EVGA Launches Its First Audio Card, the Nu Audio

83 Comments

View All Comments

• nevcairiel - Tuesday, January 22, 2019 - link

  Do people really still buy audio cards?
  
  If I wanted something high-end, I would go external, and avoid any electromagnetic interference inside the PC entirely, instead of hoping some vendors EMI shield actually works decently.

• Ian Cutress - Tuesday, January 22, 2019 - link

  The people behind the design are audiophiles that design custom solutions for the high end. If anyone knows what they're doing, it'll be them

• mlkj - Tuesday, January 22, 2019 - link

  Hhmm, I'm sure they have some experience as audiophiles, but that doesn't stop them from taking the "bigger numbers are better game" too far.
  
  Let's take their dynamic range for example.
  24-bits are enough to encode 140dB of dynamic range, and arguably even 16-bits with dithering reaches 120dB of dynamic range. And then their hardware only supports 123dB of dynamic range!
  That 32bit audio support is just pretty misleading marketing then, if your hardware can't output a full 24bit by a wide margin.
  
  But at least 32bit audio is not *worse* than 24 or 16. It's just not better.
  The really dishonest trend is those 192kHz and 384kHz sample rates.
  Counter-intuitively, higher sample rates is NOT something you want in digital audio =)
  
  The short of it is that 44.1kHz and 48kHz are legitimately enough to perfectly (per the awesome Sampling theorem!) encode all sounds your hears can perceive, anything higher is ultrasonics that you can't hear, but that can and does add noise over the whole spectrum.
  That audio is there and has to be processed when you chose a high sample rate, the problem is that audio hardware can't low-noise over the whole spectrum, not at any reasonable price.
  
  Depending on how much your audio hardware prioritizes quality in those ultrasonic frequencies you can't hear (and that's unheard of, pun intended!) the noise might go from imperceptible to shockingly awful.
  
  All in all the card might be good, the THD+N value looks totally acceptable for a PCIe card. But the truth is it won't be better quality than a cheap USB DAC.
  If they know what they're doing like you say, then this card's trying to take advantage of audiophiles who don't understand all those numbers, and I can't recommend supporting this.
  
  Angry internet comment over! =)

• willis936 - Tuesday, January 22, 2019 - link

  >Counter-intuitively, higher sample rates is NOT something you want in digital audio
  
  Well, this isn't really accurate. Higher sample rates are useful because they loosen the requirements of the analog filters in the DACs and ADCs, which are limiting factors. There is no audio information above 40 kSamp/s but if you want those dynamic range and SNR numbers then you need the higher sample rates.

• mlkj - Tuesday, January 22, 2019 - link

  I won't say there's no justification whatsoever for higher sample rates. Some margin is useful in analog filters (but again within reason, or any non-linearity near high-frequencies in your filters will further degrade the lower part of the spectrum).
  
  But at the end of the day I can't take those numbers at face value and tell people that this is a superior product.
  I'm happy to claim that card's 384kHz/32bit output versus a similarly priced USB DAC would win in neither listening tests would nor objective noise metrics in the audible range.

• nevcairiel - Tuesday, January 22, 2019 - link

  The majority of DACs will internally oversample the input for filtering before analog conversion. Due to those attributes of most DACs, a higher input samplerate is not required.

• mode_13h - Tuesday, January 22, 2019 - link

  And how good are those oversampling filters? I'll grant you than 384 kHz is overkill, but if the source material (e.g. mastering software, digital synths, etc.) can produce real signal at that rate, then oversampling can only be worse (i.e. by introducing artifacts).
  
  And for source material not at 384 kHz, your software can use whatever oversampling technique it wants, rather than being at the mercy of whatever the DAC does inside its little black box.

• willis936 - Tuesday, January 22, 2019 - link

  There's no reason to have source files above 48 kSamp/s. The extra 8 kSamp/s covers the non-linearities in the ADCs. Since all of the data you want to reproduce is 20 kHz and below there's nothing to be gained from higher frequencies. Upsampling is a very simple process and the minutiae of the interpolation filter doesn't really matter. It's all simple digital logic running at painfully low clock rates.

• mode_13h - Wednesday, January 23, 2019 - link

  So, you're arguing against oversampling DACs? I think that argument was lost over 30 years ago.
  
  While not as common as 44.1 or 48 kHz, 96 and 192 kHz content has been with us since the mid 1990's. Since then, most recordings have probably been mastered in 24-bit/96 kHz or above, providing a massive catalog of potential content. These days, you don't need a new distribution format, like DVD Audio - you can simply add the format to new or existing streaming services.
  
  And if you think the minutiae of the oversampling filters doesn't matter, then you're no audiophile. But, I think we've established that.

• willis936 - Wednesday, January 23, 2019 - link

  If you'd read the other comments you'll see that I'm not arguing against oversampling DACs. I'm arguing against oversampled content.
  
  Also throwing labels around isn't something I'm interested in, but since you've opened that up: no one should be proud to be an audiophile. Audiophile is a label that's been hijacked by a delusional group who throw objective measurements out the window because they don't want to put the effort in to understand what's going on. They're happy to be fooled and be taken for a ride. You're right. I am no audiophile.
  
  The minutiae of the interpolation filters don't matter in that their audibility is incredibly low and as long as you're not picking something silly like NN or linear you won't see large performance differences. The differences will be there and you can go as far as picking neural net based algorithms, but you won't be able to hear the difference.