Problem with test release v1.01 noise in power supply!!!

[grimm]

Very possible, must have been in too much of a hurry.

Jason

[pixelrush]

Not sure if this is helpful or not but from a review of the FSP Everest 800W:

The electrical performance of the FSP Everest 800W 80Plus unit is disappointing, given that this is marketed as a premium class product. The electrical efficiency of the unit is high and relatively stable, staying above 80% at all times as the 80Plus specifications suggest, but the voltage regulation circuit is quite disappointing. A voltage ripple of 50mV is satisfactory for the 12V lines, which ironically never even reached as high, but it is simply an unacceptable number when talking about the 3.3V/5V lines. Matter of fact, the 5V line falls way out of the ATX specifications which command a voltage ripple of 50mV maximum on the 3.3V/5V lines.

When it comes to thermal performance, the performance of the Everest is nothing more than acceptable for a premium unit. It could perform much better if larger or better heatsinks were installed, but it will remain cool enough and not at the expense of the acoustics performance. The Everest is not inaudible but it operates quietly under low to medium load and becomes clearly audible when the load increases beyond 60%, which behavior is typical for an 800W unit.

The FSP Everest 800W unit could certainly use several improvements before it can compete with other products in the premium class. If only for things like the missing surge protection varistor or the mediocre thermal performance we could say it is an expensive middle class product, but with the voltage regulation falling out of specification we find this unacceptable for any kind of power supply.

The build quality is good, the protection circuits work very well and the unit appears to be reliable, but only if you compare it against other mainstream products. Most other units around its price range will annihilate it in terms of design and performance. Unfortunately the only saving grace of the FSP Everest 800W 80Plus unit is its different appearance, which modders and enthusiasts who care about the aesthetics of their systems might like.

[Slide3D]

Not sure if this is helpful or not but from a review of the FSP Everest 800W...

This means the user must buy only premium class PSU for 200-300$ only that post-processing in octane does not cause noise?
if so send in the requirements it will be ... not good

[Zay]

Got the sound noise too from a Coolermaster Silent Pro 850W.

[jitendra]

Thanks for reporting the issue. We didn't experience the problem, but our office is not very silent either and I guess it occurs only on single GPU setups. We mainly use dual-GPU setups...

I'm not sure what is happening, but my guess is the following: The post-processing is run on the GPU of course, but it's a whole bunch of smaller kernel calls of constant complexity and execution time, which make the GPU alternate between 100% usage and idle very quickle, roughly in the frequency range you can hear. As a consequence, the power supply has to alternate the provided power accordingly, which means alternating high/low currents on the PSU. This may cause some expansion/contraction on some parts (capacitors for example, but I really don't know), which you can eventually hear.

If you use multiple GPUs they don't run synchronously, which means that the post-processing doesn't overlap exactly, which may reduce the strength of the power cycle.

Is it dangerous to the hardware? I don't know. Can we do something about it? I think so, but I don't know for sure.

Cheers,
Marcus

I did an interesting test.

I have gtx 580 and 590.
With gtx 580 for computing alone (there is no voice issue even with post processing)

with gtx 590 (msi with dual fan onto it) ... when I use it for rendering without any post processing there is a small continuous voice. (which is happening from the very first day I bought the card. I ignored it until now).
Now, If I enable the post-processing... this continuous voice is increasing and at the same time, there is a new similar voice which is getting on and off. Like happening at a regular interval...

Now, I am not sure if the voice is from GPU or PSU. Because, its difficult to find an exact source.

Thanks.

[matej]

Some more questions: Does the sound change, when you change the screen resolution? For example, does it have a lower frequency if you render larger images and higher a higher frequency, when the images become smaller? What happens when you change the number of glare rays?

Cheers,
Marcus

The bigger the resolution the louder is the noise and more glare rays = louder noise too.

Some graphic cards or PSUs make "squealing" noises when they are 100% overloaded, like for example running stress tests or games at hundreds FPS without vsync. Example of this noise. Strangely enough, GPU rendering is supposed to tax the GPU at max, but I never hear such noises when rendering with Octane. OTOH I do hear such squealing noises when I run simple opengl programs (like graphical engine test programs) that don't have mechanisms that limit PFS / refresh rate, so the card is rendering at 1000+ FPS.

Anyway, the sound in this case is a different: a more deep "scratching" sound that happens each time the viewport is updated, like there was a cat in my case trying to get out. For the time being I'm not using the postpro tools.

I'm still not sure if it's the PSU or card. I hope it's the PSU, cause mine is pretty old and mediocre and I plan to change it anyway. But the card is almost new, bought two months ago What do you guys suggest? I'm still running my rig with only one GPU, would using one additional for the monitor really help in this case?

[pixelrush]

>What do you guys suggest?

I suggest Otoy guys change the way the code works to stop the very rapid major fluctuations in load to the gpu because clearly this is overly demanding of components. The power supplies and other circuitry wont have been engineered for this type of use, or for instance the possibility that fan speeds might be continuously cycled if it is linked to a load like this has probably not been considered. Although it might be made more obvious if your PSU is not an especially good design and its voicing its general distress by chirping and burping this is a symptom of a condition all systems will be experiencing to some degree. This is not good for power regulation given the gpu are such a large part of the overall consumption when running full on as Octane does. I also wonder how the gpu chips themselves would physically respond to this intensive use. Perhaps some artificial aging/fatigue/resonance/hysteris effects?
Presently its a bit like driving your car in top gear and pumping madly up and down on the gas pedal and expecting the overall progress to average out the same way as if you smoothly accelerated through the gears.

[franbec]

The sound coincides with a momentary decrease of the load of the gpu, not overload. can be seen in gpuz utility.

[matej]

I suggest Otoy guys change the way the code works to stop the very rapid major fluctuations in load to the gpu because clearly this is overly demanding of components

I agree. Top-notch components are still stressed out the same as budget ones, they just don't physically show it. I'm no expert, but such load fluctuations most likely have some impact on the hardware longevity.

Its a good thing post-pro is not tied to the render process, so it can be turned on just at the end, before you save the render. But for animations and batch rendering you would still need to keep it on all the time.

[steveps3]

No noises here, phew!!