WiFi 6 + Intel AX200. What to really expect?

Hello bros,

Today I pulled the trigger and bought a new router. I bought the Huawei AX3 PRO. I really don't have 100% clear what's the difference between the regular and the PRO version other than the fact that the regular is a dual core device and the PRO version is a quad core device. The difference was only $20 so I went for the PRO version.



I'm also going to buy an Intel AX200 M.2 2230. So what LAN speed should I expect? According to Intel the AX200 can go up to 2400 Mbps and according to Huawei it's AX3 can up to 2402 Mbps. But let's get real guys, do you think I could be able to get actually 1 Gbps of useful data rate?

* I'll be with less than 10 feet of distance from the wireless router.
* No walls will be in between the AX3 and my laptop.
* The server I'll be accessing will be connected through Giga Ethernet to the AX3

Any experience or thoughts are welcome.

UPDATE:

So I finally got my AX setup. Intel AX200 + Huawei AX3 Pro.

The router has solid performance for my needs, at least, the only issue I had is the firmware is in Chinese but I sorted out using the translate tool from Google Chrome. The Router has little software options but I really don't need QoS/VLANS or any advanced stuff.

I feel that the speed really got so much better compared to my old AC setup.

This is my results in Tamosoft, I left the software running for like 1 hour:



This is what I got in real world file transfer from my laptop to my remote server. The server is connected via Gigabit Ethernet and the laptop is using AX200:



I have done several tests and always get that speed or similar behavior, which is more than good to me. Previously I was getting most of the time like 20 MB/s and best case scenario 45 MB/s using AC 867 Mbps with the AC9260 (server in wired connected to gigabit ethernet as always) and not as solid performance more fluctuating speeds I'd say.

This is using the speeds I get using the laptop connected via gigabit ethernet to the same switch as the remote server which is also connected using gigabit ethernet:



The difference is not that big and maybe the mechanical HDD is being a bottleneck to the communication link. At the end of the day I'd say that the performance difference is quite high and also the uplink is more stable.

 

I remember my Plex server spitting out 900Mbps for some obnoxiously high bit rate video transcode and my client actually sucked it up quite agreeably. No experience with that router though, I went with another brand that I cant remember at the moment, want to say D-Link

 

Actually my PC server which I'm connecting via wireless with my laptop has three main purposes:

• To run my algorithms of trading
• To run my Plex server
• To storage all my media content (movies, games installs, series, etc)

My Plex server is connected to the Apple TV 4K via gigabit ethernet network so I'm not aware of the bit rate when transcoding. But do the maths a 30 GB movie with a length of 2 hrs has an average bit rate of 34 Mbps. But that's the average, at moments it can easy go over 100 Mbps. A peak of 900 Mbps seems like an outlier, but not impossible.

The thing is it's not uncommon for me to move large amounts of data from my laptop to the PC server or vice versa and my current AC network is so unstable I hate it. I get most of the time speeds as low as 80 Mbps which is ridiculous, I have to disconnect and connect to the wireless network to speed up the bandwidth to 300 Mbps, which is also pretty slow to my taste, especially for a "866 Mbps" uplink

 

Its worth noting that I use the router in the room I primarily use while on my computer, so I dont have any obstructions and such. I do my cardio on a stationary bike so I like to use Steam streaming to my t440p (which is wired) and my desktop is wireless (AX200)

The 900Mbps was for some obnoxious anime with a high audio and video bit rate, I havent used it in a while but I plan to get that network back up and running again as I need to get back in shape (have been lazy)

I would wait for others to chime in about wireless distance and its potential woes.

 

I can get around 1200-1350 Mbps real world transfers to my NAS either 1 floor below or same floor with two walls. Same room as router it’s also about Mbps 1350 Mbps or so. My home is mostly drywall/wood interior. If you have more concrete walls or granite flooring expect a major hit though.

2400 Mbps is just a link rate, ie theoretical transfer rate. Real world will be at best about 65-75% of that. Speeds also depend on your router and how the antennas are on the client device, as in two laptops with the same card can have differing speeds based on antenna layout.

You can already cross gigabit speeds with HT160 on decent AC HT160 capable routers like the R7800 paired with the Intel AX200 or 9260AC. AX on 5Ghz will generally give you a 5-10% boost and at best around 20-25% over AC. AX real benefit is doubling of the 2.4 GHz throughout. Other stuff like spectrum sharing which is one AX biggest touted benefit, won’t really make difference till some years in the future when AX is actually more common. Another important thing is the AX3/AX3 Plus does not have a Multi-GiG (NBASE-T) Port ie capable of 2.5/5/10 Gbps so you will be limited to around 940-960 Mbps for anything that goes through the WAN or wired LAN. Only Transmitting between two WiFi devices might net you faster speeds.

TLDR: Same room expect to cross gigabit speeds using HT160 on the router ie 2400 Mbps mode. Even on HT80 you should be able to get peak sustained a little above 800 Mbps, at least in my environment it’s possible.

As for differences between the AX and AX3 Plus:
"The main difference is that the AX3 is equipped with two independent signal amplifiers, and the AX3 Pro uses a 2.4GHz / 5GHz amplifier (four in total). In addition, the AX3 Pro also supports one-touch networking. So turning on the NFC function of the smartphone and tapping the router’s NFC area will ensure a fast and secret-free connection."
Source: https://www.gizchina.com/2020/03/30/huawei-releases-two-wifi-6-routers-meet-ax3-and-ax3-pro/

 

That sound quite promising. 1 Gbps is my target, as I said my laptop would be most of the time pretty close to the router in the same room actually, so there is no much reason to expect otherwise. I blame either Windows 10 or the Intel 9260 for my poor AC performance. I say this because I have a fiber connection of 500 Mbps and 90% of the time when using speedtest to check speed connection my cellphone beats my laptop by easy 100 Mbps

 

In the past I got a functional 1.7gbps link rate with a R7800 and 9260 when all of the settings were correct on both sides.

The issue you will likely run into are
1. WAN is a single port and maxes at 1gbps
2. LAN same as 1 unless you can bridge the ports into a LACP / LAG to exceed 1gbps limitations
3. you can go adhoc and hit the speeds being marketed between 2 devices directly (wouldn't need the router to do)

The way to get around these limitations though is to make your own device with mutliport cards in a PC case or look into Ubiquiti EdgeRouter Pro + their PRO / HD AP's where you can bundle the ports on the WAN/LAN/AP to get a higher throughput along the path and then your bottleneck will be your ISP unless you're on a 2Gbps connection or higher.

 

What was your experience in terms of real throughput with the 1,7 Gbps uplink?

About LACP I'm not sure if it would be useful in my case. If I remember correctly in etherchannel you can bundle several switch ports and create one "big" virtual port, but every single speed "connection" is still limited to the speed of the single port speed. LACP is useful when you have multiple end points connected to a switch so they all can benefit of the port channel, i.e. several 1 Gbps connections at the same time, using the etherchannel but end to end connection will still be limited to 1 Gbps

 

Max I saw on he R7800 at a 1.7 Gbps link rate was 120 MB/s (960 Mbps) sustained even one floor below the router, the limiting factor is the gigabit ports. The RAX120 could hit ~140 MB/s (1,120 Mbps) in AC mode (AX mode adds another 100-200 Mbps at best), so I think theoretically the R7800 could reach around 1120 Mbps if it had a 2.5 or 5 Gbps port.

 

Well my LAN server (the desktop PC) has only mechanical disk drives, which are limited to about 700-800 Mbps write/read speed and has a gigabit ethernet connection to a gigabit switch. So my bottleneck will most definitely be the HDD. Or even worst my laptop has a pretty fast NVMe but also has a pretty slow 2 TB HDD which I've never seen go any faster than 600 Mbps. So if I can get a real 1 Gbps of real throughput over WiFi I'll be more than happy.

Probably in the future when SSD could compete in $/GB with HDD I'll migrate to a SSD storage server, buy a PCIe 10 Gbps ethernet card or SFP card and 10 Gbps switched network, but I don't see that happening in less than 5 years

 

If you do later upgrade to a SFP+ card, look at the Chelsio T540-CR 4 port its like $100-140 on the used market. It's an excellent card, probably will drop further in price by the time you switch. You can use RJ45 ethernet transceivers for regular connections if you have a mix of SFP+ and RJ45, most transceivers are capable of 2.5 & 5 gbps alongside 1/10 Gbps modes. Wiitek and 10Gtek have the best ones, with the Wiitek being the best priced.

And if just for fun if benchmarking and you wan't to circumvent the gigabit limit to test WiFi speeds, if your router has fast storage performance you can test directly to a USB drive connected to the router. Unlike the RAX120 which has a 5Gbps port, the RAX80 only has gigabit ports but to test peak wireless speeds I tested my Samsung T5 SSD on the router, it also gave around 1120 Mbps on AC and around 1300-1350 Mbps on AX just like RAX120. The R7800 doesn't have as good USB storage performance (~560 Mbps) so it's hard to test max wireless speeds of the chipset without a WiFi to WiFi test but that's less ideal in regards to interference etc.

 

HDD's can be sped up a little bit if you do a flash cache setup.

10GE is the way to go internally until providers start bringing 10GE to residential locations like some of the public utilities around the country pushing out the ISP's with their own community networks like in CO/TN/OR/WA and some other places where they'll give you 10GE for the same price as the big ISP's are charging for 2GE.

10GE cards aren't all that expensive anymore compared to where they used to be. I've seen 2 port 10G's going fairly cheap and if you go SFP instead of copper they're cheaper because you end up supplying your own optics instead of copper built in.

The issue with 10GE though is you would still top out at 1GB/s w/ overhead or less if there's an I/O bottleneck in the controllers handling the data. Flash only goes so far before other issues arise like heat throttling things. I can burst between 2 NVME drives about 2.5GB's but, sustained they drop to 1-1.5GB's depending on the types of files with both drives rated over 3GB/s.

With the R7800 I could hit about 98% of the bandwidth from the provider on speedtests on multiple servers over wifi. AX sounds like the new shiny kid on the block and there are some features that are appealing more so if you're hosting wifi in a small gathering of people constantly competing for packet windows. If you have a large family constantly using devices at 50% utilization then it might help alleviate issues but just a single or couple of people it won't really be noticeable.

If I could find a reasonably priced multi port 10GE card that I could segment into different networks it might be fun to play with but, realistically I don't really need it since the physical devices connected run @ 100M (TV/Soundbar) since I consolidated all of my peripherals into a single box approach instead of 5-6 different things plugged in drawing power all of the time. If they would make a cable modem card I would toss that in there too and eliminate 1 more item.

Speaking of which.... when they do open the Docsis 3.1 FDX flood gates over 1gbps the Netgear CM2000 or something along those lines (MB8611) would be a way to get beyond the bottleneck w/o going to Fiber to the prem. Of course they're slow as molasses when it comes to coughing up more bandwidth unlike other countries do for le$$.

 

Thanks for the advice on the PCIe card I'll look in to it. Well the Huawei AX3 has no USB ports IIRC and I don't think the internal storage is accessible. But with stable 600 Mbps I'll be happy. As I said now for some obnoxious reason my AC network is way too crippled and unstable, which really makes me angry, lol

 

Just to clarify you're talking about 1 GB/s = 8000 Mbps ????? ...if so, that's really impressive, even though there shouldn't be such a big loss on speed. If I remember well Ethernet efficiency is around 90% where the 10% is waist in header and tails of packet data. You get better efficiency of you use jumbo packets

 

Well, when you convert bits to bytes you have a factor of 8... real world throughput should be able to push 97%-98% w/ packet overhead if you have good optics in place. In reality from a business load planning though you try to stick to 50% of capacity in the event that if 1 link fails the other can handle the load while restoring the failure.

On my gigabit connection with 2 LACP ports into my cable modem I can exceed 1gbps with the over provisioning in place from the provider. While it doesn't make all that much difference it's nice to prove that it's functioning as expected and works when needed. Since implementing wireguard throughout the connection in addition to the firewall it doesn't quite hit those speeds due to the packet inspections but it does still hit about 97% over VPN if connected to the right server w/ a low load of users on it.

I don't use jumbo packets either but, in a business environment we typically set them to 9216 vs 1500/1514/1492. The MTU is changed depending on the network requirements.... typically anything under 1500 is for pppoe / vpn so the packets conform to the typical 1500 when accounting for protocol overhead + ethernet overhead.

 

Do you know any software to test the speed from/towards a LAN server with a host, similar to speedtest.net but for a local network???? What I usually do I copy and paste a large file as a "benchmark" but there has to be a better way to test the local network speed but I haven't found the proper software to do that

 

Lan Speed Test ( https://totusoft.com/lanspeed) or TamoSoft throughput Test ( https://www.tamos.com/products/throughput-test/) or iperf.

LanSpeed Test: Can run with just client application and test to remote SMB or install server application on 2nd device if you want more accuracy.
iperf and TamoSoft: Require application on both client and 2nd device/server to function. I like Tamo

I personally just do a large transfer to check peak sustained speeds, a mix of small file sizes may be useful as well. Depends more on what you are looking at in terms of performance.

 

Well, if you want to get nerdy about speed testing you could setup iperf on both sides to run tests. I base my speeds off actual usage though with file moves. If you want to bypass the windows performance issues then maybe use WinSCP / SecureFX or a similar program to benchmark the transfer speeds.

 

I used TamoSoft to test my current wireless setup:

• Server: Wired gigabit ethernet (Windows 7)
  
• Client: Wireless Intel AC 9260 (Windows 10)
• Acess Point: A generic one, provided by the ISP (a device that is modem, router, AP and switch)

I got this results:



Seems about in line with the performance I get with just a copy-paste from my laptop to the server. I know WiFi is half-duplex, is that why TCP up + TCP down is similar to the link connection 867 Mbps?
Or I was supposed to get 867 Mbps down and 867 up in this test (in an ideal world)? Couldn't find this information on the documentation of the software

 

Yes WiFi and the link rate are half duplex, so you can’t really add up and down. Never expect to get anywhere near the link rate in real life, that’s just a theoretical max in “ideal” conditions as in no overheads, attenuation, interference etc. Link rates are adjusted by devices based on signal attenuation/interference.

At best expect like 65-75% of the link rate in real world performance in good conditions.

 

I know that 867 Mbps is the maximum throughput including the 802.11ac header and IP header, so the max theoretical of TCP payload has to be around 80% of 867 Mbps = 694 Mbps

In that case I'm getting up-average 39% of max and down-average 65% of max. Do you think my speeds are ok?

I found this interesting article:

MAC INEFFICIENCYThe degradation in MAC efficiency is asmall factthat is usually left unspoken. In the ‘old’ 802.11n standard operating at 20 MHz the MAC operated with over85% efficiency.By comparison, the 802.11ac MAC efficiency is 65%. We are ‘wasting’ 20% more time on overheads.

https://www.winncom.com/images/stories/GoNet_WP_Things_they_dont_tell_about_80211ac.pdf

 

Yeah it's pretty decent, also remember the client such as a laptop or mobile device generally has a weaker output (though not always) than the router so it makes sense in regards to the upload. Best sustained on actual transfers I've seen on a 9260 AC or AX200 at HT80 on 5 Ghz (AC) is around 80-85 MB/s (640-680 Mbps) down and 60-65 MB/s Up (480-520 Mbps) on the few Dell Latitudes and Precisions I tested.

 

@Tyranus07 Your average speeds are allright-ish. Keep in mind that most of the time people will mention speeds they reach as in the best one they have seen, not an actual average. So you'd be comparing your actual speed to someone else's best effort.

Apart from that there are so many environmental variables with Wi-Fi that you have to leave a lot of wiggle room for what would be considered confidence interval in statistical terms.

 

I tend to hit 60-70MB/s over AC. It's consistent with the settings I have in my self built box of many trades and expert of none. If I could get hostapd to unlock 160mhz channels and work in a stable manner I would blow it out of the water with up to 1.7gbps like the Netgear router I had successfully gotten working. The card I'm using with hostapd has the same chipsets as the Netgear so in theory if hostapd was working to the extent that documentation says it should then life would be just a smidge better than it is now.

Mind you the 60-70B/s is symmetrical speeds when pulling / pushing files. Meaning I'll see those speeds in either direction but, usually only dong one direction at any given time other than some basic beacons in the background from other devices.

The ideal situation is to make sure both the Card / AP have a similar setup whether 2x2 or 4x4 or more depending on the technology. Maximize the throughput with the antennae on both sides. OEM cards tend to ship 1x1 and simple upgrades like a 9260 fix some of that doubling you from 433 to 867 which is typical even these days for most clients / AP's to negotiate to unless you opt for AX then the options are there for more but, it's mostly marketing since you can't typically get those speeds end-to-end anyway.

If you're in need of a refresh anyway then go for it... otherwise it's not really worth the cash unless you're speeds are really really slow and unbearable. For instance if you weren't on AC already just that upgrade gives you 8X the speed of BGN. Going from AX >> AX just introduces more streams available for more devices and a way to keep them passing traffic simultaneously in congested environments. Sure, if you're in a dense area of signals it might alleviate some issues if you're close enough to the AP all of the time to use the 3rd band to negate the interference from others around you.

 

Forgot about this but I went through my Router and set it back up. I have the AX50.

Im sitting with the router in the room and the only user is me. I mostly use it for my steam streaming and plex stream.

Sitting with a 1.2Gbps connection speed at present with 160Mhz + 5Ghz band.

No updates on the router or the Ax200, might get better speeds with different drivers / firmware versions.

 

Would you mind testing with TamoSoft throughput Test, to check the real TCP throughput? I should be getting my new router and AX200 next week or so, and I would be reporting results

 

Wired 1Gb/s desktop client to Wireless Server AX1650x (13 R3)


I have more screens but for some reason NBR is goofing up on me.

In short though, WiFi to WiFi you lose throughput, wired to wireless looks like the above.

Capture6 = Desktop client wireless (ax200) to wired t440p (1Gb/s ethernet)

Never used the application before, if you want more tests let me know. I have the AX1650x in my 13 R3, AX200 in my desktop, and an old N card in my t440p (which is why I used it for server point mostly)

 

The speeds for wired desktop/wireless laptop seems pretty good, I'd be happy with those speeds. But the wireless AX200 to the wired Lenovo speeds are odd. I don't it has to do with the AX cards, I read somewhere that the AX200 and AX1650 are essentially the same chip. Maybe the wireless desktop has worst antenna rx/tx?

 

Thats because I didnt re-read your request for TCP only speeds yet, UDP was enabled so it takes a portion of the throughput

Here is the reverse above...

Desktop wired as server (1Gbps) with the 13 R3 and the AX1650x as the client

 

Why the performance is so much better when the client is the wired desktopt and the server is wireless laptop? shouldn't be the same? I mean it shouldn't matter which one is the server and which one is the client in terms of throughput or am I missing something?

 

The full screen pics are essentially the same in terms of throughput. I just reversed the roles and replayed that benchmark.

The second test was just ran long enough to saturate the chart but the first test was for about 5 minutes, had I left it on, it may have leveled off to a similar point.

 

Oh I get it, well the longer the time the test run, the more representative is the average throughput. Something I couldn't find in the Tamasoft documentation is if the software test is uploading and downloading at the same time, and since wi-fi is half duplex if the test were to run uploading/downloading only then the throughput would remain the same (i.e. ~1300 Mbps uploading / 0 Mbps downloading...... ~0 Mbps uploading / 1300 Mbps downloading .... ~500 Mbps uploading / 800 Mbps downloading) What a shame that the software doesn't allow to test one direction data flow only, which is my case most of the time.

Also there is like a 50% difference between the speed of downloading and uploading, this issue is relevant in case the throughput isn't constant across the uploading/downloading different states (downloading only, uploading only, uploading and downloading)

 

I think that it does test one direction at a time but you can confirm it easily - instead of searching through documentation just start the test and look in process manager at Wi-Fi throughput.

 

So I finally got my AX setup. Intel AX200 + Huawei AX3 Pro.

The router has solid performance for my needs, at least, the only issue I had is the firmware is in Chinese but I sorted out using the translate tool from Google Chrome. The Router has little software options but I really don't need QoS/VLANS or any advanced stuff.

I feel that the speed really got so much better compared to my old AC setup.

This is my results in Tamosoft, I let the software running for like 1 hour:



This is what I got in real world file transfer from my laptop to my remote server. The server is connected via Gigabit Ethernet and the laptop is using AX200:



I have done several tests and always get that speed or similar behavior, which is more than good to me. Previously I was getting most of the time like 20 MB/s and best case scenario 45 MB/s using AC 867 Mbps with the AC9260 (server in wired connected to gigabit ethernet as always) and not as solid performance more fluctuating speeds I'd say.

This is using the speeds I get using the laptop connected via gigabit ethernet to the same switch as the remote server which is also connected using gigabit ethernet:



The difference is not that big and maybe the mechanical HDD is being a bottleneck to the communication link. At the end of the day I'd say that the performance difference is quite high and also the uplink is more stable.

By the way this is what I see in the resources manager while using Tamosoft:



Is a bit weird isn't? Maybe Tamosoft is extrapolating data? I don't know really...but while tamosoft is measuring like 600 Mbps the resource manager is reading like 4 Mbps

 

80-90MB/s is about all a 1Gbps link will handle.

If you want more or a challenge read on...

If you want the full boost you need to enable 160mhz & qos.

When disabling qos on another router I wasn't able to hit the max speeds / link with the 160mhz channel option.

There are also some things you need to change in the advanced options of the wifi adapter as well to get the max performance out of it.

LAN performance should hit the Gigabit threshold (check) unless the router has a 2.5Gbps port or the ability to LAG/LACP 2+ ports together to provide a bigger pipe to pass traffic.

For example:
NWA210AX Zyxel AX AP has a 2.5gbps port for the data
TPE-215GI 2.5gbps POE injector / or you can source a 12v2A power adapter
CD0674 (USB) or SD-PEX24065 (PCIE) 2.5gbps ethernet adapter (for devices you want to connect to at full speed i.e.NAS, Server, etc.)
QNAP QSW-1105-5T is a 2.5gbps switch (for internal devices 4 + 1 to existing router)

This would get you end to end 2.5gbps max throughput or at least 2.4gbps on 5ghz ax / additional 575mbps on 2.4ghz

Oh, and if you're using cable there's a new modem out that has a 2.5gbps port on it - CM2000 / Dell is selling them for $269
***These are so new that the Motorola MB8611 isn't even showing up for sale anywhere yet as a competitor***

Build your own router from an old PC and throw in the 2.5gbps NIC above and the POE injector and you have a full fledged AX router that beats the speeds of most you can get off the shelf but at a bit of a premium but with better coverage w/o adding more extenders/pods to the mix.

CM2000 <2.5> BYOR <2.5> switch <2.5> AP + POE
<2.5> LAN devices
Once providers flip the switch you can have 2.5g end to end w/o having to buy thousand $ devices or pay rental fees for one. If you want to upgrade to 5gbps/10gbps service down the road it's a simple swap of your NIC for a higher 10gbps NIC that's adaptive like the Asus down below.

A little less complicated -
Asus RT-AX86U - Cheaper solutions as well at $249
(2.5 port is wan OR lan though) + switch to add more 2.5 ports to LAN devices

NETGEAR Nighthawk AX12 (RAX120) AX6000 - ~$500
This one has a single LAN 2.5/5gbps port
- there are 5gbps cards out there but, a single port 10gbps card might be more cost effective and future proof
- 5gbps to your storage might be worth it as it may have the ability to fun at 5gbps or add a 5gbps -
TUC-ET5G or NT-SS5G ($60) USB
ASUS XG-C100C ($100) 10g/5g/2.5g/1g/100m NIC

10 port 10G switch w/ POE and multiple 2.5/5/10g ports - MS510TXPP @ $349
Same w/o POE - MS510TX @ $269

 

Thanks for the good tips. In the future I'll be updating my server with a new motherboard, processor, etc. Maybe I'll take your advice and mount a virtual machine on my server and make that virtual machine a router. But that's not gonna happen any time soon, let's say I'm struggling a little bit with money and I rather update my server with new 10th gen Intel processor and RTX 3000 series. I guess there are tons of Linux distributions that emulates Cisco routers and can be installed in a virtual machine??

 

VM = Cheaper
Metal = a bit more secure

If you went USB instead of NIC limiting yourself a bit to 5G you could do a MicroPC that's the size of a coaster but still have the swapability of a PC while limited still possible to have a in/out port setup where you can enforce FW options more stringently w/o possible software / buffer issues w/ VM. Even an ITX case would work for 2 PCI cards whether you went 2.5/G or 10G since those options are cheaper than looking at dual port cards (2.5 don't exist unless dual 10G nbase-t) or 10G which avg's around $180 either nbase-t/SFP (extra $35/port and compatibility issues)

Right now I'm just running a 4 port Gig I350 in a LACP bundle but, you could do the dual 10G and that would work for in/out designations and allow flexibility down the road for higher speeds as they get released by providers.

It's all about planning before you start and deciding the priorities of wants vs needs.

For me I put several devices into a single case condensing 5-6 different things into 1. Dropped my power usage, saves space, performs better than the items did separately, and just easier to manage. Pricing wasn't too bad until I went from a single spinner for storage to a Raid 10 setup for speed / redundancy. Mind you tinkering and tweaking things lead to using 3 different mobo's / cases to get to the final result as things changed.

 

Asus RT-AX82U Gaming Router Review: Solid Mid-Range Performance, Plus RGB Tomshardware.com

Yeah, the gamers will sit enjoying the RGB glow. I hope they don't forget what happen on the screen What next? Glowing antennas?

 

As long as I can turn off or significntly dim the LEDs I'll be okay with it (once they add 6GHz). As is, it looks a bit too much like a scene from 2005's War Of The Worlds! Or a 1970s Cylon (by your command!).

 

While yes bare metal may be best, no need to waste extra space or money for hardware if you don't need to, plenty of people I know/spoken/forumed with run VMs for routing without issue even at 10Gbe, no longer is this the early 2000s where VMs were far more finicky, have sufficient resources allocated and you will be fine. If you wanna get closer to bare metal try to go for a something like ESXi, Proxmox etc and have hardware passthrough for NICs, modern Intel adapters should be able to do hardware passthrough for individual ports (haven't done it myself). For NBase-T on Ebay you can find the X550-T2 for (Dual Port) for ~$150. Yes you can get cheaper cards like Realtek ones but more hassle and relatively worse drivers and at times incompetent offloads. For routing I'd stick to Intel/Chelsio/Mellanox adapters, pretty sure almost any networking forum, routing/firewall/NAS distro devs/forums would likely discourage Realtek. Ok for endpoints/clients maybe but I'd discourage Rtk for firewall/router use as well.

 

@ Aivxtla

I threw together my 5GE upgrade and it's sufficient for the speed of the disks in a Raid 10.

I went with a QNAP QXG-5G4T-111C quad port 5GE card ~$200 shipped, swapped from the internal AP via QWA-AC2600 >> NWA210AX AP ~$230, picked up a USB 5GE NT-SS5G adapter for ~$50 for speed of copying bulk folders to the laptop since TB3 just doesn't make sense right now since the Raid is capping out at about ~400MB/s. Also picked up the new AX210 6E Intel adapter on Amazon on the cheap.

With the new 5GE Aquantia it didn't have any issues getting it up and running though the port #'s Linux assigns is a bit odd as normally you go 0->3 well this thing is 8/9/5/6 which I'm sure is something ASIC related. Split them between WAN/LAN for LACP on the WAN and just bridging the other 2 for LAN since the AP has a 2.5GE port on it for uplink / LAN.

The AP seems to have some software limitations though as the QOS seems to cap the speeds right at 1gbps so I opened a case with them to see if they have a way to override the limitation. Also the 160mhz option seems kind of iffy and is marked as beta in the configuration GUI. It seems stable though even w/ the speed limitation at the moment until they unlock the QOS option and potentially add on 80+80 in the next firmware release.

 

I got one of them bad boys in my cheap dell. One thing I find funny on this thread is the mix of AP connex being used. Hardware being compared apples and oranges here.

My WRT1900AC has never been adjusted since it's config back 3 years ago. I have a few low connection areas that some devices have issues with. Wifi cards ranging from Broadcom, Qualcomm, and Raylink. This Intel mofo, beats them hands down. I can get a connex link of 867 where others fail at getting a full connex speed up and down.

As for signalspeeds going up and down. That part was explained 3 years or more ago on AC. Different hardware and the firmwares will effect throttling between connections. The WRT1900AC, does not throttle like MIMO, so them graphs don't mean much in comparing through put, unless you are trying to identify connection issues.

Back when I configured my WRT1900AC, I found some wee changes can affect how much throughput a connection has.

Setting up unlimited bandwidth on media allows little throttling. Also selecting upper or lower end of the channel when channels are congested helps.

Then you have features that are not shared between other client adapters. Disabling some of the features keeps them in line with playing well on the network.

All in all for it being not as robust as a 4x4, it can connect at max speeds very well with the proper AP configuration, and hardware.

Not all routers are made the same, ASUS and Motorola on the AC side are pretry goOD for home use. The linksys I have does very well for mixed interplay with both MIMO and non MIMO connections.

 

@ Not-meee

I concur with your observation on the AX210 in stability. Even switching from AX200 -> 210 made a difference in stability.

Older cards like the 7 series and downward to the older models are more of a backup solution while waiting on a new card to arrive. AC was a leap forward but 6/6E are more of a rocketing head in stability and speed. From the AP standpoint switching from AC to AX boosted things ~200mbps to a full 1gbps speed in actual data not just link speed.

If you're happy with the LAN performance of the WRT and want to just upgrade the WIFI to 6 attaching an AP like NWA110AX (cheaper than 210 model) would give you a boost without having to fully replace things. Sometimes though router/wifi combos have issues on the ports you don't notice w/o texting within the house as many people don't want / need gbps ISP plans. I noticed this in the past when I upgraded to gbps service where the LAN was being limited to ~250mbps with gig ports. Swapping out the router for something else resolved the issue.

 

@Not-meee I think you are confusing MU-MIMO with MIMO... not the same thing. Practically most multi antenna devices out in the wild right now are MIMO (Multiple In Multiple Out), a few exceptions like MISO products (Multiple In Single Out) like the Chromecast 2 (2 antennas used for receive but only 1 of those at any given time is used for transmit. 2x1)

Most clients in the wild are MIMO but not all of them are MU-MIMO capable. You can also have MU capable SISO (single antenna 1x1) devices, MU just means MU capable client devices can receive data in parallel rather than one after the other ie a 4 antenna router (4x4) can transmit to 2 dual antenna devices. MU on client uplink was only added with WiFi 6 spec, in AC it’s receive side only.

Also MU on a 2 (2Rx2T) or 3 (3Rx3T) antenna router makes no real sense as most clients are dual antenna anyway, with a 3 or 2 antenna router you’d only get MU active if you had 2 single antenna devices active with a 2x2 router or 1 single antenna and 1 dual antenna for a 3x3 router. And some older MU capable routers (ie R7500v2) function in an N-1 fashion as in only 3 streams for MU.

A lot of Apple users also seem to get confused about MU and regular MIMO as for some reason Apple mentions MIMO on their product specs as if it’s something special... As far as I know no Apple devices support MU. The iPhone 12 has a MU capable Broadcom chipset but I don’t think MU is active.

Either way SmallNetbuilder did a MU test and it was found that only the Qualcomm based routers like the R7800 actually showed tangible gains, and on Broadcom routers it even led to performance losses... Granted things may have changed with newer AX BCM chips.

Broadcom clients have had issues with MU regardless of BCM or QCA chipset routers in use where dual antenna BCM clients like the Galaxy S6 or S7 would drop to single antenna mode.

As for link rates alone (which you refer to as connection speed) they don’t mean much without actually testing the real world throughput. 2 routers or clients can have varying real world transfer rates despite having same ie 866 Mbps link rate.

I also just want to add that unlike most 4 antenna (4Rx4T) routers the WRT3200ACM and WRT1900AC are 4:3Rx3T devices as in only any 3 of the 4 antennas are used for transmit/receive at any given time. The buffer of an extra antenna is for antenna diversity/optimizing signal transmit/receive.


Also it’s not really that much of an Apples to Oranges at all since people are showing what they can get out of their AX200s in their own environments with their routers/laptops, still gives a good idea to potential upgraders...and others experiences may even help in choosing what router may be better. Would be more Apples to oranges if this thread was not about a specific card...

 

Well, on my apples and oranges... it was on how comparisons are being drawn by graphs with differing AP. We all have the same card, though not all have default settings. That to me is apples and oranges.

As for 867 connex speeds... if your on a fringe connex, being three rooms away, it means alot. Being that your not so tied by distance to achieve a decent or a maximum bandwidth.

There are two ways to check throttling. One at the router, being more accurate than from a client side. Most of the traffic throttling is done at the provider. Many forget, since Corona providers limit traffic to compensate for home use activities being doubled or trippled. Even though you may have provisioned one thing, you may average well below it, with peaky tops and bottoms or ramps. My router is a media server and ftp server. Media will always be throttled, but ftp can allow maxium throughput. Try downloading a 4GB file from your home based meadia / file server and measuring to see how much spikey or rampy signaling you get.

Eliminate external connections before expecting one issue over another.

As for multiple antenna configurations... yes mine is 3 + 1, it comes in handy with mixed 2.4 / 5 signals. As for MIMO, it don't do that.

What I should have stated, is distance is not all what it is cracked up to be. Some routers are better at one thing but are not so good in all things. Can't expect perfect signals when congested, so find out what your ap is doing with wifi scanning, you may find its signal doing a ramp signal from interferance or configuration.

I have never gotten a bad issue with wifi cards, it's mostly how the ap functions. Oh, I forgot... my Samsung tablet... had to replace the wifi drIvers with a phone driver with like hardware. The driver was old and Samsung would not update the firmware.

I assume to make you AP play well, you will find waiting for a newer driver, as a cheap alternative in correcting ap connectivity, if there is an incompatibility that can not be corrected with the AP configuration.

 

@ Not-meee

It's never going to be apples to apples when comparing technologies when you're link rate on AC is 867 max and AX can do 2.4gbps with the proper card / AP. That's 300% more than AC will ever be able to do with a single client regardless of how it's configured. Even with the ability to handle AC Wave 2 @ 1.7gbps you're still stuck at a max outbound of 1gbps due to port limitations and unless you're doing wifi adhoc between 2 clients you'll never achieve 1.7gbps throughput.

You mentioned the WRT but then replied with your media / ftp server being your router.... which is it?

I setup my server and it's directly connected to my cable modem and performs as the router/wifi/vpn/firewall/media/dvr/etc. I have the flexibility with upgrading the links between devices rather than being stuck with prefabbed routers i.e Linksys / Netgear / etc..

My "router" is running an 8700K and doesn't break a sweat with most functions it performs and only really warms up when it's transcoding video files after a recording finishes and gets processed.

As to WIFI anything released in 2020/1 will be 2x2 unless it's a PCI card which has the option of going 4x4 w/ 2 per band 2.4/5ghz.

Adding tablets / phones to the mix just complicates your testing other than being able to push the STA limit upwards to test MU-MIMO but most of the chips in phones are inadequate for testing higher end throughput unless you bought it in the last 12 months due to the changing specs and HW availability in production runs. Sure there are some devices that aren't "certified" but work well, otherwise there's only a handful of handsets that will perform at WIFI6 speeds.

"Compatibility" is subjective as there are standards and approvals that need to happen before things get released into the wild. The compatibility is mixing technologies and not isolating slower speed technologies from your 5G/6G spectrum as in most cases your network functions at the slowest speed client and those clients tie up the bandwidth during transmission when they're allowed to talk which causes other devices to wait an extended period of time between bursts.

Link Rate is the proper term for the PHY rate in which your device connects to the AP. Just because your LR is 867 doesn't mean you'll get that in throughput as you know already but it is a bit of an indicator of how strong the signal from the AP to your STA is.

Throttling depends on the type of data being passed. Different data has different markings to trigger faster/slower data transmission. If you're mainly moving files back and forth you should be able to internally sustain whatever your port speeds are physically whether 1gig/10gig unless there's an ASIC issue in the connecting device in which could result in slower internal speeds or you didn't configure it for full duplex.

To throw things into contention:
For ACw1 your max LR is 867 with a real world throughput max of ~600mbps
For ACw2 1.7gbps you need to enable 160mhz spectrum - on this I was able to externally/internally test up to wire speed gigabit
For AX 5 you can hit 1.2gbps on 80mhz channels - on this I max out internally @ 125MB/s or 1gigabit due to a software limitation on the AP I'm working with the OEM to resolve due to the AP having a 2.5gbps uplink back to my server along with them unlocking 80+80 in addition to 160
For AX 6 you'll be able to exceed 3gbps

 

Well there is the rare AC router with a 10 Gbps port (SFP+) like the R9000 which lets you take advantage of HT160. You can likely hit 1120-1200 Mbps on the 9260AC or AX200 clients with that router. But yeah I agree 10 Gig was rare on AC routers and I know none that supported Multi-Gig. As you said that pretty much limited most of those to around 900-940 Mbps peak. Even on AX routers most don’t have multi gig but it’s definitely more readily available than on AC models.

AX will net you like 5-10% more over AC on 5Ghz on average and at best like 15-20%. Real gain is on the 2.4 GHz where it doubles over the N standard. Gain was more pronounced for me on HT20, on HT40 it wasn’t nearly as much probably due to interference with so many neighbors nearby. AX is more about spectrum sharing/optimization rather than raw throughput over AC.


FYI: I only forced HT40 on 2.4 GHz for testing but normally I force HT20 so as not to hog 2/3 of the 2.4 GHz spectrum and annoy my neighbors LOL plus the worse performance and drop outs at times due to interference. My router at times likes to stay at HT40 even if fallback is enabled which is kinda sad.

 

Hmm, looks like we get diminishing returns on this newest WiFi tech. I get 70MB/s on file transfers via ac/160Mhz at close (1-2m range). This is through Netgear Nighthawk EX8000 extender in AP mode via 1GE. Even though they advertise this as a "3GB" extender, it still doesn't have a 2.5GE or link aggregation to support that theoretical max bandwidth. Their top of the line "6GB" EAX80 extender also still only supports 1G Ethernet. Seems like the real performance payoff on getting the Wifi6 EAX80 at twice the price plus new network cards would be fairly marginal. In contrast, the effect of upgrade from b/g to a/c is massive: 5-10 times faster.

BTW I think the "2400Mbs" is a bit of a marketing trick. What they mean is 1200 up + 1200 down. It's like they advertise ac 160Mhz as 1733Mbs - they just mean 866 up + 866 down if I'm not mistaken.

 

That is marketing trick but it's even more convoluted than that. 1733 Mbps is actually 1733 Mbps one way (either up or down) - that is a theoretical link speed so it's not achievable in reality but it's not achieved by adding things that don't work together, so it's no more lie than say 300 Mbps 802.11n link speed.

As for 2400 it gets more complicated - AX200 is capable of 2400 Mbps link speed in one direction exactly as 9260 is of 1733 Mbps but there are 802.11ac routers on the market that are branded "AC 2400" that have nothing to do with that speed.
They just add the max theoretical speed of all radios onboard to come up with that number. An example from Asus: "High-speed WiFi performance up to 2400Mbps, 2.4G 3T3R + 5G 4T4R concurrent speed up to a combined 2400 Mbps data rate".

So 2400 Mbps is real... or it isn't

I agree with your conclusion about diminishing returns though - although I might have already agreed with you on this in a post I made yesterday. In case I already did: I still agree with you on that.

 

OK, does the same apply to ac? I have this Netgear EX8000 extender, the 1733 band is enabled on non-default channel 112, no overlap with neighbours, my wireless adapter is Killer ac1550 which claims support for 1733 bandwidth and 160MHz channel. Connection is at 866 1M from the AP.... What am I missing? Many thanks in advance

Edit: actually, this extender is triband, and both 5GHz channels are 80Mhz.... that probably explains it. There is no explicit channel width setting for that band, it just says "Speeds up to 1733Mhz). Perhaps it's a matter of channel choice - maybe some channels don't support 160Mhz?

The extender settings shows this channel info, even though I selected 112: 100 + 104 + 108 + 112(p). I guess (p) stands for preferred?

KIller Wifi analyzer is showing the following info:
Physical type: VHT.
VHT Channel Width: 80Mhz
Short Guard Interval: 80, 160, 80+80

Help lol

Edit 2: OK, so the 1733 support in that extender requires 4x4 MU-MIMO, whereas the Inhell based Killer card supports 2x2 over 160Mhz channel. Apparently, the 4x4 MU-MIMO approach is more robust, but sadly not as prevalent. Not sure if I can get a 4x4 MU-MIMO laptop card. If I want a Netgear extender, the next step up is WiFi 6 + 160Mhz at double the price. Not sure it's worth it.

Edit 3: 4x4 MU-MIMO 1733 ac is a scam. There is literally just one desktop WiFi card which supports it. Sorry for a little OT. I guess it could help someone wondering why bother with 1.2Gbps Wifi 6 if one can get 1.7Gbps using wifi 5

Might return this extender and look for some WiFi 6 router with Link aggregation and 2.5GE instead. Can I have 2 WiFi routers share the same SSID for both 2.4Ghz and 5Ghz channels?