Linksys EA8500 Max-Stream AC2600 MU-MIMO Smart Wi-Fi Router Reviewed

[sfx2000]

Review was done with 166845 firmware, which was installed on the router and the only one released to the wild AFAIK on production (not beta test) product.

5 GHz channel width was left at its default Auto, which enables 80 MHz bandwidth mode. I've corrected that in the review.

There is nothing that says an AC2600 class (or any other class for that matter) router must have the ability to force mode and bandwidth mode.

For most folks, anything aboveAC1900 class is at best AC1900 class...

Just saying - there's a distinct lack of clients that could do more...

edit - damn autocorrect...

 

[ddarko]

Review was done with 166845 firmware, which was installed on the router and the only one released to the wild AFAIK on production (not beta test) product.

5 GHz channel width was left at its default Auto, which enables 80 MHz bandwidth mode. I've corrected that in the review.

There is nothing that says an AC2600 class (or any other class for that matter) router must have the ability to force mode and bandwidth mode.

Thanks for the clarification. I can tell you that the routers being sold in retail and online are shipping with the earlier 166556 firmware installed on them. I bought two copies, one from Amazon and one from Best Buy, and both came with 166556 on them (fyi, the Amazon EA8500 comes in a smaller box, presumably to save weight).

Although I do have a laptop with a 3-stream 802.11ac antenna, I'm not interested in forcing an 80 MHz channel. When I didn't see 80 MHz as a manually selectable option, I was alarmed that it meant 80 MHz channel width was not supported at all. It has been a long time since I had a Linksys router and I guess I am not familiar with the way they label their settings. I did not realize the 80 MHz channel mode would "hidden" so to speak behind the Auto channel mode. I assume the same goes for 802.11ac only mode, which is selected by setting the router in "mixed" mode? I am used to Netgear and Asus routers where these options are explicitly spelled out in the firmware settings and assumed it was the same with Linksys.

 

[chadster766]

Thanks for the clarification. I can tell you that the routers being sold in retail and online are shipping with the earlier 166556 firmware installed on them. I bought two copies, one from Amazon and one from Best Buy, and both came with 166556 on them (fyi, the Amazon EA8500 comes in a smaller box, presumably to save weight).

Although I do have a laptop with a 3-stream 802.11ac antenna, I'm not interested in forcing an 80 MHz channel. When I didn't see 80 MHz as a manually selectable option, I was alarmed that it meant 80 MHz channel width was not supported at all. It has been a long time since I had a Linksys router and I guess I am not familiar with the way they label their settings. I did not realize the 80 MHz channel mode would "hidden" so to speak behind the Auto channel mode. I assume the same goes for 802.11ac only mode, which is selected by setting the router in "mixed" mode? I am used to Netgear and Asus routers where these options are explicitly spelled out in the firmware settings and assumed it was the same with Linksys.

I should clarify that (AFAIK) AC mode with Linksys routers is actually Mixed or A/N mode with Auto Channel Width.

 

[sfx2000]

It's been my experience with any AC class device on 5GHz, just leave them in auto-mode - the chipsets are smart enough to sort themselves out.

 

[TonyH]

Hi,
Anyone ran speed test from LAN and 2.4 or 5GHz bands? Here this the only router reporting fster sped than
my 50/3 I have. It is consistently giving ~57/3.7 Mbps always. No other router did it. They always gave 50/3
or little less. ????!

 

[chadster766]

Have you had a chance to do more testing with the EA8500? I'm noticing that LAN ping times start increasing with network traffic which may account for some delays seeing on network TCP communications. Can MU-MIMO scheduling somehow overflow onto LAN devices?

It turns out that my issue was on the ISP side of things.

Honestly my family of power users (IT Admin\Dev, Gamers, Streaming Media and Social Media ) are reporting that since installing the EA8500 our network is working better than ever.

 

[System Error Message]

when it comes to qualcomm snapdragon performance the krait cores only beat the cortext ARM a9 (which broadcom uses) but not A15 however some snapdragons include the stock ARM Cortex variants. I have the snapdragon 810 developer tablet which has very nice specs and performance but it uses the standard 64 bit arm variant of the ARM A9 and A15s instead of krait. If the snapdragons in routers did retain the GPU they could also use the GPU as an offload processor letting the CPU handle other things. The issue with MU-MIMO is when you have multiple clients and one of them isnt MU-MIMO which is why i chose the AC3200 instead as it would be more practical. When it comes to optimisation qualcomm does a good job at it as evident as their development of krait.

Routers dont need 64 bit CPUs do be much faster. Some vendors have hardware encryption acceleration instead which would mean faster VPN and such. Forwarding of traffic doesnt require the CPU to have more bits but rather be faster at passing data/addresses. If a CPU has to handle data calculations bigger than its bit than it will use a cycle or 2 more than handle it so clock speeds are also important. The better the logic IPC, bigger cache and clock speed of the CPU the faster it is at networking. Even with 64 bit arm it is not necessarily faster than a 32 bit x86 CPU at the same clock speed when it comes to VPN and firewall because many intel and AMD CPUs have built in encryption and SSE so they can handle bigger data types using less clock cycles since VPN uses 128-256bits or higher crypto.

Any wifi AP with more than AC1900 divided into multiple streams or channels can handle multiple clients much better. In my opinion MU-MIMO can work better with multiple clients automatically while using multiple channels are a better option if you place more control over the network division.

 

[chadster766]

when it comes to qualcomm snapdragon performance the krait cores only beat the cortext ARM a9 (which broadcom uses) but not A15 however some snapdragons include the stock ARM Cortex variants. I have the snapdragon 810 developer tablet which has very nice specs and performance but it uses the standard 64 bit arm variant of the ARM A9 and A15s instead of krait. If the snapdragons in routers did retain the GPU they could also use the GPU as an offload processor letting the CPU handle other things. The issue with MU-MIMO is when you have multiple clients and one of them isnt MU-MIMO which is why i chose the AC3200 instead as it would be more practical. When it comes to optimisation qualcomm does a good job at it as evident as their development of krait.

Routers dont need 64 bit CPUs do be much faster. Some vendors have hardware encryption acceleration instead which would mean faster VPN and such. Forwarding of traffic doesnt require the CPU to have more bits but rather be faster at passing data/addresses. If a CPU has to handle data calculations bigger than its bit than it will use a cycle or 2 more than handle it so clock speeds are also important. The better the logic IPC, bigger cache and clock speed of the CPU the faster it is at networking. Even with 64 bit arm it is not necessarily faster than a 32 bit x86 CPU at the same clock speed when it comes to VPN and firewall because many intel and AMD CPUs have built in encryption and SSE so they can handle bigger data types using less clock cycles since VPN uses 128-256bits or higher crypto.

Any wifi AP with more than AC1900 divided into multiple streams or channels can handle multiple clients much better. In my opinion MU-MIMO can work better with multiple clients automatically while using multiple channels are a better option if you place more control over the network division.

I know of a user that just returned a brand new AC3200 router. He didn't like the fact that any one device couldn't receive the total bandwidth but that the total overall throughput was increased because he had so few devices. Not a happy camper but I explained that this technology might not be right for him because he is mostly concerned about the throughput for a single device.

 

[System Error Message]

i dont think MU-MIMO of AC2600 is going to give more throughput for a single device either. Both technologies only work when you have multiple devices. There are however other advantages such as better router cpu and ram, and more features. For example the recent feature of trendmicro router anti virus is similar to what enterprises used in the past in their networks, a networked anti virus and is a good feature for the non technical as it also helps when you have guests.

i know for me a wifi router with a fast CPU is a moot point since i already have a 36 core router but they run linux so i also use them for other things that would be too slow on MIPS CPUs.

 

[sfx2000]

when it comes to qualcomm snapdragon performance the krait cores only beat the cortext ARM a9 (which broadcom uses) but not A15 however some snapdragons include the stock ARM Cortex variants. I have the snapdragon 810 developer tablet which has very nice specs and performance but it uses the standard 64 bit arm variant of the ARM A9 and A15s instead of krait.

Krait crushes Cortex-A9 in terms of memory bandwidth performance - that's A9's weak spot, and why Qualcomm, Apple, and Marvell spun their own cores based on the ARMv7 ISA...

In my experience, there is no 64-bit variant of Cortex-A9

 

[sfx2000]

ARMv7, aka Cortex-A5 thru A9, A12, A15, and A17 - all 32 bit

ARM Licensed IP blocks - just about everyone except below:

These companies have Architecture Licenses for the ARMv7 ISA...

Qualcomm - Krait, Scorpion
Apple - A6/A6x - aka Swift
Marvell - Sheeva

hth...

sfx

 

[sfx2000]

I know of a user that just returned a brand new AC3200 router. He didn't like the fact that any one device couldn't receive the total bandwidth but that the total overall throughput was increased because he had so few devices.

Hence the problem right now with the whole AC**** rating system... it's gotten completely out of hand - saw a mention earlier about an AC5300 class router being demo'ed...

It's setting bad expectations for end-users, and allows marketeers to inflate things to ever increasing levels..

Heck, most folks now days, you buy a new low to midrange laptop off the shelf, and the included 802.11n wifi chip is a single stream, single band 2.4GHz, which may or may not even do wide channels...

 

[System Error Message]

i think they also need to advertise low to midrange routers too that are better than the default routers given by ISPs.

snapdragon 810 CPU :Quad-core ARM® Cortex™ A57 and quad-core A53 with 64-bit support

I have the hardware. The A53 is the improved version of the A9 while A57 is the improved version of the A15 not to mention they both are 64 bit CPUs. If you do a bit of research you'll find them to be the successors of the A9 and A15. There are some qualcomm CPUs that use the stock ARM design and various CPUs from them too. Some of the CPUs come with either kraits or stock ARM.

 

[sfx2000]

Routers dont need 64 bit CPUs do be much faster

Yes they do... in many ways, even in the SOHO space - 64bit cleans up a lot of cruft...

Some vendors have hardware encryption acceleration instead which would mean faster VPN and such.

Most do... whether the compiler and the SW devs avail themselves of it, that's another story - OpenVPN specifically doesn't, but it's on their to-do list...

Forwarding of traffic doesnt require the CPU to have more bits but rather be faster at passing data/addresses.

And that's why MIPS is still being used - it's remarkably efficient at these kind of tasks... not much different than the old SGI days where OpenGL and 3D rendering was mostly displaylists and bitslicing...

If a CPU has to handle data calculations bigger than its bit than it will use a cycle or 2 more than handle it so clock speeds are also important. The better the logic IPC, bigger cache and clock speed of the CPU the faster it is at networking. Even with 64 bit arm it is not necessarily faster than a 32 bit x86 CPU at the same clock speed when it comes to VPN and firewall because many intel and AMD CPUs have built in encryption and SSE so they can handle bigger data types using less clock cycles since VPN uses 128-256bits or higher crypto.

If one has to crack a 64-bit dataset into two pieces... well, that kills throughput...

Similar to AMD's Bobcat - it could do SSE, but it took twice the time as the SSE stack was only 64 bits wide...

sfx

 

[System Error Message]

im quite aware of the advantages and details of different hardware architectures. Some CPUs automatically use hardware acceleration such as hardware encryption depending on the instruction while some require the compiler to make use of it.

The CPU bit can be a confusing thing because it can mean that each instruction is x bits long or each part of the instruction is x bits long since each instruction has a few parts to it, it is not an actual indication of performance. The CPU can be 32 bits for the instructions but have 64 or 128 bit units to deal with data which allows it to perform 64 bit or 128 bit math in 1 cycle. More bits just means that it has more instruction types and larger addresses but it is not an indicator of having more performance. More instruction types means less instructions are required to represent a piece of higher level code so it also boils down to the compilers too. For example when games started utilising 64bits there werent much performance improvements, it didnt double 3D calculations, scripts and AI performance because there wasnt much change in representing them in machine language. Infact newer games are starting to demand much more of the CPU.

So whether a CPU is 32 bits or 64 bits it is a reference to the instruction set/instruction size. bigger CPUs handle data seperately from instructions (in parallel) but smaller CPUs have data as part of the instruction or a much smaller pipeline that can leave gaps in codes that arent compiled in a certain way.

 

[sfx2000]

im quite aware of the advantages and details of different hardware architectures. Some CPUs automatically use hardware acceleration such as hardware encryption depending on the instruction while some require the compiler to make use of it.

I think we're on the same page - in a general arch perspective, having logic units that can accelerate AES or h264 is pretty cool rather that work them software on general registers...

And generally, yes, it requires some compiler smarts to make it work.

64-bit is the future...

sfx

 

[System Error Message]

64bit instruction sets are bigger so it is very likely that we'll still see 32 bit, 16 bit and 8 bit CPUs around. although you dont hear of it but there are many tiny 16 and 8 bit chips/CPUs coupled together on boards that perform certain tasks. The advantage of less bits is a smaller CPU making it cheaper to fab and design while also using much less power.

64bit ARM CPUs use more power which is why my tablet has double the battery watts than normal so unless the home environment demands it i dont think consumer networking hardware will use 64bit CPUs. By using small CPUs like MIPS and ARM with some hardware acceleration the power usage gets reduced to a few watts. Since it is expected to keep the hardware turned on all the time any increase in power use means more cost of the item.

Ive also seen some devices that while they utilise low power technology have a bad design on power that ends up using many times more watts than it should.

 

[sfx2000]

The A53 is the improved version of the A9 while A57 is the improved version of the A15 not to mention they both are 64 bit CPUs. If you do a bit of research you'll find them to be the successors of the A9 and A15. There are some qualcomm CPUs that use the stock ARM design and various CPUs from them too. Some of the CPUs come with either kraits or stock ARM.

Trust me - I'm ex-Qualcomm, and a Qualcomm customer...

A9 is what it is... A15 is an improvement, but Krait/Scorpion is better... but A15 still 32-bit

A53/A57 is where they (ARM) are going after not having design wins otherwise...

 

[System Error Message]

Krait/scorpion isnt faster than A15 but does have better power usage. They still havent introduced 64 bit kraits yet.

 

[sfx2000]

64bit instruction sets are bigger so it is very likely that we'll still see 32 bit, 16 bit and 8 bit CPUs around. although you dont hear of it but there are many tiny 16 and 8 bit chips/CPUs coupled together on boards that perform certain tasks. The advantage of less bits is a smaller CPU making it cheaper to fab and design while also using much less power.

64bit ARM CPUs use more power which is why my tablet has double the battery watts than normal.

Nope...

aarm64 takes a bit more space, but like all ARM's, incredibly dense (good thing), but executes faster... so the power argument is moot, and most newer ARM's are on smaller geometries... so less power perhaps..

Your 64 bit tablet is perhaps a bad implementation, but that's outside of the ARM 64 bit discussion...