Upcoming ASUS ROG Gaming Router: GT-AC5300

[clvk07]

Basically it's a merge between AC88U and AC5300

So does it mean there is no successor to the AC88U? ie. AC89U?

 

[dlandiss]

This should probably go in the greybeard thread...

I spent many late nights as a young teenager sorting 6502 assembler* - came in handy later on in life...

* some wise guy will likely ask... so yes, Apple ][

Kids these days! Fortran loaded from punched cards to an IBM 650, running 12AU7 vacuum tubes and a rotating drum memory.

 

[sentinelvdx]

And not sure about this model....

Sent from my SM-G935FD using TapaTalk

 

[IAAI]

And not sure about this model.... View attachment 8200

Sent from my SM-G935FD using TapaTalk

Looks like

https://www.asus.com/Commercial-Networking/BRT-AC828/

 

[sfx2000]

So does it mean there is no successor to the AC88U? ie. AC89U?

Who knows what tomorrow brings, but my guess is that it won't be a BHR...

Things are changing...

 

[RMerlin]

So does it mean there is no successor to the AC88U? ie. AC89U?

No, it simply means that they don't have anything to announce yet.

Traditionally, Asus tends to make most of their announcement at Computex rather than CES. You'll note that the only press event they had at CES was to announce two phones.

 

[Gitsum]

That new web interface they have with the GT-AC5300 looks pretty damn cool. I like all the graphs, charts and numbers.

 

[el pescador]

Yeah the interface does looks nice.....could do with buying out netduma though...if they are serious about being pro gamer.

last attempt was nothing more than a gimmick

 

[maylyn]

I'm surprised Maylyn hasn't flooded us with photos yet LOL

You paging for me?

Sorry .... have been out of action for a few months due to work commitments overseas.

I will be back with more insider news and more teasers in time to come

 

[Shonk]

My god whats the AC88U 40nm? 55nm?
my 2014 Exynos Note4 (N910C) is 20nm geez broadcom like the old nodes
they should be launching at 14nm in 2017 not 28nm

And nice to see the die-shrink to 28nm, so it should run a bit cooler...

 

[kvic]

My god whats the AC88U 40nm? 55nm?
my 2014 Exynos Note4 (N910C) is 20nm geez broadcom like the old nodes
they should be launching at 14nm in 2017 not 28nm

Routers don't have to shrink as hard as phones. They're actually growing bigger every year but so do phones..

 

[Shonk]

why not exactly?
it has nothing to do with the size but the performance / power use
we pay more for some routers than most phones
they run 24/7 a 14nm cpu is going to use under half that of a 28nm cpu power wise and perform better

Routers don't have to shrink as hard as phones. They're actually growing bigger every year but so do phones..

 

[sfx2000]

My god whats the AC88U 40nm? 55nm?
my 2014 Exynos Note4 (N910C) is 20nm geez broadcom like the old nodes
they should be launching at 14nm in 2017 not 28nm

IIRC, current is on the 40nm node... 28nm is good.

14/16nm nodes are still pretty expensive - which is why only the high end silicon (CPU/GPU) are using it..

FWIW, the long pull on chips like this is not the CPU cores, as these are commodity ARM functional blocks, it's the switch element, which is broadcom's custom IP block.

(RPi3's BCM2710 is still on 40nm, again, long pull there is the videocore GPU, not the arm cores)

 

[kvic]

IIRC, current is on the 40nm node... 28nm is good.

14/16nm nodes are still pretty expensive - which is why only the high end silicon (CPU/GPU) are using it..

CPU/GPU and SoC are totally on different paths of process upgrade. SoC always lag behind a few generations. But as reference point, Huawei's Kirin 950 (2015) was already manufactured using TSMC's 16nm FinFET process.

 

[sfx2000]

CPU/GPU and SoC are totally on different paths of process upgrade. SoC always lag behind a few generations. But as reference point, Huawei's Kirin 950 (2015) was already manufactured using TSMC's 16nm FinFET process.

Apple's A9 is on both 14nm and 16nm (taped out twice for two fabs, TSMC and Samsung) - imagine the cost there (and twice the validation work as well)

 

[sfx2000]

Tossing some hard numbers for Cortex-A53 vs. Cortex-A7 - nice to see as both are on the same geometry node - 40nm, and exactly the same rest of board config...

Why this is relevant?

Cortex-A53 @ 1.2GHz is pretty close to Cortex-A9 at 1.0GHz

Broadcom's new chip is on Cortex-A53, QC-Atheros chip is Cortex-A7, so this is a valid test for the CPU cores if not the ethernet functional block - but there, QC-Atheros and Broadcom are roughly similar (within std deviation - both are very good)

TL;DR - Cortex A53 wins here...

Quick Study - ARMv7 across Cortex-A53 vs Cortex-A7
--------------------------------------------------
CPU Speed - 900 MHz
Core Clock - 250 Mhz
--------------------------------------------------
DUT's - Raspberry Pi2, Raspberry Pi3

Pi3 has been configured to same core/gpu/cpu/ram clocks as Pi2

/boot/config.txt - this item is added for the Pi3

# edit to underclock pi3 to pi2 clocks
core_freq=250
arm_freq=900
arm_freq_min=600
# end edit

Both DUT's are headless to remove the GPU from any test artifacts
--------------------------------------------------
SW Environment

OS - Raspbian 8 (Jessie)
SW Rev => 4.4.42-v7+ #946 SMP Sat Jan 14 10:14:34 GMT 2017
FW Rev => Jan 14 2017 11:57:33 
Copyright (c) 2012 Broadcom
version 5215826715b1baa53cd3a59458926588aefbe0e7 (clean) (release
--------------------------------------------------
Tests - exploring OpenSSL, gnutls-cli, and Sysbench across different
core architectures. This test is probably the best indicator of
improvements across ARM's A7/A53 as the rest of the HW is the same.
==============================

Benchmarks follow...
--------------------------------------------------
openssl speed aes-128-cbc aes-256-cbc bf-cbc

Note - this is a common bench on the SNB forums

Higher is better

A53

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     23961.03k    27038.46k    27929.86k    28178.09k    28254.21k
aes-128 cbc      30607.47k    34732.03k    36018.77k    36318.78k    36252.33k
aes-256 cbc      24148.63k    26515.22k    27253.93k    27383.13k    27440.47k


A7

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
blowfish cbc     17994.17k    20029.20k    20531.89k    20666.24k    20677.27k
aes-128 cbc      19525.01k    21070.93k    21693.70k    21925.24k    21921.13k
aes-256 cbc      15291.62k    16247.96k    16559.85k    16683.27k    16703.96k
---------------------------------------------------
gnutls-cli --benchmark-ciphers

Higher is better

A53

Checking cipher-MAC combinations, payload size: 16384
     SALSA20-256-SHA1 27.99 MB/sec
     AES-128-CBC-SHA1 13.91 MB/sec
     AES-128-CBC-SHA256 12.19 MB/sec
     AES-128-GCM 10.58 MB/sec

Checking MAC algorithms, payload size: 16384
            SHA1 61.87 MB/sec
          SHA256 37.77 MB/sec
          SHA512 9.06 MB/sec

Checking ciphers, payload size: 16384
        3DES-CBC 4.69 MB/sec
     AES-128-CBC 17.90 MB/sec
     ARCFOUR-128 44.69 MB/sec
     SALSA20-256 50.73 MB/sec

A7

Checking cipher-MAC combinations, payload size: 16384
     SALSA20-256-SHA1 19.15 MB/sec
     AES-128-CBC-SHA1 10.29 MB/sec
     AES-128-CBC-SHA256 8.82 MB/sec
     AES-128-GCM 8.19 MB/sec

Checking MAC algorithms, payload size: 16384
            SHA1 36.87 MB/sec
          SHA256 22.92 MB/sec
          SHA512 6.73 MB/sec

Checking ciphers, payload size: 16384
        3DES-CBC 3.24 MB/sec
     AES-128-CBC 14.13 MB/sec
     ARCFOUR-128 40.60 MB/sec
     SALSA20-256 38.80 MB/sec
------------------------------------------------
sysbench --num-threads=1 --test=cpu --cpu-max-prime=20000 --validate run

Lower is better

A53

Test execution summary:
    total time:                          637.2841s
    total number of events:              10000
    total time taken by event execution: 637.2725
    per-request statistics:
         min:                                 63.61ms
         avg:                                 63.73ms
         max:                                 95.73ms
         approx.  95 percentile:              63.96ms

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   637.2725/0.00


A7

Test execution summary:
    total time:                          767.7772s
    total number of events:              10000
    total time taken by event execution: 767.7574
    per-request statistics:
         min:                                 76.39ms
         avg:                                 76.78ms
         max:                                105.65ms
         approx.  95 percentile:              77.47ms

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   767.7574/0.00
-----------------------------------------------
sysbench --num-threads=4 --test=cpu --cpu-max-prime=20000 --validate run

Lower is better

A53

Test execution summary:
    total time:                          160.0435s
    total number of events:              10000
    total time taken by event execution: 640.0145
    per-request statistics:
         min:                                 63.60ms
         avg:                                 64.00ms
         max:                                121.08ms
         approx.  95 percentile:              64.08ms

Threads fairness:
    events (avg/stddev):           2500.0000/14.05
    execution time (avg/stddev):   160.0036/0.02

A7

Test execution summary:
    total time:                          192.6860s
    total number of events:              10000
    total time taken by event execution: 770.5622
    per-request statistics:
         min:                                 76.39ms
         avg:                                 77.06ms
         max:                                131.29ms
         approx.  95 percentile:              77.59ms

Threads fairness:
    events (avg/stddev):           2500.0000/17.25
    execution time (avg/stddev):   192.6405/0.03

 

[unsynaps]

All this talk the biggest thing I want out of my home router is more horsepower.

I find myself using VPN more and more being that I am now on FiOS (150/150) from Comcast (150/10... yes 10... TEN! With ZERO options to get anything faster <spits on Comcast>).

Would love to know if there will be Merlin support for the BRT-AC828. Mostly because out of the box it is dead to me without OpenVPN server support.

 

[RMerlin]

Would love to know if there will be Merlin support for the BRT-AC828.

No, for multiple reasons, but the main one being the BRT-AC828 is not based on Broadcom.

 

[RTN]

Based on the description, this looks to be a BCM4908, with BCM4366E wifi. 1 GB RAM, 256 MB flash.

ASUS GT-AC5300: https://wikidevi.com/wiki/ASUS_GT-AC5300
• CPU: Broadcom BCM49408 @1.8GHz Quad Core
• RAM: 1 GB (DDR3), Flash: 256 MB (NAND)
• WLAN: 3x Broadcom BCM4366E (4x4 802.11abgn/ac)
• WAN: 1x GbE, LAN: 8x GbE, USB: 2x USB 3.0

 

[RMerlin]

ASUS GT-AC5300: https://wikidevi.com/wiki/ASUS_GT-AC5300
• CPU: Broadcom BCM49408 @1.8GHz Quad Core
• RAM: 1 GB (DDR3), Flash: 256 MB (NAND)
• WLAN: 3x Broadcom BCM4366E (4x4 802.11abgn/ac)
• WAN: 1x GbE, LAN: 8x GbE, USB: 2x USB 3.0

Wikidevi often bases their entries on pure speculations rather than hard fact. When they list the specs of an unreleased device, don't blindly trust what they publish. I remember one case where the pre-release specs they posted was different from the pre-release sample I actually had in hands.