MLO performance and Limitations

[snakebite3]

Is has been over a month since the feature has been available yet the discussion on the feature is almost nonexistent. Like 100% of the reviews are outdated and hasn't been updated with the new MLO firmware. What's the deal with the MLO feature? Where are the performance numbers? What's the limitation having it enable? Does IOT wifi still work? Does it live up the hype or bust?

 

[jixiangyuan]

It's crap at the moment (GT-BE98).

From my testing, I have both MLO compatible devices and legacy devices droppoing to use 2.4GHz only, and there's no latency improvement (at a non-congested location) with very little performance difference.

One of the limitations is that you can't separate each band after enabling MLO (although if you were using smart connect before, this makes no difference to you). I still don't know why they don't leave all bands separated for those who need it and create a new separate SSID for MLO.


Hopefully ASUS will release another firmware update for the BE98 next week.

 

[jzchen]

At least for the BE98 Pro when you toggle the MLO switch from OFF to ON there's a pop up asking for two SSIDs (and passwords). The first is the MLO SSID, which should be different. The 2nd is the Legacy SSID which all non-WiFi 7 devices can connect to. You can then create IoT/Guest Networks in Guest Network Pro which enforce VLANs on each one you create...

 

[baller54]

I've been testing MLO on my BE96U for a bit now, and so far it's OK.. Hard to definitively say since I only have 1 MLO device right now (Pixel 8 Pro), and it's kinda nerfed as it can only connect to 2 bands on MLO, not all 3.

I set MLO up after the fact not out of the box, so I just created a separate MLO SSID under Guest Network Pro. I have 2 other SSIDs, 1 IoT, and my main SSID for everything else. Haven't had any issues w/ device disconnects.

MLO Performance on stock firmware didn't yield better results vs. non-MLO, but after flashing Merlin's Beta firmware it's definitely improved. I'm consistently seeing 1.5 down 1.4 up from my phone. Have hit 1.8 down max (2 gig connection) twice. Should be faster, but I'm chalking that up to Pixel's horrible samsung radio and/or firmware.

Not being able to customize which bands we can use for MLO is definitely limiting, hopefully that'll come in future releases. My phone typically connects to 2.4 + 6ghz bands on one side of the house, and 5 + 6hz bands on the other side of the house. Sometimes it'll connect to a single band only .

I also think Smart Connect is buggy in this latest stock firmware though because the values are strange on the 6ghz band, and won't save when i change them. Have seen a couple others report this

 

[sfx2000]

I've been testing MLO on my BE96U for a bit now, and so far it's OK.. Hard to definitively say since I only have 1 MLO device right now (Pixel 8 Pro), and it's kinda nerfed as it can only connect to 2 bands on MLO, not all 3.

I set MLO up after the fact not out of the box, so I just created a separate MLO SSID under Guest Network Pro. I have 2 other SSIDs, 1 IoT, and my main SSID for everything else. Haven't had any issues w/ device disconnects.

Starting to sound like MLO is more of the promises like MU-MIMO, OFDMA, etc....

All of these features introduce a level of complexity that has little real-world benefit.

To be honest - WiFi4 and WiFi5 brought forward the best bang - 802.11n in 2.4 has pretty decent performance on 2*2:2 devices, and 802.11ac in 5GHz has been the best bump forward - 11ax and 11be add some features, but it's not really paying off compared to 802.11ac, it's incremental at best...

Probably why the RT-AC68U was still relevant for Wireless...

 

[RMerlin]

Starting to sound like MLO is more of the promises like MU-MIMO, OFDMA, etc....

MLO is a major improvement that is measurable in real life, it`s not just situational/theorical improvements like MU-MIMO or OFDMA were.

 

[Tech9]

My realistic expectations - MLO won't be supported on all Wi-Fi 7 APs/routers and not all Wi-Fi 7 devices will be MLD. Otherwise good demo.

 

[pdixon0]

Well here's my real world experience using MLO.

I have 3 MLO capable devices, my S24 Ultra and 2 gaming PC's I've retro-fitted with Intel BE200 cards.

Now the PC's are running the Intel 23.60.1 driver they are running MLO correctly. 1 PC is on the 1st floor (middle floor) of our 3 story Townhouse the 2nd is on the 2nd floor (top floor) while my BE98 is on the ground floor (bottom floor) on the opposite side of the house to the PC's as that's where my Fibre connection enters the house. (I am looking to move the router to the 1st floor but I need someone to route some CAT6A in the wall and between floors).

Anyway, the PC on the 2nd floor gets quite poor signal on 5Ghz, 6GHz doesn't even display as an option and of course 2.4GHz has excellent signal, but obviously poor throughout. On MLO it shows as connecting to 2.4GHz and my 5GHz-2 (DFS channel 100 - 160MHz). This machine, when on 5GHz-2 only will see 1-2 bars on the Windows taskbar, now on MLO it sees 3 or 4, which I know doesn't really mean anything - however, it's pings are in the high teens / low 20's and it sees around half of my 1Tb download speed and nearly all of my 100Mb upload speed. It's the pings that have improved the most being on MLO for that PC, albeit they were mainly in the high 20's to 50's on occasion which still isn't bad.

My 2nd PC (my machine) is luckily able to use the full 360 MHz 6GHz channel (69) with 3-4 bars on the Windows taskbar, but sometimes it will connect to 5GHz-2 as well (so probably 160MHz on 5GHz-2 and 160MHz on 6GHz), that machine sees my full ISP bandwidth on both upload and download with pings in the mid-teens.

This is a far sight better than my previous router, an ASUS RT-AX88U Pro, which required an RP-AX58 to get decent connection for the PC on the 2nd floor, but with the obvious hit on the ping. The throughout was roughly the same.

 

[sfx2000]

MLO is a major improvement that is measurable in real life, it`s not just situational/theorical improvements like MU-MIMO or OFDMA were.

MU-MIMO and OFDMA also have benefit, just the same as MLO...

The improvements over time have been incremental, getting us ever closer to the shannon limit, but this also comes at additional cost due to the complexity of these features.

FWIW - I'm still on WiFi5 for my primary wlan deployment - for development I do have 11ax as a mature project, and recently acquired a QCAtheros WiFi7 development board with three radios (2.4, 5, 6) along with a PCIe client adapter...

It's hard to objectively measure performance with my WiFi7 dev platform, as I only have the single client - MU-MIMO and OFDMA features need multiple clients so that the AP can actually trigger those features, otherwise it agressively sticks with SU-MIMO and OFDM - and I'm in somewhat of a noisy environment, so I rarely see QAM4096 used at all.

1 meter away, one can see benefit on the bench, but 10 meters away thru three walls, and 11ac, ax, and be are all about the same in the real world...

 

[sfx2000]

MLO is a major improvement that is measurable in real life, it`s not just situational/theorical improvements like MU-MIMO or OFDMA were.

IF using MLO across 2.4 and 5GHz - erm, nope, that ain't going to work...

MLO might work better across two radios - in the same band - e.g. 5Ghz or 6Ghz, but even there, if you split the bands, it becomes a bit of a problem for the client radios perhaps...

I do say "might" as this is heavily dependent on the client station capabilities along with the router resources.. a single client sta for bench testing might be one thing, but if one has 25-40 clients on a radio, then MLO isn't really an option is it?

 

[RMerlin]

if one has 25-40 clients on a radio, then MLO isn't really an option is it?

MLO is actually going to help there, since clients connected through MLO won't be stuck to a single band - if that band is currently clogged by a client doing a lot of transfer, then the client can use another band to do its own concurrent transfer, where there is more airtime available. That's why they say that MLO can improve latency - if having difficulties getting airtime allocated on one band, you can get that allocated on another band.

 

[sfx2000]

MLO is actually going to help there, since clients connected through MLO won't be stuck to a single band - if that band is currently clogged by a client doing a lot of transfer, then the client can use another band to do its own concurrent transfer, where there is more airtime available. That's why they say that MLO can improve latency - if having difficulties getting airtime allocated on one band, you can get that allocated on another band.

Doesn't work like that - with MLO, one now has to reserve hostapd AID's for MLO based clients in case they may be a candidate for the connection across two radios.

In an environment where capacity is a priority, MLO should likely be disabled...

Esp for dual-band AP's with only two radios, as in my testing, an MLO connection with 2.4 and 5Ghz doesn't add enough to justify the hit...

 

[sfx2000]

Doesn't work like that - with MLO, one now has to reserve hostapd AID's for MLO based clients in case they may be a candidate for the connection across two radios.

And in my testing so far - MLO is a bit of a hot-mess...

1) What happens when the Client STA is in good RSSI for 2.4, but poor RSSI for 5GHz - this happens quite often

2) What happens with a Client STA in MLO mode roams in an ESS with multiple AP's - we can get into a situation where 2.4GHz is camped on AP1, and 5GHz radio wants to move to AP2 - this is now a problem as how do you manage the upper layers in the stack.

3) MLO only happens when the SSID and Security Creds are the same - if one has different SSID"s per band, then MLO isn't going to work.

WRT items 2 and 3 - common SSID across the AP's and Bands makes a lot of sense...

Going back to Item 1 - this is about resource management across radios - this would even occur on 2 radios in the same band...

The schedulers in the AP (and in the case of multiple AP's the WLC) need to take this all into consideration - imagine a cohort of clients that a mixed base of capabilities - some might in 802.11n with WPA2 in 2.4, some might be 802.11ax/be with all the good stuff - how to manage this across 32 to 50 clients per radio, and how much do one reserves for MLO in a high traffic environment - does was one reserve 20 percent perhaps for MLO candidates (and those are not real connections, just possible ones).

IN the current state of the art - we barely have schedulers that can handle UL/DO-MU, OFDMA...

As I mentioned earlier - it does seem like MLO is the 3D-TV feature of WiFi7...

And also consider the use case of a mobile phone - power is everything - why would one add an additional radio considering the the constraints above?

And of course, there's the business aspect - who is going to pay for all that testing across bands, including DFS channels in 5G, and the whole managed spectrum stuff in 6G - yeah - it's a mess...

And let's not talk about the regulatory testing cost, plus internal QA - yeah, that's a spendy amount...

 

[Tech9]

What's the deal with the MLO feature?

Will be used mainly for marketing purposes. This is the deal.

 

[Clark Griswald]

As I mentioned earlier - it does seem like MLO is the 3D-TV feature of WiFi7...

That statement puts MLO into a perspective most will be able to relate.