Home Network: The ASUS ROG Rapture GT-AX11000 WiFi-6 Router & QNAP QSW-M2108-2C mixed 2.5GbE + 10GbE switch. (Internet jitter and "one number" audio objectivism.)

QNAP multi-gigabit switch being installed behind the basement home electrical panel.

Well guys & gals, it seems like every year, I'll post at least an article on computer networking.

I know, this isn't specifically audiophile-related but for us computer/streaming digital audiophiles, that computer network infrastructure we run at home is important and certainly being home-bound in the midst of the COVID-19 pandemic, a smoothly running network has become even more important for work, education, socialization and play these days.

So with both my kids now teenagers in high school, myself and the wife having to do more work from home, and the ongoing (exhausting) use of video conferencing, I figure I might as well optimize the system and hopefully forego thinking about this for a few years. ;-)

So today, let's talk about wireless, router features, and multi-gig wired home ethernet.

The top image is my new WiFi router. The 4lb, 802.11ax "Wi-Fi 6" capable, tri-band (2.4 + dual-5GHz), up to 160MHz bandwidth in 5GHz, DFS, 12-stream, 8-antennae, 1.8GHz quad ARMv8 64-bit cores, 1GB RAM, 256MB firmware, 2.5GbE-port, ASUS ROG Rapture GT-AX11000; from now on I'll just call it the "GT-AX11000".

[I know that's a lot of specs and acronyms in one sentence; that's just how it works in technology after generations of evolution, for more tech specs, see here.]

As with other "high end" computer peripherals these days, this device came well packaged:

With 8 antennae, and the unusual angulation of the box, this is far from "pretty" IMO. However it is a "gaming" product so there's a certain "aggressive" look I suppose they're going for here.

That shiny part on top of the box is just a plastic protector for the "Aura RGB" multicolor LED on top which adds bling and of course makes you a better gamer for it. ;-)

Out back are some expected ports:

We have a couple of USB3.1, the DC power plug, a convenient power switch. Nice to see >1Gbit ethernet ports more common these days (I spoke about 10GbE years ago for my home system). The 2.5GbE port can be used either as standard ethernet link or as a WAN port. That's cool because at some point down the road once we have >1Gbit internet speeds, one could reconfigure the 2.5GbE port to connect to an ultra-high-speed modem. For now, my Internet download speed is limited to 300Mbps and much slower uplink so it makes more sense to use the 2.5GbE port on the LAN side.

The front just has a few white/red LEDs that are small and relatively unobtrusive. They'll do the job of conveying basic information like whether your WAN is active, whether the 2.5GbE port is in use, etc...

The only other side of the box worth showing are the 3 buttons on the right (when facing the front). Turn on/off the WiFi, set up WPS link, and the Boost button allows you to turn on/off the game packet prioritization for those of you who might need absolute lowest gaming latency! I've mapped the button so I can turn on/off the bling LED instead.

One more thing. This router is not meant to be hung on a wall if you look at the bottom plate:

Notice the lack of any slots to hang it off a wall mount. I've seen folks use an inexpensive "Wonder Mount" like this but note that this will affect the orientation of the antennae and signal quality will vary as a result.

In daily use, this is the first time I've had a router where the 5GHz channel(s) even had a hope of covering most of my home across the 3 levels. Also, this is the first time I've been able to turn on the tri-band Smart Connect feature and leave it on without family members complaining they're dropping signals too often as they move around the house. Smart Connect dynamically changes the connection between 2.4 and the 2 5GHz channels based on signal strength. Load balancing between the different channels works quite well also, including prioritizing the two 5GHz channels when possible.

The stock ASUS firmware is thorough and includes a ton of tweaking options including stuff like wireless power levels (check out the "Professional" tab in the Wireless settings), a pretty good adaptive QoS feature, and even the rules for how Smart Connect functions in "Network Tools". Something I find useful is to optimize power utilization and reduce overall transmission radiation. Health risks are worth keeping in mind, and putting the transmission power down to ~50% "Balanced" (as opposed to higher output described as ~75% "Good" and 100% "Performance") across the 3 channels is a way to lower the radiation while still maintaining good home coverage. Another nice feature is the ability to have a WiFi channel schedule. For example, if peak wireless utilization is from about 3:00PM to 10:00PM, then I can turn on the second 5GHz channel during those times, otherwise, I'll just leave one of the channels on the rest of the time. From midnight to 7:00AM, maybe all that's needed is the 2.4GHz channel operating while everyone's asleep... Nice to have options like this.

The "Wi-Fi Radar" feature and "Wireless Channel Statistics" feature on this router is also excellent for checking out all the other wireless base stations nearby to help you select the best channels to avoid interference. These days, if you're living in a condo or apartment, 2.4GHz probably is already pretty saturated, but it's still good to find whatever "edge" you can. While perhaps not as dire, it's possibly important to know what's happening on the 5GHz channel(s) as well especially with wide bandwidth 40/80/160MHz settings.

The VPN Fusion feature is pretty cool, allowing you to assign VPN tunnels to certain devices on your network such as IoTs (Internet of Things):

For example, I can use my Private Internet Access VPN account in this example and throw the 4 Nest connected thermostats on there to anonymize the IP address to my home and encrypt data transfers from the device to VPN. You can read more about this here.

Of course, a modern router is able to act as a VPN server as well and the GT-AX11000 can create PPTP, OpenVPN and IPSec variants for your use. A simple PPTP tunnel will allow me to access Samba shares from my work to the home network when needed - very convenient on the rare occasion I forget to copy something over onto a USB stick from my main Server computer and need to log in from work.

Speed is excellent as you might expect from a quad-core CPU device (probably based on something like the Broadcom BCM4908) faster than a Raspberry Pi 4 with a snappy web-based GUI. I've had good stability with uptimes of 4+ weeks without needing a reboot on the current non-beta firmware 3.0.0.4.386_41700 (only rebooting for setting change). With QoS active, large transfers going across the LAN and WAN, multiple wireless devices streaming, my son gaming, etc. I have not seen the CPU use go >50%. I'm impressed by the stability and performance.

Power utilization is around 14W on average with all 3 channels on using my Kill-A-Watt meter which is on the higher end of Wi-Fi routers since most consumer routers stay <10W (check out this interesting database). Just slightly warm to touch after weeks of operation.

It's worth thinking about the extra features you might need when looking for a router. For example, with the kids these days needing the hours of inevitable "screen time" for online education, it's good to have parental controls over adult content. The GT-AX11000 will give you at least some control over this with their "AiProtection" feature included in the price through Trend Micro (you have to agree to some data collection). There's also time scheduling for when you don't want the kids to stay up to midnight watching YouTube!

Even without a built-in website filtering feature, you can still implement OpenDNS to control accessible sites. Basically, you'll have to sign up for free, identify which IP you'll be implementing the restrictions on, set the restriction level, and then in the router, have the DNS nameservers point to OpenDNS' IPs: 208.67.222.222 and 208.67.220.220. This has worked pretty well for me over the years with restricting access to "untrustworthy" sites.

Addendum: If you run into issues with Roon playback over the network using the ASUS router, make sure to check that multicast is turned on in the settings. See this post for details. I noticed that it was deactivated by default on my router.

--------------------

Let's talk now about another important device - the ethernet switch, especially the "central" switch often located in the electrical panel of the home where stuff like cable TV and phone lines come in.

While 10-gigabit RJ45 ethernet (as compared to the ubiquitous 1Gb network) is still a bit expensive, it's great to see lower-priced 2.5GbE switches like the QNAP QSW-M2108-2S (~US$250) or the TRENDnet TEG-S350 (~US$125) with five 2.5GbE ports. At this kind of price point, there's IMO no need to do stuff like dual-ethernet link aggregation any more to get past 1 gigabit. BTW if you still want to do link aggregation, the GT-AX11000 router does support gigabit ports 1 and 2 bonding.

The QNAP QSW Series of switches present an interesting array of options. While each has a different complement of RJ45 ports and SFP+ for multi-gigabit speed, the architecture has a total of 80Gb/s switching speed. I was tempted with the 4-port 10GbE / 8-port 1GbE QNAP QSW-M408-4C but decided that the QSW-M2108-2C with 2-port 10GbE but 8 ports of 2.5GbE would, on balance, fill my needs better by upgrading the backbone of my home network to a baseline speed of 2.5GbE for any device with that capacity.

QNAP QSW-M2108-2C: excellent switch with 8x2.5GbE, 2x10GbE ports. The round rotating power connector is convenient.

The QNAP is a Layer 2 managed switch so there are a few settings you can play with:

Realtime port traffic. I'm copying a large file over the 10Gbps ports hence the traffic shown on the right.

Given the high switching speed and being mostly plastic construction, notice that there is a small fan running around 2k rpm as shown in the screenshot above. Fan noise is soft and will not be a problem behind my electrical panel but it would be audible in a very quiet room if you're nearby. I would not put this in a bedroom for example.

So now, with the Syba SD-PEX24066 PCI-E 2.1x1 (500MB/s) network card (only CAD$50) in my upstairs computer, I can access the Server storage at double the transfer rate. One can save a few bucks with the single-port TRENDnet TEG-25GECTX, Syba SD-PEX24065, or generic card instead.

Syba SD-PEX24066 dual 2.5GbE network card. Based on dual Realtek RTL8125 ethernet controllers.

The Realtek chips can get a bit warm. Small heatsinks will do the job...

The ASUS GT-AX11000 router and wireless devices can also now connect at 2.5Gb/s to the LAN.

I'll leave the 10GbE ultra-high-speed Server <--> Workstation communication axis as described previously unchanged which is really the only connection where that transfer rate can routinely be of benefit.

Here's what the network at home looked like back in 2018 after adding the initial 10GbE components:

And here's my network as of April 2021:

Bit-by-bit, the home network gets faster with the addition of "multi-gigabit" links. No need to upgrade my home's Cat-5e/6 network cables.

Notice that there is currently no need at all for the sound room network to be upgraded. Data rate to the TV box even with 4K HDR video and lossless TrueHD-Atmos audio does not need anything more than 1Gbps, and obviously music streaming (even multichannel) needs way less. Don't forget the excellent Netgear Nighthawk S8000 as a nice "audiophile" switch. :-)

Let's have a quick look at the transfer speed across the 2.5GbE connection from Server, through the QNAP QSW-M2108-2S switch in the basement to the "Upstairs Computer" which is my old i5-6500 build from a few years back using the inexpensive Syba SD-PEX24066 PCI-E ethernet card:

Looks like the raw throughput is able to reach close to 250MB/s. Here's a real-world transfer of an 8GB file from Server to the "Upstairs Computer" before and after the 2.5GbE network card:

Notice that with this i5-6500 computer, the copy speed doesn't quite reach 2x the speed of the 1GbE network and we can see from the graph that the speed fluctuates a bit over time. I believe this is a result of the old 240GB SanDisk Ultra II SSD attached to a SATA II 3Gbps port in the computer which is not quite able to stably maintain the full ethernet transfer rate. Notice the initial 2.5GbE transfer rate on the graph is faster; once the network buffer is full, the rate drops down to the nominal SSD speed in the computer of around 200MB/s.

On my main Workstation computer communicating with the Server, using faster M.2 SSD drives (writing to ADATA XPG SX8200 Pro in this example), I can achieve 1GB/s over the 10GbE network:

Protocol overhead and storage speed will obviously affect the maximum transmission rate once buffers are overloaded. One way to reduce inefficiencies is to use "jumbo frames" which I have not done. In the past I've run into issues with compatibility using jumbo frames so have not bothered.

So, with the ability for >200MB/s transfer rate around the house through the 2.5GbE network, WAN Internet speed and slower hard drives/SSDs are really the main choke points. This kind of transfer speed is great for data backups. For audiophiles who might not be familiar with relative storage capacities, >200MB/s transfer speed means an uncompressed music CD can be backed up in 3 seconds or less across the network.

Well, hopefully after this post, there's really no need for another router/central ethernet switch upgrade for a number of years. I don't foresee any need until the day my Internet speed goes >1Gbps, maybe. ;-)

Gradually, I'll see more Wi-Fi 6 devices used here at home. So far, my wife's iPhone 11 and son's recent Lenovo IdeaPad Flex 5 laptop are the only devices capable of communicating in 802.11ax. Link speed on these remain high even at a distance from the router. Likewise, most of my wired ethernet devices are not 2.5GbE capable, much less 10GbE so unless specially called for, 2.5GbE wired networking will be more than sufficient for the foreseeable future.

For those who really want to live on the bleeding edge, you can consider the upcoming Wi-Fi 6E routers! ASUS ROG GT-AXE11000 is an example of such a router. The difference is that the 6GHz band is available and this would be great I think if you're living in an apartment, condo or closely spaced townhouse complex where there could be congestion from the neighbours. For a detached single-family home, I don't think there's a need especially if your router already supports DFS channels. Already with 5GHz, the coverage here at home isn't fantastic, so with 6GHz, range limitation would be even more apparent.

BTW, if you're confused about all the channels available to 5GHz WiFi including DFS, the various UNII bands, make sure to have a look at this BCCA.org page. Obviously it can get complicated and not everyone needs to be a geek about such things.

--------------------

Let's spend a few moments thinking about the network and the fact that streaming music through the Internet is a key form of audio delivery these days. Rather than provide answers, sometimes it's best to first ask questions and maybe examine how some audiophiles think...

There's a pretty cool tool in the GT-AX11000 router that is used to measure your internet speed and keeps a record over time (under the Game Acceleration tab). This way, you can make sure that your connection stays consistent and easily monitor that your ISP is giving you the promised speed you paid for.

Check this out - 3 readings done on the same day:

Notice that at around 5:30PM (arrow), while the download and upload speeds are what I paid for (>300Mbps down, 16Mbps up), the measured data Internet packet jitter to the same server (Shaw Cable being the internet company) jumped from <0.5ms to over 2ms (yeah... that's milliseconds, not picoseconds or femtoseconds). My guess is that there's some kind of peak utilization around the neighborhood during this time of the afternoon (Saturday) with a number of households concurrently hitting the network. This finding is actually quite repeatable in my neighborhood, every day around 5:00PM, network jitter does get higher than other times of the day.

Here's a serious question to the audiophile "Golden Ears" (Homo sapiens subspecies aureus auribus) who ostensibly can hear network differences even down to the switch level:

Do you hear a difference or believe you hear a difference streaming lossless audio from Qobuz / Tidal / Amazon / Deezer / etc. during various times of the day? 

Would you say higher data packet jitter to/from the streaming service affects the sound?

Does audio streaming sound "better" in the quiet of night compared to "rush hour" early evening when the network is busier? Is this Internet jitter related? Or is this mainly noise over the network/electrical system? How do we tell the difference?

If you've invested in an expensive ethernet switch (like this guy who claims "measurement morons aren't even real audiophiles", or this Network Acoustics ENO "filter" that drops 1Gbps to 100Mbps in the absence of measurements), do you think the specialized switch/filter will protect the audiophile from this?

It might be good to ask this question to people like Hans Beekhuyzen, Jay Leung, folks here, and John Darko. It would be interesting to understand if there is a consensus on the beliefs and attributions...

--------------------

In other audiophile news, it's somewhat fascinating hearing John Darko asking the interviewee about a supposed "obsession with THD", or the idea that objective folks just wanting "one number" in his podcast. That's patently ridiculous and Darko's just setting up strawmen fallacies all over the place, creating an unrealistic characterization of the depth and breadth that reviews with objective testing are able to explore. This does not help but create enmity through perpetuating "black or white" stereotypes. I seriously hope that these days we can get beyond such simplicities. Even if one were to use a single "summary" like either a THD(+N) or maybe frequency response graph, nobody is suggesting that this captures everything about the performance of a specific device even though it could be very important and actually might explain a large part of what is heard (eg. frequency response with headphones)! At 46:30, discussion of the "psychology that surround people's obsession around numbers" is also interesting although I think very wrong with irrelevant correlations being made. It's also dangerous to talk about Dunning-Kruger or implying this in others when a person conducts only subjective reviews and doesn't open up for comments and criticisms!

Despite the interview, I do like Jonathan Novick's RMAF 2015 presentation. If we look at the RMAF 2015 talk, around 40:00, his summary of testing with reminders that "not all distortions sound alike", "graphs tell more than specs", "measure where you listen" (ie. at normal power levels for example), etc... you'll see that folks who test with measurements generally will address these points. There's no mystery in there. In fact, objective audiophiles like myself publishing these tests are not acting like a manufacturer's advertising department who might want to hype up specs and throw out only a single big number with the intent to sell stuff. If anything, the testing punishes the devices and I'm just as happy to report problems as I am to see excellent results; that should be the case for all independent reviewers I hope.

I think the most interesting and important part of the Novick 2015 talk is a reminder to listen and test for "transitory conditions" which is a good point about idiosyncrasies that might not show up on static signals. The other important revelation he makes is "we should be able to see anything that we hear in the waveform" - yes, it's all testable objectively. An axiom of objective testing of sound quality which I don't think Darko explored with his guest.

I think it would be nice for folks like Darko to just be more real and provide examples they're referring to instead of these dramatic handwaving generalizations and vague claims of what "some people" believe (like that single numbers are all that important). I'm really not sure if these days, many people believe strongly in what manufacturers publish as a single number in the specs anyways. So maybe at the end of the day, the message should be: Be wary with manufacturer-promoted single-number specs, but encourage careful, detailed measurements when reviewing!

Conveniently, Darko also seems to be the only audiophile media personality commenting on the recent GoldenSound MQA video exposé with a convenient "I understand some of it, but not all of it". Sure, when you can't argue, plead ignorance.

Funny that he picks up on the AudioQuest Carbon 1m S/PDIF cable measurement in the MQA video showing the difference in +/-250Hz sidebands down at -133dB around the 12kHz synthetic J-Test signal ;-). Yeah, that's significant value right there for your US$200 cable! Music will clearly sound much better with the AudioQuest because of that difference?! Have a good look at the magnitude of MQA's effect compared to that jitter measurement. (As usual, have a listen to jitter effects yourself.)

Okay... Enough with MQA, and commenting on other audiophile sites / commenters for a few weeks at least. I promise. ;-)

I watched Nobody with my son the other day. A turn-your-brain-off-plethora-body-count type movie - enjoyable if you're into the genre, like the John Wick series but more humor. Great to see Christopher Lloyd on the screen again. Bob Odenkirk does a great job as an "everyman" older suburban dad action hero. The score by David Buckley is enjoyable and propulsive but they really need to release a proper full soundtrack with all the classic songs featured in the movie which added to the atmosphere substantially!

That's all for this week, folks. Stay safe, enjoy the music, and may good network speed be with you and yours...