Best Ethernet Cards for Streaming

[TomJ]

Hi all,

 

Since I am looking for the cause of the different sound of LAN cables and switches, I had bought on Ebay Tektronik TDS784A 1GHz oscilloscope.

To get a good 100Base TX reference for my cable measurements I wanted to check which network card produces the best Ethernet signal and also sounds best.

 

NIC in my Test

 

• Intel X540 T2 10GbE Dual Server NIC
  • The question was if a 10GbE card produces a better 100Base TX
• Supermicro AOC-STG-i2T 10GbE Dual Server NIC
  • The question was if a 10GbE card produces a better 100Base TX and how does the Supermicro competes with the Intel
• Intel PRO/1000 PT Dual Gigabit Server NIC with 82571EB chipset
• Intel i350 T2 Dueal Gigabit  Server NIC
  • Question: How does the i350 card compete with the older 82571EB card?
• Intel i210 Desktop 1 Port Nic
  • very small card that doesn't have much power conditioning and isn't that complex
  • Question: means less complexity a better sound?
• TP-LINK TG-3468 1 Port Nic
  • This card was already praised in other threads here
  • very small card that doesn't have much power conditioning and isn't that complex
  • Question: means less complexity a better sound?

 

Test Setup

• Host system for the NICs
  • Gigabyte Z97N-WIFI motherboard with WLAN deactivated, PCIe slot and HDD raid system
  • Windows home Server 2011, which is actually a Windows Server 2008
• Cable
  • 1 meter of a shielded Gore ePTFE CAT cable, which is my current reference and with best measurements so far and best signal.
  • Shielding only connected to the NIC
• End device and measuring point
  • A discarded Asus WIFI router on which the measurement was made at the RJ45 connector's release points - in front of the router's transformer. So no Magjack.
  • The router produces a very distinctive signal, so that one can quickly recognize in which auto MDI-X mode the router and the card are.

• Measuring instruments

  • Oscilloscope: An old Tektronix TDS784A 1Ghz

  • Probes: 2 x Tektronix P6245 active 1.5Ghz probes

  • The grounding of the probes was bridged and the Ethernet differential signal was generated via a Math function in the scope

 

Result

• I don't want to go too much into the sound differences, as this is always very subjective, but roughly give a tendency. The sound was determined by a direct connection of the network cards with my KEF LS50 W II.

 

Intel X540 T2 10GbE Dual Server NIC

• The 10GbE card is apparently really optimized for 10Gb, but does not produce the best 100Base TX signal. The sound is a bit brighter and shriller.

Supermicro AOC-STG-i2T 10GbE Dual Server NIC

• Similar results to the Intel X540 card. It also has the same chipset.

Intel PRO/1000 PT Dual Gigabit Server NIC with 82571EB chipset

• A somewhat wavy eye patter. A bit less nervous than the 10GbE cards.

The Intel i350 Dual Gigabit Server NIC

• produces the cleanest signal and is also the winner sonically in my system with a very clean, quiet sound. It has the same chipset as the Jcat cards.

Intel i210 Desktop 1 Port Nic

• Poor signal and eye patter. Sound is a bit dull.

TP-LINK TG-3468 1 Port Nic

• The very worst card in the test. Also very dull and undifferentiated on the sound side. No idea what the colleagues find sonic about this one.

 

Summary

 

If you want a very good network card and do not necessarily want to buy audiophile cards with super clocks (although I would like to see measurements of this), then I can highly recommend the Intel i350 T2.

 

Measurements in next posts

 

 

[TomJ]

Intel X540 T2 10GbE Dual Server NIC

[TomJ]

Supermicro AOC-STG-i2T 10GbE Dual Server NIC

[TomJ]

Intel PRO/1000 PT Dual Gigabit Server NIC with 82571EB chipset

[TomJ]

Intel i350 T2 Dueal Gigabit  Server NIC

[TomJ]

Intel i210 Desktop 1 Port Nic

[TomJ]

TP-LINK TG-3468 1 Port Nic

[AudioDoctor]

This is a good post!

[The Computer Audiophile]

Thread cleaned up. Thanks @bobflood for letting me know, I hadn't been following this thread. 

 

@TomJ do you think your thread is best placed in the objective section or left in this section due to your attempts to correlate sound quality with this stuff?

 

 

[Exocer]

Interesting post! Thanks for sharing. I had always wondered how these Aqtion AQC107 nics stacked up but never got a chance to try:

 

Syba 1 Port 10 Gigabit Ethernet Network Card - PCIe x4 10Gb 10GBASE-T NIC AQTION AQC107-10Gbps Ethernet PCI-Express x4 Adapter SD-PEX24055 https://www.amazon.com/dp/B07G2DV9K2/ref=cm_sw_r_apan_glt_fabc_S2JM4JY52VHRBSNJPVFN?_encoding=UTF8&psc=1

 

I was looking into all sorts of industrial class PHYs and stumbled upon what appeared to be an impressive PHY at a reasonable price. I'm not expert in this area so I'll take a step back and learn 😃.

 

The chipset itself doesn't seem to be common and will likely not be supported on some OSes but I may just pick one up for subjective reasons.

 

I had actually purchased that TP link you tested based on some comments I read here and it just never did anything special for me.

 

Cheers,

-Rob

[TomJ]

1 hour ago, The Computer Audiophile said:

@TomJ do you think your thread is best placed in the objective section or left in this section due to your attempts to correlate sound quality with this stuff?

Thats a good question.

 

Since I am still quite new in the forum, I put the question back to you: Where would the topic be better located?

 

My mission is to bring more objectivity into the Ethernet sound discussion with my measurements and maybe find reasons for this. I am a very analytic person and I am very familiar with ethernet and its protocols. Nevertheless I hear differences from cables, isolators and switches.

 

 

 

[TomJ]

4 minutes ago, Exocer said:

Syba 1 Port 10 Gigabit Ethernet Network Card - PCIe x4 10Gb 10GBASE-T NIC AQTION AQC107-10Gbps Ethernet PCI-Express x4 Adapter SD-PEX24055 https://www.amazon.com/dp/B07G2DV9K2/ref=cm_sw_r_apan_glt_fabc_S2JM4JY52VHRBSNJPVFN?_encoding=UTF8&psc=1

I wouldn't bet too much on this card. 
Maybe the PHY is very good, but the power supply might be the weak point.

[Exocer]

9 minutes ago, TomJ said:

I wouldn't bet too much on this card. 
Maybe the PHY is very good, but the power supply might be the weak point.

Thanks for your feedback. I will scratch this off of the list!

[TomJ]

16 minutes ago, Exocer said:

Thanks for your feedback. I will scratch this off of the list!

What else is on your list?

My measurements tell me that 10GbE is not necessarily the way to go.
I am currently looking for new objects for my measurements and am currently browsing through the Supermicro catalog.
I would also like to test a Jcat net card xe or the new Sotm card to compare, but I exclude these currently, because I think that the manufacturers of such exclusive parts should actually publish such measurements themselves and apparently shy away from the comparison.

[Exocer]

14 minutes ago, TomJ said:

What else is on your list?

My measurements tell me that 10GbE is not necessarily the way to go.
 

Interestingly enough, the XT540 was on my to do list as well. I currently use this Startech with good results: https://www.startech.com/en-us/networking-io/pex1000sfp2

 

I was interested in going back to copper for the sake of experimentation...since some people who's ears I REALLY trust are not too much into fiber nics.

 

I had nothing else on my radar. Will keep on eye on your experiments though!

 

Thanks,

-Rob

[One and a half]

The Intel 350 is fitted to the JCAT femto net cards, for both generations. 

[TomJ]

59 minutes ago, One and a half said:

The Intel 350 is fitted to the JCAT femto net cards, for both generations.

I know. But it depends on the implementation - so a comparison would be interesting. 

[shum3s]

I have been using this Ethernet card in a NAA setup. Do you have a opinion on its quality? 
https://www.sonnettech.com/product/solo10g-pcie-card.html 

Thanks, Sam

[TomJ]

5 hours ago, shum3s said:

I have been using this Ethernet card in a NAA setup. Do you have a opinion on its quality? 
https://www.sonnettech.com/product/solo10g-pcie-card.html 

Thanks, Sam

Hi Sam,

 

it is quite difficult to make statements without measurements, as much depends on the implementation.
As I have already written, I would not necessarily prefer 10GbE. In addition, discrete transformers are usually superior to magjacks with integrated transformers in my opinion. The card has integrated transformers. If you look at the cards of the measurements in comparison, the cards that have more effort with the power supply have a clear advantage. This card doesn't seem to make too much of an effort here. But you don't know exactly what's under the heatsink.

 

The only way to find out would be to measure it. If you would send it to me, I could measure it - but I have no ambitions to become a second Amir with an EthernetScienceReview ;-) But I think this is illusory, since I live in Germany and you most likely in the USA.

 

Best regards,

Tom
 

[shum3s]

Thanks Tom, makes a lot of sense for the need to measure. Yes I am in the USA. I sure like what your trying to accomplished in starting this thread, I will continue to follow it with great interest.  Sam 

 

[bibo01]

@TomJLike another Tom friend of mine would say: you are punching a hole in the water.
You are measuring some small inconsistencies in the digital protocol.
If you really want to investigate any differences on the streaming, you have to take the analog signal from a converter after connecting with all the interfaces to be tested.
This is the only way to evaluate the possible difference and the ability of the device to recover the error.
Many interfaces generate different performances with relative easily measurable jitteral alteration. However, when you connect a DAC, the result from the analog outputs does not change because the jitter recovery system sweeps it all away.

[TomJ]

@bibo01 That's what my head says too, but unfortunately my ears say something else.

And because no measurement method has yet managed to measure the differences on the analog side, I had concentrated on the digital side. I am not so naive that I would consider jitter in the Ethernet signal as a problem. As already written, I think the sound difference could come from spikes in the signal.

 

The manufacturers of audiophile network devices all advertise their clean signal as the origin of the better sound - unfortunately so far without proving this with measurement results.
There are some that supposedly achieve a better sound with super clocks upgrades and better power supply. I have my doubts, when I look at the result of the i350 card, whether this can really be achieved a significant improvement.

 

 

[Miska]

I would also compare things like amount of hardware offload features. This affects how much CPU load network traffic generates. Usually server controllers offer more offload features.

 

[bibo01]

3 hours ago, TomJ said:

And because no measurement method has yet managed to measure the differences on the analog side

...well, that's not really true. Perhaps "classical measurement methods" haven't, but otherwise it is possible with the proper infrastructure in place by capturing and comparing analog output with original signal. 

[Cogito]

On 10/9/2021 at 12:56 PM, TomJ said:

To get a good 100Base TX reference for my cable measurements I wanted to check which network card produces the best Ethernet signal and also sounds best.

Digital network signals (Ethernet) have no bearing on sound quality. The shape of the signals you saw on the oscilloscope have no impact on the sound. 
 

 Network signals are at 2.5v which is divided into 2 states, 1 and 0. I don’t know the exact details, voltage ranges to represent 0 and 1 are huge, something like 0.5v and above represents 1 and anything below 0.5v represents 0. Even in the worst case, any lost data packet is retransmitted. In a home LAN, retransmissions do not cause any I interruption in the music stream.

 

The noise in the network is what causes problem to the audio. Ethernet cables act line antennas. They pickup all kinds of noise which enters the computer and eventually the DAC thru’ USB interface. Noise filtration/elimination is the key to good audio.