ASUS ROG Rapture GT-AC5300 Wireless-AC5300 Tri-Band Gaming Router Reviewed

[sfx2000]

BCM4366E is a nice step up - there's more to it that just a simple model number change - they beefed up the microcontroller and SRAM significantly on the "E" model compared to the 4366 - mostly because of the additional processing requirements for the MU group handling...

Even without MU - one should see some improvement - and since Beamforming is part of Wave1 - mid-range/long-range performance should be a bit better in SU mode compared to 4366/4360 - dependent of course on the quality of the front end obviously...

 

[avtella]

C0 stepping of which chip specifically ? The CPU stepping (i.e. BCM4709C0) doesn't have any impact on MU-MIMO. AFAIK, a BCM4366E is required (as opposed to BCM4366).

Sorry, I should have clarified, I meant 4366C0 not the CPU.

Looks like Linksys updated from the 4366B1 on the EA9500 (https://fccid.io/document.php?id=2975586) and also the Netgear R8500 silently to this, I assume the AC88U also has a newer revision with it.

 

[somms]

Sorry if I am confused but does this mean this router wouldn't come close to hitting my max 1Gbps up/down throughput? Is CTF not enabled by default?

 

[Makaveli]

Sorry if I am confused but does this mean this router wouldn't come close to hitting my max 1Gbps up/down throughput? Is CTF not enabled by default?

None of these routers can hit 1Gbps speed unless CTF is enabled.

 

[somms]

None of these routers can hit 1Gbps speed unless CTF is enabled.

"Best" open source firmware for Netgear X10

Individual at the above linked thread claims the X10 could attain his bandwidth throughput cap of 1Gbps under DD-WRT which does not include the CTF module AFAIK...

http://www.dd-wrt.com/phpBB2/viewtopic.php?p=1049650

CTF is just a hack that bypasses parts of the firewall it is closed source and requires, that an old kernel is used it also takes away the ability to fully customize the kernel. Thus CTF is not an option, because:

-you have to use old unsupported kernels, no security fixes from upstream
-is insecure by design as it only works if it pybasses certain frewall functionality
-causes instability if you recompile the kernel with changed network functionality

If you need a router that can handle 1Gbps, then you should just choose the right unit. Netgear R7800 has a different cpu architecture and allows gigabit speeds under dd-wrt without any hacks.

I'm sure broadcom will soon have a unit that can also do it without CTF.

We tested a CTF port some time ago and removed it again as it broke some features.

FWIW: DD-WRT firmware developer Kong also claims the R7800 can hit 1Gbps w/o CTF loaded...

 

[RMerlin]

Sorry, I should have clarified, I meant 4366C0 not the CPU.

Looks like Linksys updated from the 4366B1 on the EA9500 (https://fccid.io/document.php?id=2975586) and also the Netgear R8500 silently to this, I assume the AC88U also has a newer revision with it.

No idea then, sorry. I never heard about the B or C stepping before. E was the first time I heard they had a new revision, and at the time it was mentioned that it was intended to address the MU-MIMO issues, among other things.

 

[RMerlin]

FWIW: DD-WRT firmware developer Kong also claims the R7800 can hit 1Gbps w/o CTF loaded...

Other (i.e. not BCM) platform might have their own technology that can improve throughput but without it being specifically CTF. That technology might possibly be less constraining than CTF in terms of feature compatibility.

 

[RMerlin]

mostly because of the additional processing requirements for the MU group handling...

Interesting. That could indicate that BCM's MU-MIMO issues with the BCM4366 weren't due to a design flaw, but simply lack of processing power?

 

[sfx2000]

Interesting. That could indicate that BCM's MU-MIMO issues with the BCM4366 weren't due to a design flaw, but simply lack of processing power?

I can't imagine it being an overall design flaw in the 4366 - the chip did handle Wave 2 MU-MIMO, just perhaps not as well as other chips did.

MU is incredibly complex from a systems perspective for scheduling the groups, doing the MU group sounding to ensure that the group is still valid, and looking at the inbound/upstream data flows (QoS marking for example for WMM) - it really is tough to model even, much less implement. QC-Atheros spent years working on their implementation as did Quantenna - Marvell was also behind the curve a bit as well.

The enhancements in the "E" chip were shared informally by a close friend and former colleague - the micro-controller was bumped up to a cortex-a9, where previously, Broadcom was using the Cortex-M cores for the control logic inside the chip.

 

[sfx2000]

I should add - even though the 4366 was a bit challenged for MU-MIMO, there was much benefit to Wave 1 usage as the 4366 is a 4*4:4 chipset, which had benefits over the already excellent 4360 for SU-MIMO... the 4366E will build upon that, so even if MU isn't needed, it's still a pretty good chip.

 

[Skyler LaVey]

For those asking about vpn performance...it’s pretty terrible and speeds are unreliable. The Netgear X10 maxed my connection at 120Mbps. DD-WRT was running on my X10. I do however get 20ms lower ping time on the GT AC5300 compared to the X10 and AC3100. My vpn speeds vary from nothing, not connecting even though it says it is, to about 40Mbps on the GT AC5300.


Sent from my iPhone using Tapatalk

 

[Makaveli]

For those asking about vpn performance...it’s pretty terrible and speeds are unreliable. The Netgear X10 maxed my connection at 120Mbps. DD-WRT was running on my X10. I do however get 20ms lower ping time on the GT AC5300 compared to the X10 and AC3100. My vpn speeds vary from nothing, not connecting even though it says it is, to about 40Mbps on the GT AC5300.


Sent from my iPhone using Tapatalk

If its 40Mbps that isn't great lower prices models with lesser cpu's can already hit that.

 

[RMerlin]

For those asking about vpn performance...it’s pretty terrible and speeds are unreliable. The Netgear X10 maxed my connection at 120Mbps. DD-WRT was running on my X10. I do however get 20ms lower ping time on the GT AC5300 compared to the X10 and AC3100. My vpn speeds vary from nothing, not connecting even though it says it is, to about 40Mbps on the GT AC5300.


Sent from my iPhone using Tapatalk

I could get 50 Mbps out of a 800 MHz RT-AC68U, so there's no reason for it to be lower than that on the GT-AC5300. What were your testing parameters?

That CPU should be able to reach close to 100 Mbps with OpenVPN and an AES-128-CBC cipher.

 

[thiggins]

Sorry if I am confused but does this mean this router wouldn't come close to hitting my max 1Gbps up/down throughput? Is CTF not enabled by default?

Correct. CTF is enabled. I tested this a few times.

 

[sfx2000]

I could get 50 Mbps out of a 800 MHz RT-AC68U, so there's no reason for it to be lower than that on the GT-AC5300. What were your testing parameters?

That CPU should be able to reach close to 100 Mbps with OpenVPN and an AES-128-CBC cipher.

It's a new chip (broadcom's B53 Cortex-A53 variant), and in my experience with A7/A53, compiler options and version can make a big difference - I think it's fair to say that there's some headroom to be found with the B53 in the near future.

Recall the migration from MIPS to ARM Cortex-A9 a few years back... this is similar... The Cortex-A9's improved over time.

 

[RMerlin]

It's a new chip (broadcom's B53 Cortex-A53 variant), and in my experience with A7/A53, compiler options and version can make a big difference - I think it's fair to say that there's some headroom to be found with the B53 in the near future.

I saw some (internal) benchmark numbers. It's definitely WAY faster than the RT-AC68U for OpenVPN.

 

[Skyler LaVey]

This is applying the vpn to the router itself not my computer. I used openvpn within the router software. The vpn feature doesn’t even work right now anyway. It’s complete trash. That’s my only complaint with the router at this time. My settings never stick.


Sent from my iPhone using Tapatalk

 

[sfx2000]

I saw some (internal) benchmark numbers. It's definitely WAY faster than the RT-AC68U for OpenVPN.

It should be... not just for clock speed, it's got more L2 cache per core, a much better memory interface, and more bandwidth across the switch inside the SoC.

BCM4708/4709

• 2 Cortex-A9@800-14000MHz (with no VFP/Neon)
• 256KB L2 Cache
• 16-bit DRAM interface
• 5 1GBe ports

to BCM4908

• 4 Cortex-A53@1800MHz (with VPF3, VFP4, Neon, and ARMv8-A features)*
• 1MB L2 Cache
• 32-bit DRAM interface
• 7 switched ethernet ports, including 2.5GBe

*Broadcom in the vendor lit calls the core B53, but generally it's considered an A53 variant by many - but to GCC, the model number is unique.

The 4908 is a pretty decent chip just looking at the cores, but when compares the entire SoC, it's a big jump from the 4708/4709 - and moving to a smaller node (40nm vs. 28nm), should run a bit cooler and use less power.

Folks should be patient - it's a new toolchain and baseline SDK...

Moving from ARMv7-a code over to 64Bit ARMv8-a is non-trivial - in another thread, the current firmware spelunking shows there is a mix of 32 and 64 bit code, along with arch splits - as more of the SDK moves over to ARMv8 64-bit native, performance and stability will improve.

 

[thiggins]

Folks should be patient...

Why should people be patient when spending almost $400 for a router?

 

[sfx2000]

Why should people be patient when spending almost $400 for a router?

It's a work in progress... as such, it does pretty good.

I see potential - and as many know first hand, I'm not a Asus Fanboi... but I see the upsides with the platform... It's not perfect, but running, and it will run better in the future I think.

Many on this thread focus on the negatives - it's a new SDK/Platform/Toolchain...

FWIW - I spent $350 for a dedicated pfSense box without wireless - and that's money well spent, considering needs at the time...