Overclocking RT-AC56U

[RMerlin]

The command is nvram get clkfreq

That shows the nvram value, but that won't necessarily show what has been applied to the hardware.

I don't know any way to read the value that's actually programmed in the memory controler.

 

[RMerlin]

Technically, 1000,800 should not be called overclocking as it is within the specs of our hardware.

Broadcom has three different SKUs for the BC470x line. I don't think they all clock to 1 GHz - in fact I believe that's one of the differences between the BCM4708 and BCM4709.

 

[RMerlin]

This is why personally I don't offer any kind of help or support for overclocking - you're entirely on your own there. There's a chance that the CFE could be programming the CPU clock BEFORE it checks for the reset/WPS buttons, in which case it means you can be permanently bricked. Chances are Sky got lucky by trying multiple times because one time the CPU worked long enough to get through the reset code. But that won't necessarily always be the case.

Unless someone can study the CFE code to confirm that the reset gets handled before the CPU clock gets set, assume the worse.

 

[Sky1111]

Broadcom has three different SKUs for the BC470x line. I don't think they all clock to 1 GHz - in fact I believe that's one of the differences between the BCM4708 and BCM4709.

Good point. Published Broadcom specs are not very clear IMO...

So R7K owners are safe at 1GHz?

Either way, what happened to Primitivo's router is a concern. Corrupted NVRAM?

So the router was running for a while at 1000,800 and one day it just died? Or it died right after applying the clocks?

 

[RMerlin]

Good point. Published Broadcom specs are not very clear IMO...

So R7K owners are safe at 1GHz?

Probably. I think Tim did mention in his article that it was one of the differences with the BCM4709.

Either way, what happened to Primitivo's router is a concern. Corrupted NVRAM?

So the router was running for a while at 1000,800 and one day it just died? Or it died right after applying the clocks?

Read what I wrote about CFE and reset handling. This is what I suspect happened to him.

Personally, I've ran into cases with the RT-AC56U where the WPS and reset buttons were no longer handled at all by the CFE. The only way I was able to recover was by using the initial console (over serial) started by the firmware - the CFE prompt itself didn't work at all either, as it wouldn't acknowledge the Ctrl-C trying to get in. And it was also failing to initialize the Ethernet interface.

The CFE on the RT-AC56U is quite quirky. That makes the RT-AC56U a very unsafe platform for anything that involves heavy experimentation. A bad firmware flash can quite easily completely brick that router.

 

[speedingcheetah]

I set my AC56U back to 1000,667.....network died randomly....(missed the last 10 min of the new episode of Criminal Minds

Not sure if it was the OC or not that caused it....hopefully wont have any mor e issues.

Also, other firmware like DDWRT and Tomato can show ram frequency and such, so wouldn't be the same command or code?


EDIT: mmmm....wifi cut out again on my laptop about 15min after I changed the clock setting....weird....used it for hours yesterday with OC settings on and had no issues.

 

[RMerlin]

Also, other firmware like DDWRT and Tomato can show ram frequency and such, so wouldn't be the same command or code?

They most likely just show the content of the clkfreq nvram setting. That does not mean that's the value that's actually used by the HW.

 

[speedingcheetah]

Weird...thought it was the setting change to the router causing my wifi on my laptop to randomly drop out and not detect my network...but my S3 is still connected to the network and works fine....

 

[Sky1111]

playing with CPU/memory clocks should not have any effect on the wifi... could be weird coincidence or something is gone wrong.

I would try few things:
- clear the NVRAM
- re-flash the firmware - whatever version you prefer - even if it is the same
- clear the NVRAM again
- setup the router from scratch
(hint - others suggested that too already - write done all your settings, so you just copy paste PSK keys/static IP adresses/logins/etc.)
- set the last known working clocks
- check the temperatures

Now about the clocks: even though we can type whatever number, CPU/memory clocks can only change in increments and in certain ratio; so it if the start was 800/533, by the same ratio the next step would be 1000/667, and then 1200/800. Merlin - please correct me - is 1000/800 kosher (not saying it is not, just asking)?

speedingcheetah - if 1000/800 was stable for you and you did not have any issues, and temps are in check, I really do not see a reason why should downlock the memory. If I am not mistaken - you were fine with 1000/800, and all you issues started when you went 1000/667?

 

[RMerlin]

Merlin - please correct me - is 1000/800 kosher (not saying it is not, just asking)?

Nothing else but the stock clock is safe. For example, you don't know if Asus is undervolting the RAM since they run it at a lower clock.

 

[Sky1111]

Nothing else but the stock clock is safe. For example, you don't know if Asus is undervolting the RAM since they run it at a lower clock.

Very good point. Overclocking is not only science, but in many cases just luck.
Some chips run fine while undervolted, and some with crap out.

This whole discussion makes the decision between getting AC68U vs Netgear R7K (which seems by all means faster than ASUS) very painful...

 

[RMerlin]

Very good point. Overclocking is not only science, but in many cases just luck.
Some chips run fine while undervolted, and some with crap out.

This whole discussion makes the decision between getting AC68U vs Netgear R7K (which seems by all means faster than ASUS) very painful...

I doubt that DD-WRT's implementation of OpenVPN on a 1 GHz R7000 will give you any significant performance boost over my own implementation on a 800 MHz RT-AC56U. I wouldn't be surprised if it was actually slower, due to lack of CTF, not having the various optimizations I did to openssl and openvpn, etc...

 

[Sky1111]

I doubt that DD-WRT's implementation of OpenVPN on a 1 GHz R7000 will give you any significant performance boost over my own implementation on a 800 MHz RT-AC56U. I wouldn't be surprised if it was actually slower, due to lack of CTF, not having the various optimizations I did to openssl and openvpn, etc...

...you have a point there; but we already know that BCM 4708 at stock engine/memory clocks has a huge performance drop even on 25Mb internet - so only if AC68U is overclockable it can fight with R7K... otherwise it will be just as slow (as far as VPN Client is concerned), but with much better wireless.

 

[RMerlin]

I was able to push 60 Mbits through an OpenVPN tunnel on an RT-AC56U running at stock clock.

 

[speedingcheetah]

playing with CPU/memory clocks should not have any effect on the wifi... could be weird coincidence or something is gone wrong.

speedingcheetah - if 1000/800 was stable for you and you did not have any issues, and temps are in check, I really do not see a reason why should downlock the memory. If I am not mistaken - you were fine with 1000/800, and all you issues started when you went 1000/667?

No...I was running at 1000/800......it was only my laptop that 2x lost the wifi signal...while my S3 still saw the network...and wired network still was fine. I set it back to 1000/667 and about 15min later, laptop went out again. I shutdown and rebooted laptop, and it hasn't had any issues as of yet...

Currently, Temp of 5ghz says 44C.....never seen it go above 48....cpu temp form CLI says 52C.

 

[speedingcheetah]

Very good point. Overclocking is not only science, but in many cases just luck.
Some chips run fine while undervolted, and some with crap out.

This whole discussion makes the decision between getting AC68U vs Netgear R7K (which seems by all means faster than ASUS) very painful...

Mini rant: I will never use another Netgear router again. I have nothing but problems and failures...not to mention numerous bricks with the Netgear routers that I have flashed custom firmware to.
Also, Dlink....their customer service is a joke...and they refuse to accept any RMA I have submitted...claiming they haven't received any of the needed information, despite my several email replies to their messages. DLink also claim they will charge you if you don't send back all the org accessorizes(pwr cord, ethernet cord, QIG and cd)...ridiculous.

 

[silvercloud]

anyone using iperf for stress test?

Sent from my SHV-E160L using Tapatalk now Free

 

[PrivateJoker]

Mini rant: I will never use another Netgear router again. I have nothing but problems and failures...not to mention numerous bricks with the Netgear routers that I have flashed custom firmware to.
Also, Dlink....their customer service is a joke...and they refuse to accept any RMA I have submitted...claiming they haven't received any of the needed information, despite my several email replies to their messages. DLink also claim they will charge you if you don't send back all the org accessorizes(pwr cord, ethernet cord, QIG and cd)...ridiculous.

Asus tells you (at least in my case) specifically not to send in cords and stuff, and I got a refurb (cosmetically new) unit back sealed in anti-static bag with new power adapter & cords and all that.

 

[primitivo]

After my unsuccessful experience I can only say to those who still consider Overclocking to OC CPU first instead of doing both at the same. At least you will be able to report here which step exactly bricked your router.

 

[speedingcheetah]

anyone using iperf for stress test?

Sent from my SHV-E160L using Tapatalk now Free

I have used iperf....but i noticed that the numbers get higher as you run it multiple times....not sure if it is due to caching or something...not sure if there is a way to set it to run continually.