POLL: What should be the reference noise level at which graphics cards should be tested?

What should be the reference noise level at which graphics cards should be tested?

  • 35 dB(A)

    Votes: 4 14.3%
  • 40 dB(A)

    Votes: 10 35.7%
  • 45 dB(A)

    Votes: 4 14.3%
  • 50 dB(A)

    Votes: 2 7.1%
  • 55 dB(A)

    Votes: 1 3.6%
  • I don't know - it's your job to figure it out!

    Votes: 7 25.0%

  • Total voters
    28

mechan

Distinguished
Jan 25, 2009
90
0
18,640
Hello Tom's Hardware community. As you may have seen in The Myths Of Graphics Card Performance: Debunked, Part 1, we've introduced a new type of test that compares graphics cards performance at a set noise level. We arbitrarily picked 40 dB(A) as a starting point. We've received quite a bit of feedback as different users feeling different levels were more/less appropriate. Now we'd like your input into what noise level should become the actual "reference" level. Answering "I don't know - it's your job to figure it out!" is fine!

As always, we appreciate your input!

- Filippo Scognamiglio
 

mechan

Distinguished
Jan 25, 2009
90
0
18,640


While in theory that might seem a good idea, in practice:
A) Different cards have different thermal tolerances
B) It is really, really, hard to set a temperature target as cards will vary fan speeds and throttle core frequencies in ways that makes temperatures bounce around quite a bit more often than not
 

A temperature limit based on what the weakest card can take makes sense, though I get its a harder test to perform between cards, much easier on cards of the same brand and model.

http://www.tomshardware.com/reviews/geforce-gtx-670-test-review,3217-13.html

The 'cooling to 70C' bit of this provides pretty definitive proof that the Asus card is best. With any other sort of test you have pretty much no idea because the temperatures and noise are all over the place.
 
You have to pick a lower than average noise level in order for the test to worth it. If you pick a high value, you will be staying within most cards' normal operating noise thresholds, allowing them to run at full throttle. The test will end up being a normal test that doesn't get at the noise-performance ratio you're seeking.
 

mamasan2000

Distinguished
BANNED
Well, a video of the cards running at a certain DB would help, like Toms did with the various R9 290X cards.
Even tho they all had about similar decibels, some sounded whinier (more aggravating sound).
So it's not just about recorded levels of decibels, also the character of the noise matters.
 
All I am really interested in is the cards performance and noise at full gaming performance. For that reason I avoided the geforce FX5800 in its time just as the R9 290(X) reference cards now which finally may have topped the aforementioned FX5800 on noise. So really I do not see the point in this comparison.
 

Karadjgne

Titan
Ambassador
I think that when gpus are tested for game performance, they should include an audio clip, so ppl can hear what that card sounds like in comparison to the other cards. A reference card should be chosen, that being the average performer and audio one.
 


Thats a hard choice, the frequency of the noise at the same dB can effect how it leaks out of the case and how every individuals perceive how loud it is.Also the type of case, or if your testing without a case. I think just use your ears and if a graphics card is overbearingly loud, just note it as part of the review.
 

cherubwings

Honorable
Feb 1, 2014
32
0
10,530
This must be the most interesting tread ever. Very delightful to read though.
I really choose the lowest db for having infants, highest during the day, semi during afternoon ect.
 

paitjsu sadff

Honorable
Jan 29, 2014
1,231
0
11,660
to me it should be more about temperatures than noise level...i usualy set a fan curve and i want my GPU to be under 70c, that i think is how GPU should be tested for noise level and heat
 

Mouldread

Distinguished
Apr 17, 2013
985
0
19,360
I can't stand noise! I think I might have some form of Misophonia.

My room (where my PC is) has about 10 dB of ambient noise if that. The only fans in my PC are the CPU fan, the PSU fan and the GPU fan. I have both side panels of my case removed and from about a year I've switched only to SSD due to HDD spinning noise bothering me. I am not using any CD or DVD rom either lol.

When idle measuring from about 1 meter away my PC is at about 20dB. I have also manually locked my GPU fan not to go over 41% as for me it's the only bearable setting.

So I guess different people have different preferences. I would appreaciate if future tests were conducted at lowest possible ambient noise levels.
 

Karadjgne

Titan
Ambassador
NASA has a 0db quiet room they use for testing astronauts. They estimate that for most mature well adjusted adults, any longer than 3 hours in that room will be enough to drive them over the edge.

If your room really is 10db ambient, I'd suggest turning on the radio. Background noise has been shown to help in concentration because it forces the brain to focus on the given task by actively ignoring the background noise.

Pretty much what that means is you are driving yourself further into nut-dom by eliminating noise sources. :D
 
Personally I think trying to bench cards at a set dB level is just as flawed and full of variables as you are saying using a set temperature is while being far less relevant or useful as part of a review. Different coolers and fan bios settings on different brands of cards will give you just the same variances as you speak of.
As a bit of fun for interest it cant hurt but for real world usefulness its a non starter to me.
Performance is relevant as after all that's what the consumer wants and temperature is important as it either indicates that there is overclocking room to be had thermally or that the cooler is not that good or the chips just run hot.
Either way its good info which people can use to help base a purchase on.
Sorry but the FPS a card can achieve for a set db figure has no value to anything.as far as I can see.
Or am i missing the point ?

Mactronix
 

Aphala

Honorable
Oct 4, 2013
159
0
10,690


Actually 95 is universal max :) otherwise the cold soldering melts at 105.

Also,

1. Number of fans variying from 1 - 3 which puts too many factors in to dictate a cards "maximum" noise level
2. Different cards, different gpu and different heat dissipation methods also make hard to factor a real optimal noise level

I guess it's really personal preference and wither or not want you GPU to Melt or sound like 200 2 stroke mopeds just crashed.
 

sr1030nx

Reputable
May 4, 2014
36
0
4,540
Personally I'd prefer 2 separate charts, one for temperature at idle and under max load(prime95?).
And the second chart depicting noise level starting at 0db, and also showing idle and max noise levels.