Goodbye To Wi-Fi Router Classes

[thiggins]

The Wi-Fi class system has outlived its usefulness. So we have switched to a new method to categorize these products.

Read on SmallNetBuilder

 

[kenrok1]

More useful than any number on the box - Thanks!

 

[Razor512]

Did Linksys really list the Velop as AC6600 and not change it after realizing the mistake? I could understand a router being listed at AC1900 where while no single device can use both bands, multiple device can connect to the single router and eventually use both bands. but the Velop, are separate devices.

That is like a car company selling 2 cars, each with a 200 horse power engine, and then marketing the 2 car pack as having 400 horsepower. Or think how some TVs use the class system (50 inch class), imagine a box with 2, 25 inch TVs in the box, and having it marketed as a 50 inch class TV.

 

[thiggins]

Did Linksys really list the Velop as AC6600 and not change it after realizing the mistake?

This was not a mistake. It is a conscious marketing decision. This is printed on the product box and plainly shown on Amazon.

 

[irev210]

This is much welcomed! Thank you, this makes much more sense.

 

[maxbraketorque]

I think its great to provide clarity on the number of streams used in the testing. I am confused thought about the results provided for the version 9 Router Charts. Have all the routers listed now been tested using the revision 9 technique?

 

[thiggins]

I think its great to provide clarity on the number of streams used in the testing. I am confused thought about the results provided for the version 9 Router Charts. Have all the routers listed now been tested using the revision 9 technique?

The Test Method selector shows the test method used for the products shown.
I am considering adding a class or streams filter for the Charts.

 

[maxbraketorque]

The Test Method selector shows the test method used for the products shown.
I am considering adding a class or streams filter for the Charts.

That's an impressive amount of router testing to get that many tested under revision 9 already. Thanks.

My understanding is that revision 9 already tests everything using a 2x2 client. Is that incorrect, or are you thinking of adding a streams filter for older test methods?

 

[thiggins]

My understanding is that revision 9 already tests everything using a 2x2 client. Is that incorrect, or are you thinking of adding a streams filter for older test methods?

Your understanding is correct.
https://www.smallnetbuilder.com/wir...2993-how-we-test-wireless-products-revision-9
https://www.smallnetbuilder.com/wir...32478-how-we-test-wireless-products-revison-8

The "streams filter" would apply to all Test Method views. It would be like the current "Type" selector to show only the selected type of product. Except it would filter by class, i.e. AC1900, AC2600.

 

[maxbraketorque]

I definitely like this way of comparing. Considering that the client is using a 2x2 antenna, one thing that's surprising is that many of the 4x4 routers do better than the 3x3 routers, even among the same brand of products. What's the reason for this?

 

[thiggins]

I definitely like this way of comparing. Considering that the client is using a 2x2 antenna, one thing that's surprising is that many of the 4x4 routers do better than the 3x3 routers, even among the same brand of products. What's the reason for this?

One of the main reasons we changed the method was to see if more expensive routers really do help improve performance.

Explanations include higher receive gain due to the additional RF chain (antenna). Possibly also more CPU power in the router.

Sort by price to find best price / performance. Be sure to look into the details, including sub-ranks before buying. Could be the thing boosting the ranking is something you don't care about.

Also keep an eye on the ranking score itself (the small number under the rank #). For example the difference between the #3 NETGEAR R9000 (38.6) and #4 NETGEAR R7800 (39.3) is pretty small. But price difference is big.

 

[maxbraketorque]

Sorry to keep asking questions, but there are some pretty large changes in wifi performance for some routers between rev 9 and whatever revision was used for the original router test. A great example is the R7800. For it's original test using Revision 3, 2.4 GHz downlink was fine, but for Revision 9, 2.4 GHz downlink is rather poor. Conversely, 5.0 GHz downlink is typical in Revision 3, but unbelievably good for Revision 9. Any comments as to what's happening here?

2.4 GHz Downlink Revision 3

2.4 GHz Downlink Revision 9

5.0 GHz Downlink Revision 3

5.0 GHz Downlink Revision 9

 

[thiggins]

There is no wireless "Revision 3" test method. I think you are comparing results from Test Method 8 and 9.

The test client in Revision 8 was a 3x3 NETGEAR R7000 in bridge mode. The Revision 9 client is a 2x2 Intel AC 8265.

 

[maxbraketorque]

There is no wireless "Revision 3" test method. I think you are comparing results from Test Method 8 and 9.

The test client in Revision 8 was a 3x3 NETGEAR R7000 in bridge mode. The Revision 9 client is a 2x2 Intel AC 8265.

Thanks on clarification of the Revision. Aside from going from 3x3 to 2x2, I don't understand how the change in client would affect the results so strongly for the differences shown for the R7800. Are you suggesting that client hardware and drivers can that strongly affect router performance?

 

[thiggins]

Are you suggesting that client hardware and drivers can that strongly affect router performance?

Yes, of course.

 

[maxbraketorque]

Yes, of course.

Thanks. Very fascinating, and considering that there a so many different clients out there, it suggests some uncertainty in how well your test results will represent real world performance. There has to be some metric though, and your tests are the best I've seen so far, so still very valuable.

 

[sfx2000]

The Wi-Fi class system has outlived its usefulness.

Speaking of which - on the inflated Vendor numbers that kind of caused this - has anyone actually observed a TurboQAM connection in 2.4GHz?

This was one of the key non-standard items - vendors would quote the TurboQAM/QAM256 numbers for 2.4GHz...

 

[hellboy2]

How does this new Revision measure the 1:n ratio of one router and multiple wifi devices in the house?

My home has 1 router and typically between 12 and 20 devices connected. Most of them are N-class devices connected at 2x2 144 mbit, and some of them are AC-class devices connected at 300 mbit or more. Several devices have been tested with www.speedtest.net, where N-class devices usually reach 30-50 mbit/sec, and AC-class devices reach >100 mbit/sec, which is my ISP's bandwidth limit. So far so good.

The real usage is 2-3 concurrent streams (Netfix, NetTV, Twitch - 1-4 mbit/sec per client), and some times 4-5 teenagers playing games, where ping ratio is more important than raw speed throughput. None of the clients need much from the router, but their combined load should be handled by the router.

Isn't it more realistically important that the router can handle all the clients at once? Each client may be some 2x2 radio device, but then the router's ability to serve them on spearate antennas would look like 2 devices at 2x2 could be better served by one Router with four antennas.

I don't know much about Routers and radios, so it is possible I have just misunderstood the whole thing.

 

[thiggins]

The current and past test processes do not test wireless capacity, which is what you are referring to.

 

[crashnburn]

Appreciate the new methodology. Is there any way we can still do a Price vs Performance HEAT MAP (found it in Old Tools) for newer Devices that you test?

Also, not sure if its therein the test TB, but Wirecutter ranked Archer C7 as their best pick and a Netgear model as their best picks (for most people). Would love it if you can dissect some of their stuff and give us theirs vs yours to help us pick.