Why You Don't Need MU-MIMO

[glugglug]

On my network are a Galaxy S6 Edge+ and a OnePlus 5T, both of which claim MU-MIMO support. Also a recently retired Galaxy S7 Edge.

After reading this article, I'm wondering... should I be disabling MU-MIMO on the RT-AC88U?

The OnePlus, at least, is consistently getting a 866Mbps link rate.

 

[thiggins]

After reading this article, I'm wondering... should I be disabling MU-MIMO on the RT-AC88U?

No need to wonder. Try an experiment and see.

 

[glugglug]

Tested, seems to depend on the device.

OnePlus 5T gets 866Mbps link speed reguardless of the router MU-MIMO setting.
Samsung Galaxy 6 Edge+ and Galaxy 7 Edge both drop to 433Mbps when MU-MIMO is enabled. However, the Android WiFi Speed Test is getting slightly better results with it *enabled* on the Galaxy S7 Edge even with the lower link rate (but horrible either way). I think there might be a latency improvement when it switches to only use 1 stream. My ping times to android devices are horrible (many > 100ms!) while I'm getting 1-2ms to Chromecasts and Rokus.

 

[thiggins]

Thanks for posting the results. You are seeing results similar to those in the ariticle, i.e. might help a little bit, regardless of fallback to single stream.

 

[avtella]

I found this old article where they used a QCA chipset router and saw decent benefits from MU:
https://www.google.com/amp/s/www.fo...looked-wireless-upgrade-over-the-nexus-5/amp/

 

[digital10]

So is MU-MIMO snake oil or has real benefits now? The article from 3 years ago says no signifcant advantage but what about now? If it doesn't hurt shall I just keep it on? I have no way to test this. Using Netgear Orbi.

I shut it off because I believe the more "auto-tune" stuff working on the router the heavier and more issues you might encounter.

 

[thiggins]

The Broadband Forums TR-398 test suite includes a MU-MIMO test that looks for total throughput gain using three MU-MIMO devices. I have seen it show modest gain in some tests.

I have not gone back and re-run my tests, not do I intend to. But my upcoming test suite will ensure that MU-MIMO and OFDMA are enabled in products that support it. The new suite will include multi-STA tests that should show a performance difference if the product produces one. I just won't be running specific A/B tests on those features.

Both OFDMA and MU-MIMO add overhead and can cause compatibility problems with older devices. If you're curious, run your own tests and decide for yourself. But most people won't go through the hassle of this. You need at least two devices that support the features and must run devices simultaneously during the test.

 

[digital10]

The Broadband Forums TR-398 test suite includes a MU-MIMO test that looks for total throughput gain using three MU-MIMO devices. I have seen it show modest gain in some tests.

I have not gone back and re-run my tests, not do I intend to. But my upcoming test suite will ensure that MU-MIMO and OFDMA are enabled in products that support it. The new suite will include multi-STA tests that should show a performance difference if the product produces one. I just won't be running specific A/B tests on those features.

Both OFDMA and MU-MIMO add overhead and can cause compatibility problems with older devices. If you're curious, run your own tests and decide for yourself. But most people won't go through the hassle of this. You need at least two devices that support the features and must run devices simultaneously during the test.

can't I just test with MU-MIMO on and off and see the difference?
What software should I use?

 

[thiggins]

can't I just test with MU-MIMO on and off and see the difference?
What software should I use?

Sure. But unless you set up a consistent test scenario, you won't really know what you are testing.
iperf3 is most widely used, but there is a learning curve.

 

[RMinNJ]

:..But my upcoming test suite will ensure that MU-MIMO and OFDMA are enabled in products that support it. The new suite will include multi-STA tests that should show a performance difference if the product produces one. I just won't be running specific A/B tests on those features..."

Can't wait to see your tests of all these routers..

 

[Krisbi]

So, Aruba tested its own access points some time ago: AP-515, AP-535 and AP-555 which all support 4x4:4 on 5 GHz. There´s hardly any benefit using MU-MIMO at all compared to SU-MIMO:

 

[L&LD]

@Krisbi, those graphs look like there are benefits, to me.

 

[Krisbi]

Well, if I look at AP-535 from the left:
- 50 clients with 802.11ac (wave 1) and SU-MIMO: 590 Mbit/s

AP-535 from the right:
- 60 clients with mixed 802.11ax/802.11ac (wave 2) and MU-MIMO: 528 Mbit/s

Those 50 "old" clients outperform the 60 "modern" clients; having +10 more clients on the access point results in less perfomance even with MU-MIMO?

 

[L&LD]

Of course.

More clients mean more overhead, less work completed.

 

[thiggins]

Of course.

More clients mean more overhead, less work completed.

Not if MU-MIMO is working as intended. It's supposed to handle multiple (up to 3 or 4, depending on the chipset version) in one frame.

There is throughput improvement between the all SU and all MU cases, with the 555 showing the highest gain.

Comparing all SU to mixed MU/SU shows negligible gain for the 515 (4%) and 555 (2%) and decrease (11%) for the 535.

Using so many STAs actually doesn't help make the case, since maximum gain occurs with only 3 (or 4) STAs.

 

[Tucu]

How do they handle MU-MIMO with 2x2 and 3x3 clients? Do the clients switch to 1x1? If not, these tests are not really testing MU-MIMO given that the APs are only 4x4.

 

[Krisbi]

For the test with 60x 802.11ax clients Intel AX200 (2x2:2) is used.

AP-515 (4x4:4)
AP-535 (4x4:4)
AP-555 (8x8:8)

According to https://www.qualcomm.com/products/networking/networking-device-finder/aruba-ap-535 AP-555 uses "Qualcomm Networking Pro 1200 Platform" Which IPQxxxx would this be?

AP-535 uses the same "Scorpio" firmware as AP-555 on Arubanetwork's support homepage. So it might be a stripped down "Networking Pro 1200", aswell.

- AP-515 uses BCM43694

 

[thiggins]

How do they handle MU-MIMO with 2x2 and 3x3 clients?

3x3, don't know.
BRCM drops 2 streams to 1, or at least that's what they've done in the past. I suspect the few 3 stream STAs out there also drop streams.

MU-MIMO works best with single stream. When you get to 2 streams, things get more complicated.

Qualcomm did a better job with MU-MIMO with AC chipsets. I'm seeing signs that MU-MIMO is handled better with AX chipsets,.

At any rate, MU-MIMO is most useful when the channel is fully loaded. This can happen with as few as two AC STAs, depending on bandwidth and application in use.

 

[sfx2000]

3x3, don't know.
BRCM drops 2 streams to 1, or at least that's what they've done in the past. I suspect the few 3 stream STAs out there also drop streams.

MU-MIMO works best with single stream. When you get to 2 streams, things get more complicated.

Qualcomm did a better job with MU-MIMO with AC chipsets. I'm seeing signs that MU-MIMO is handled better with AX chipsets,.

At any rate, MU-MIMO is most useful when the channel is fully loaded. This can happen with as few as two AC STAs, depending on bandwidth and application in use.

I would agree in the early days of Wave2 support for MU-MIMO...

Now that we have multiple vendors on the client side with Wave2 support for MU on 11ac - it's less of an issue that it was a few years back with Broadcom and Quantenna...

QC-Atheros was never really a problem on the AP side for their Wave2 chipsets on the AP side

IPQ40xx - which is in most of the mesh solutions on the home/consumer market - they're 2*2:2 solutions with MU support - and recent versions of QSDK have tuned things there.

PC Client side - Realtek, QCA, Intel - getting better - and Broadcom on the mobile side is getting better as well.

Enabling MU is the past was perhaps a performance hit, but these days, it's not so much...

Interesting to note however - Apple still doesn't do MU on their 802.11ac/ax clients, and that says much about their thoughts.

 

[digital10]

is there an update on this?

is MU-MIMO good now or still will give degrading performance and problems with the wifi network?