• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Measurements of MSI MPG Z390 GAMING PRO CARBON

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
Here are some measurements of the audio outputs from a motherboard I just bought for my main workstation.

71ARHkWTa6L._SL500_.jpg


Equipment under test (EUT): MSI MPG Z390 GAMING PRO CARBON motherboard; front panel wired through a Fractal Design Define XL R2 computer case; standard HD Audio Microsoft driver (10.0.17763.1); Windows 10 build 17763; test signals played using FlexASIOTest 1.4 in WASAPI Exclusive Mode for bit-perfectness. EUT is running at 24-bit, 48 kHz, maximum gain unless otherwise noted.

Test equipment: QuantAsylum QA401 Audio Analyzer (using L-/R- inputs; L+/R+ properly terminated; attenuator disabled; baseline results); ASIO401 1.1; REW V5.20 Beta 6 and custom analysis scripts

All the raw data and measurements, including a few more not detailed in this post, can be downloaded here.

Before we begin, a few interesting things I noticed:
  • According to the marketing material of the motherboard, it's supposed to have an headphone amplifier that "auto-detects impedance". I was unable to find any evidence of this; neither the front nor rear outputs seem to behave any differently with a 15 Ω load, 100 Ω load, or no load at all.
  • The rear panel main stereo output is 6 dB louder (unloaded) than the other channels (surround, center, etc.). Be careful about level alignment when using more than 2 channels.
  • I noticed this motherboard seems to be way more sensitive to ground loops on the audio outputs than my previous motherboard (a Gigabyte Z87X-UD3H). That said this is probably highly dependent on external factors, so YMMV.
  • That's the first time I see a DAC that can be used in 32-bit in Windows. Which is of course pointless, but I found it amusing.

UNLOADED MAXIMUM OUTPUT VOLTAGE

Rear panel LR output: 7.1 dBV (2.3 Vrms)
Front panel output: 5.4 dBV (1.9 Vrms)

Nice to see around 2 Vrms output. This should be good enough for most use cases.

UNLOADED 997 Hz -1 dBFS SPECTRUM

Rear panel LR output:
997.png

THD: -84.9 dB THD+N: -83.1 dB

Front panel output:
997.png

THD: -88.0 dB THD+N: -84.1 dB

A result that won't make anyone jump up and down, but is in line with expectations for that kind of product.

DYNAMIC RANGE

Measured in accordance with AES17-2015 6.4.1 and IEC 61606-3:2008 6.2.3.3, unloaded:

Rear panel LR output: 102 dB CCIR-RMS (17 bits)
Front panel output: 103 dB CCIR-RMS (17 bits)

Not too bad considering the kind of product we're dealing with. The ads from the manufacturer claim 120 dB, which, unsurprisingly, is very far from reality (it looks like the manufacturer merely copy-pasted the figure from the Realtek ALC1120 datasheet).

UNLOADED THD+N VS. OUTPUT LEVEL

Measured in accordance with AES17-2015 6.3.4 and IEC 61606-3:2008 6.2.2.3.

Rear panel LR output:
thdn.png


Front panel output:
thdn.png


I'm not sure why there is so much variance in the rear panel measurement. I suspect it might have to do with poor noise isolation from computer activity, making the noise floor randomly jumping up and down. This would be consistent with the fact that the variance is reduced when a load is applied.

LINEARITY

Rear panel LR output (unloaded):
linearity.png


Front panel output (unloaded):
linearity.png


The step on the rear panel is quite weird. It's still there even after re-doing the measurement. The rear panel output also trends downwards at the lowest levels which is usually a bad sign as it suggests dithering issues. The front panel output is better behaved. On the other hand, the rear output still keeps within 0.5 dB down to around -110 dBFS (18 bits) which is pretty good, while the front panel output loses linearity at only -100 dBFS (17 bits). Still doesn't look particularly bad for this kind of product.

UNLOADED FREQUENCY RESPONSE

Rear panel LR output:
fr.png


No objections here.

Spectrograms reveal a sharp reconstruction filter that compromises on imaging artefacts at the top end of the spectrum, which is pretty common. There are no artefacts to be found within the audible range, which is nice.
spectrograms.PNG

(spectrogram parameters: 20 dB gain, 100 dB range)

OUTPUT IMPEDANCE

Measured using a 15 Ω load and further confirmed using a 100 Ω load:

Rear panel LR output: 73 Ω (constant with frequency)
Front panel output: 101 Ω (constant with frequency)

Output impedance this high makes this product extremely unsuitable for driving most headphones. Manufacturer claims of a "dedicated headphone amplifier" are quite laughable in light of these results.

I did a few other measurements at 15 Ω and 100 Ω. Most of them didn't reveal anything particularly interesting, so I won't detail them here, especially considering the output impedance disqualifies this product for headphone use anyway. You can find them in the measurement package if you're interested. The only things worth noting is that I was unable to make the output clip into 100 Ω (1.16 Vrms max voltage, 13 mW max power), however, the rear panel output clips above -3 dBFS into 15 Ω, suggesting a current limit of around 18 mA.

CONCLUSION

As a basic DAC, performance of this product is average at best, but should be enough for most users. The comfortable output voltage (around 2 Vrms) is nice, and might interest users of high-impedance headphones. However, due to the extremely bad output impedance, this product is unsuitable for any kind of critical listening with most headphones; I would recommend treating these outputs purely as line-outs.
 
Last edited:

bunkbail

Addicted to Fun and Learning
Forum Donor
Joined
Feb 27, 2018
Messages
520
Likes
662
This is a very nice review. Worthy of front page promotion, IMO.
 

maty

Major Contributor
Joined
Dec 12, 2017
Messages
4,596
Likes
3,160
Location
Tarragona (Spain)
Off topic

Very interesting the FlexASIO, universal ASIO driver. With foobar2000 and JRMC 64 bits I always use Kernel Streaming to enjoy the music, with a buffer of 50 ms. I will try it to see if it improves the sound.
 

maty

Major Contributor
Joined
Dec 12, 2017
Messages
4,596
Likes
3,160
Location
Tarragona (Spain)
OP
edechamps

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
Very interesting the FlexASIO, universal ASIO driver. With foobar2000 and JRMC 64 bits I always use Kernel Streaming to enjoy the music, with a buffer of 50 ms. I will try it to see if it improves the sound.

There is no reason to believe it would. In general, output methods will not make any sound quality difference unless their configurations diverge in terms of sample rate and bit depth conversions that take place. With the default FlexASIO configuration (which uses DirectSound), Windows will automatically insert a resampler if the sample rate of the stream doesn't match the sample rate of the device as configured in Windows, which might or might not make an audible difference.

Also, if you're using foobar2000, there is literally no point whatsoever in using FlexASIO. The point of FlexASIO is to provide a glue layer to use various Windows audio API backends (DirectSound, MME, WASAPI, WDM-KS) in an ASIO-only application. foobar2000 already has plugins to use WASAPI and WDM-KS directly, so going through FlexASIO just makes things more complicated (and more likely to go wrong) for absolutely zero benefit.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
Audio performance is irrelevant because this motherboard has a 'Mystic Light' AND it's inspired by the shape of supercars- there's even a picture of one on the box...

1553472792973.png


It sure looks like there's nothing in the way of a dedicated H/P amplifier on the board. There may be a small opamp near the front audio header. There's also a smattering of Nichicon UFGs (Fine Golds) near the audio chip.
 
Last edited:
OP
edechamps

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
@amirm: I just stumbled upon your Gigabyte B8 measurements. I was really intrigued to see that your linearity measurement is showing the exact same phenomenon as the one I noticed on the output of the motherboard I tested here: a small 0.1 dB error below around -45 dBFS.

Gigabyte B8:
index.php


MSI MPG Z390 PRO CARBON, rear panel output:
index.php


I wonder if this could be some kind of "signature" behaviour of the Realtek ALC1220 DAC. Not that it really matters of course, but I'm intrigued.
 
Last edited:

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
Very nice review edechamps!

Was always wondering how motherboards might measure - with Realtek being basically the monopoly and whatnot. But boy.. that output impedance tho >_>
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
Although the non-linearity is similar in relative magnitude, the levels where it occurs are very different.

Amir's measurement jumps at -42.5dBFS, yours at approx -52dBFS. How do you explain the 10dB difference? If this is actually inherent in the Realtek D/A converters, one test condition/setup is out by a significant margin.
 
OP
edechamps

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
I don't know. Maybe the DAC is configured differently, or it has something to do with the environment around the DAC.

My point is, this behaviour is very peculiar. I've never seen a linearity graph show that kind of step function before, at any level. I've only seen it in these two sets of measurements from amirm and I. Both motherboards happen to be using a Realtek ALC1220 DAC, which makes me suspect this is specific to this particular chip.
 

Ron Texas

Master Contributor
Joined
Jun 10, 2018
Messages
6,074
Likes
8,906
Nice MoBo for a workstation.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
I've never seen a linearity graph show that kind of step function before, at any level.

I've seen that before. :)

That type of behaviour was common in the early days on ladder (R2R) D/A converters, where a consistent level error kicked in as the upper MSBs bit levels switched off. An offset one way was counteracted by an offset the other, and as soon as that high level bit switched off (as overall level dropped), the consistent error kicked in.

Many D/A converters offered optional trimming of the upper MSBs to obtain the best linearity and lowest level THD. The Burr Brown PCM-58P for example had trimpot options for the top 4 MSBs.

It's possible the Realtek D/A converter chip uses a hybrid PWM and resistor or current source ladder arrangement internally and that's why these errors manifest themselves.

Interesting for sure.
 
Top Bottom