04-19-2008, 09:26 AM
Crysis DIDN'T work!? Interesting.

I got the impression Crysis is meant to be a hardware-killer, that' happens to have great graphics and something you could call "gameplay"...

Seth Hoffert
04-19-2008, 11:52 AM
Hmm, I've noticed this on my 8800 too in Linux, especially when I disable vsync and am rendering close to nothing (high framerates - ~4k fps).

Eosie
04-19-2008, 03:24 PM
Same issue here (on ATI). It seems to have something to do with very high framerates, no matter the scene complexity.

predaeus
04-20-2008, 02:38 AM
Same here on gf6800.

wizard
04-20-2008, 07:49 AM
Really interresting, because for me it happened on very low framerates and was lessened when I further reduced the workload of the GPU :) ...only to vanish after a rewrite (that effectively did the same thing as the orignal piece of code). Why would the GPU require more juice when running barely anything, anyway?

OT about Crysis: What is also interresting is that when I tested on Crysis I saw that the game was using only one core. That's quite bad really. It seems to have game code, scripting, physics and graphics all running on the same thread. Correct me if I'm wrong but this was my observation...

Ilian Dinev
04-21-2008, 01:48 AM
It may be the power-supply getting overloaded and not filtering properly.
Crysis seems to not produce that noise only because the noise there will be very low-pitched or high-pitched for your ears to hear.
I'm not really sure why this problem causes mechanical fluctuations... I thought piezoelectrical effect is strict about the material.

Nothing can beat the noise produced when I'm flashing a PIC microcontroller at low-speed with a possibly too low-power power-adapter. (I haven't tried with a better adapter yet, and can't now)

P.S. it could also be sound, created from quick thermal expanding and contraction of materials?

V-man
04-21-2008, 08:04 AM
It has nothing to do with thermal expansion. It's just that there is some high frequency current passing that is causing something to vibrate at high speed. Putting some silicone or something like that on that particular part would silence it.
That's what they do with the flyback transformer inside TVs (the CRT kind)

rockeye
04-22-2008, 12:24 AM
Same effect on a Quadro FX1700.

In my case, the noise happens even in GDI under XP, on every refresh events.

Nicolai de Haan Brøgger
04-22-2008, 04:51 AM
I too had some coil whine on GF7800GT but much less on GF8800GTS512. Like V-man says, it is caused by high-frequency vibration. You can try and poke the components with a platic pen when it's whining to find out which one to glue :)

wizard
04-22-2008, 01:57 PM
Sounds like a sound explanation :) Still can't figure out what caused it. Maybe the ways of the driver/GPU are too complicated to comprehend.