Why You Don't Need MU-MIMO

  • ATTENTION! As of November 1, 2020, you are not able to reply to threads 6 months after the thread is opened if there are more than 500 posts in the thread.
    Threads will not be locked, so posts may still be edited by their authors.
    Just start a new thread on the topic to post if you get an error message when trying to reply to a thread.

sfx2000

Part of the Furniture
It gets better. Apparently some 18 companies in a prticipating SIG ("DensiFi") got caught colluding, which set things back a bit. Fun reading if you have nothing better to do. I see that the Wikipedia entry for 802.11ax has had the incident edited out.

I've seen this stuff before - 802.16 had similar concerns (successfully resolved) and 802.20 (where things got very bad - the group was frozen, the chair removed, and voting changes were put into place as entity votes vs. the normal member votes)
 

Razor512

Very Senior Member
With MU-MIMO providing some benefit in most cases, isn't a router with it still objectively better than one without it if it does not have much impact on the price?

Also, I was wondering, is it possible to implement features such as 802.11v and 802.11r and have them work across brands and models of access point? For example, allowing a user to have a high end wireless router, and then later buying an extra access point or 2 (possibly from a different OEM with a different wireless chipset) but have functions like 802.11v and 802.11r work seamlessly between them.

Aside from doing a complete replacement, there does not seem to be any system that will allow a user to buy an additional AP to expand their coverage for situations where a single wireless router is not providing enough coverage, and then allow the user to seamlessly move from one part of the house, to another and not have the connection drop for a few seconds while it switches?
 

sfx2000

Part of the Furniture
Also, I was wondering, is it possible to implement features such as 802.11v and 802.11r and have them work across brands and models of access point?

11v/11r - just describes how one could do it - doesn't define the actual implementation - keeping things in the same vendor is the best bet there.

Similar to MU at this stage - similar vendor chipsets can and often do perform better - as all are reading from the same spec sheet. The challenge, which was brought out in the article - is when one team reads things on the AP/STA side, and another team reads things from a Client/STA side - and those differences show in the results.

MU is hard stuff - it really is...
 

RMerlin

Asuswrt-Merlin dev
Tim, do you know if your RT-AC88U used the BCM4366 or BCM4366E? First one has non-working MU-MIMO, so its test results aren't really indicative of how a properly working MU-MIMO might behave.
 

thiggins

Mr. Easy
Staff member
Tim, do you know if your RT-AC88U used the BCM4366 or BCM4366E? First one has non-working MU-MIMO, so its test results aren't really indicative of how a properly working MU-MIMO might behave.
It's one of the first produced, so has whatever ASUS used when it first started producing the product.
But it's moot, since even the results using Qualcomm AP and Qualcomm STA, which should be "best evidence" show marginal benefit.

You really think Broadcom AP and Qualcomm STA would be better?
 

sfx2000

Part of the Furniture
Getting back to the origin - MU capable AP's are there - and where I find them to perform best is not in MU mode... there's benefit with the updated silicon for SU use.

MU is a real pain...
 

RMerlin

Asuswrt-Merlin dev
It's one of the first produced, so has whatever ASUS used when it first started producing the product.

So it's the original BCM4366 which has non-working MU-MIMO.

You really think Broadcom AP and Qualcomm STA would be better?

I don't know, and quite frankly I've also buried MU-MIMO a long time ago as yet another technology that looks great on paper, but is virtually impossible to implement in a reliable manner in hardware/software. Tossed on the same pile as NitroQAM.

A BCM4366E results would be better than your original RT-AC88U results, but I don't know if they would be better than the others. If they are, it's probably not by a meaningful difference.

I was asking because someone I was talking to earlier today was looking into upgrading his laptop's NIC to get MU-MIMO support, and I was actually telling him it was pointless, and to have a look at your article. After seeing it, he asked me whether your BCM test results were with the BCM4366 or BCM4366E (since he uses a BCM-based Asus router).
 

abj

New Around Here
Very interesting read! Thanks for the write up! MU-MIMO has been a buzz word for a while now, and I'm glad that you shed some light on reality.
 

L&LD

Part of the Furniture
Tim, thank you for the article and additional info on MU-MIMO.

I am concerned though with tests involving handheld devices. That may be the most common scenario for a lot of people, but I would be interested in tests with current routers with current hardware chips and revisions and more powerful laptops instead. Also in a more realistic setting. I don't think 8' is far enough.

What you presented was eye-opening though. Certainly explains a few customers' experience with their mobile phones and tablets when expecting at least the same if not better than what they had before they upgraded their networks.

Looking forward to an update on this issue, when you have time. Thanks.
 

thiggins

Mr. Easy
Staff member
I plan no further testing on MU-MIMO. It's a dead technology IMO. It will get another shot with 11ax, when devices eventually appear that support it in a few years.
 

thiggins

Mr. Easy
Staff member
SFX can probably answer this better than I. But it does not appear to be, from its requirement of very large numbers of antennas.
 

sfx2000

Part of the Furniture
Massive MIMO is very different from MU-MIMO as applied to WiFi.

MU-MIMO in 4G/LTE is very different than it is with WiFi.
 

Cosmin Cazan

New Around Here
(I'm a bit late to this article, but..)

Tim - Very nice article and I agree with the MU-MIMO sentiment (especially when it comes to 2x2).

However - I'm surprised you didn't go further with 1x1 in this instance (more devices). the 62% improvement with two 1x1 devices in MU mode was not bad. I think three 1x1 devices would have done even better (maybe even upwards of 100% improvement).

I'm currently thinking that MU-MIMOs problems have (mostly) been related to both QCA and BRCM chipsets having difficulty with 2x2 clients (especially since most APs are just 3SS for MU-MIMO). Would you agree that's a fair assessment? Or have you seen MU-MIMO degradation in 1x1 also?
 

thiggins

Mr. Easy
Staff member
That article is the last testing I've done on MU-MIMO.

Let me put it this way, if you get a router and devices with working MU-MIMO, it's most likely to provide total throughput gain with single stream STAs.

But whether you actually notice the benefit in real life is doubtful. It's certainly not something where you will think "Wow, that MU-MIMO is great!"
 

sfx2000

Part of the Furniture
Getting back to the original post on the thread - for the home market, I'm not seeing this as a must-have or needed feature - e.g. don't go out of the way to choose a device or AP just because of MU.

If it's there - give it a try, but know that like mentioned, even with just two MU capable stations, it's going to be hard to notice any difference, and there can be a hit on performance depending on client chipsets and the AP chipset used (including firmware for that chipset).

In my experience - MU is best for devices that don't move around - so desktop, set top boxes, maybe... devices that move around like tablets and mobile phones - MU isn't helpful there, as the sounding matrix needed for beamforming and group membership decisions for the MU frames incurs a lot of overhead and airtime, and this can impact performance, even for the clients that are not MU capable.

The engineering effort to make MU work in WiFi is pretty impressive - I'm not knocking the technology - if one looks at the math alone, it's a big step in processing requirements in real time - and that also contributes to better SU performance if MU is disabled. If one looks at the earlier AC1900 Wave 1 AP's, the current crop of AP's on Wave 2 silicon do perform better, and the results generally show, even with older client chipsets.
 

sfx2000

Part of the Furniture

I think one of the key takeaways from 802.11ac MU is below:
  • 802.11ac introduced DL MU-MIMO, but we’re experiencing the following issues:
  • Many client devices are single antenna, and many two antenna clients switch to single stream mode for DL MU-MIMO for protection against interference
  • With 4 antenna AP, gains compared to Single User are modest
  • Even if we built an 8 antenna AP, groupings are limited to 4 users
  • Channel sounding responses from the users are transmitted serially in time resulting in high overhead
  • TCP/IP on downlink with TCP ACK on uplink is impaired with no UL MU enhancement
 

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top