Why no talk about X850 yet???

[Rys]

I'd love to know how they measured 20dBA without

a) utter silence from everything in the system and room they tested in, bar the cooler on the graphics card

b) multi thousand dollar measuring equipment

As you can see, it costs $100+ just for a meter that'll measure accurately < 30dBA most of the time.

 

[aaronspink]

Oh, doh.

It primarily means that you'll have to measure from a much closer distance than the standard 1m.

Aaron Spink
speaking for myself inc.

 

[anaqer]

I'd love to know how they measured 20dBA without

a) utter silence from everything in the system and room they tested in, bar the cooler on the graphics card

b) multi thousand dollar measuring equipment

As you can see, it costs $100+ just for a meter that'll measure accurately < 30dBA most of the time.

I was told that the values have been obtained from a 16mm distance relative to the center of the fan, with all the other fans in the system shut off for 30 seconds - the highest recorded value during this time is taken as a baseline. Then they subtract 35.9 dB based on a calculation the tester can't explain off the top of his head (...) but this should give the same results as measuring from 1m.

BTW, looks like I might be wrong about the 6dB figure too - according to this site, it's more like 10dB...

 

[Pete]

My mistake, sorry for the power/noise mixup/misinformation.

 

[Entropy]

Oh, doh.

It primarily means that you'll have to measure from a much closer distance than the standard 1m.

Aaron Spink
speaking for myself inc.

True, but there are difficulties with moving closer as well.
Since you want to test the card under load, moving the mic closer to the gfx-card also moves it closer to other noise sources in the system. While you may get a high enough reading to register, it doesn't help with isolating the contribution from the gfx card unless you go very close, but that emphasizes the most important difficulty with moving closer which is directionality. While many HSF setups can be assumed to be the most noisy from the side of the chip/sink/fan, some new and interesting designs have this area shrouded and evict the air out the back of the case. This would tend to lower the noise measured from the side, but increase it from the back. There are other problems as well. So in order to reduce directionality concerns, as well as better mimic actual use, I'd prefer if the measurement was made outside a closed case at some semi-reasonable distance. The standard 1 m would be OK I guess. However, this would bring its own can of worms
* What cabinet is used? How open to air carried noise, how prone to resonances?
* How is it placed in the measurement environment? Open, or with a reflecting/dampening wall behind it?
Plus of course the fact that this requires a more sensitive SPL-meter, (as well as possibly lower ambient noise).

While the sub-20db measuring rigs are probably a bit too expensive for Dave's tastes unless he can find a second hand B&K, it might be worth it to look around for something more sensitive than this 40+ dB job, as it would make measuring a fair bit easier.

Still, it's damn fun and interesting to have an SPL meter and measure stuff, so if he can't return this device, he might as well get some mileage out of it by getting aquainted with the ins and outs of SPL measuring. Take readings in various rooms, the car, office environments, the neighbours dog, whatever. He can try various ways of testing - what this device can let him do, and where it fails him. Get some hands on experience with resonance and standing waves. Think through the issues. Fire off a mail to the guys at silentpcreview, and ask them for a bit of advice. Try some new stuff, and mull it over some more. Then he can start thinking about how or if he would really want to approach the problem of getting good information to his readers, and what new equipment that would require, if indeed any.

I'm really glad to see you taking an interest Dave, and I'm sure you'll find it very interesting.
Plus I'm sure you'll enjoy having a near silent test rig.

 

[Dave Baumann]

We'll see how it goes, but at the moment, I'm figuring on taking system noise tests, i.e. place the meter at the botton of the case and take readings with each board but keeping the rest of the system the same. A graphics card can make more "noise" (and heat) than from its own fan alone, i.e. if its dumping a load of vertex ops (or has no vertex shaders) to the CPU that can atually put a bigger load on the CPU, meaning it generates more heat which in turn means the CPU fan would need to spin faster.

Yes, this won't give use the definite "noise" for the graphics fan alone for the graphics fan alone, but we hopefully should get a baseline for the test system being used in a particular test. Of course, PCIe and AGP systems may not be directly comparable as the AGP / PCIe slots may not be in the same position on each motherboard. We aren't going to be able to directly compare power either, though, since although we can now get close with the CPU and memory staying the same with the Athlon platform, the motherboards will never have the same power draw.

 

[Entropy]

We'll see how it goes, but at the moment, I'm figuring on taking system noise tests, i.e. place the meter at the botton of the case and take readings with each board but keeping the rest of the system the same. A graphics card can make more "noise" (and heat) than from its own fan alone, i.e. if its dumping a load of vertex ops (or has no vertex shaders) to the CPU that can atually put a bigger load on the CPU, meaning it generates more heat which in turn means the CPU fan would need to spin faster.

The big issue here is probably PSU noise. Not only does the gfx card load the power supply with its own draw, we also have PSU inefficiency to deal with, with most PSUs being in the region of 70% efficient over most of their range. For instance if you add a passively cooled 50W graphics card, you still add load to the PSU, and you still have to evict some 75W of extra heat. All PSUs, even those with well designed fan control, will increase their fan speed and noise under those circumstances. I see that you've considered that, since you ordered the power meter as well.

Yes, this won't give use the definite "noise" for the graphics fan alone for the graphics fan alone, but we hopefully should get a baseline for the test system being used in a particular test. Of course, PCIe and AGP systems may not be directly comparable as the AGP / PCIe slots may not be in the same position on each motherboard. We aren't going to be able to directly compare power either, though, since although we can now get close with the CPU and memory staying the same with the Athlon platform, the motherboards will never have the same power draw.

True, and a drawback to testing the whole system. Measuring the card in isolation is only half the story, but measuring the system brings a whole new set of variables into play. (BTW, it is interesting how the same issues, Component vs. System, crop up in almost all testing even regardless of whether it's physical or something like benchmarking!)
Measuring both is the straightforward answer. But it's more work, and does not necessarily bring all that much more enlightenment.
However, if you configure the rest of your system so as to be as quiet as possible, you will to some extent be able to catch both by making system wide measurements, as the noise then will be dominated by the gfx card. In the case where the gfx card is cooled passively (or extremely quietly) you will still be able to measure its secondary effects on the rest of the system. In terms of practicality, this is the path I would first try to pursue myself. Good nearfield measurements are tricky, and it can be difficult/impossible to make a relevant nearfield comparison between say, an X800XT and an X800XL, where the cooling apparatus is so different in terms of directional properties.
While you could place the probe inside the (opened?) case, I would suggest you try to place the the mic on a stand outside the closed cabinet, with the case standing in the middle of a quiet room, on top of something that dampens vibration. A thick carpet will do in a pinch. The mic should probably be placed directly behind the cabinet, preferably a fair distance away, a meter or so.
This is where the not-so-sensitive SPL-meter might well be a limitation. Try it. You could creep closer if necessary.

By all means, try placing it in the case as well to see how that plays out, making experiments with different placements should be fairly quick. Might be a good idea to put it on some foam rubber though.

Fun stuff.

 

[Dave Baumann]

Not only does the gfx card load the power supply with its own draw, we also have PSU inefficiency to deal with, with most PSUs being in the region of 70% efficient over most of their range. For instance if you add a passively cooled 50W graphics card, you still add load to the PSU, and you still have to evict some 75W of extra heat. All PSUs, even those with well designed fan control, will increase their fan speed and noise under those circumstances. I see that you've considered that, since you ordered the power meter as well.

Yeah, precisley, its not much cop having a passive cooler if it just means you end up generating more noise elsewhere to cope with the heat, alternatively a slightly noiser graphics fan that exhausts out the back may actually produce less noise overall as it is assisting in lowering the case temps.

While you could place the probe inside the (opened?) case, I would suggest you try to place the the mic on a stand outside the closed cabinet, with the case standing in the middle of a quiet room, on top of something that dampens vibration.

Unfortunatly, not particularly practical - if you ever saw the test room you know why!

 

[Entropy]

While you could place the probe inside the (opened?) case, I would suggest you try to place the the mic on a stand outside the closed cabinet, with the case standing in the middle of a quiet room, on top of something that dampens vibration.

Unfortunatly, not particularly practical - if you ever saw the test room you know why!

Ah.
This is where the flowers to your wife pay off.
The quietest room is often the bedroom, and it's only a 5s lift away.
Set up, measure idle, fire up a suitable demo, let stand, measure full draw, shut down, carry back, kiss indulgently smiling wife who think you're a swell guy, bringing her flowers 'n all. Simple.