Overclock and Merlin 380.58 / AC3200 can not make it work

[FTC]

Hello,

I am attempting to overclock a bit my router, since having good ventilation and temps, in order to gain some speed in USB disk sharing. I am using the shared disk to do image backups and every MB/sec gained ends up shaving minutes from the whole process..

The problem is that I am not able to make it work. I added the nvram set clkfreq=1200,800 / nvram commit sentences in the jffs scripts init-start and services-stop, but after reboot (tried twice) the overclock has not been executed (bogomips still around 2000), even if the clkfreq is 1200,800.

Has this been seen before ? any idea on what to test ? (and yes, the scripts have execute permission and are coded with nano, i.e linux no LF style).. or anyone with a 3200 and merlin 380.58 can confirm overclock works ?
Regards

 

[hggomes]

That was working on 380.57 builds (via JFFS script), probably the new FW code changed it like it happens now at every new FW release lately, new lockups, now you can't even overclock your router? Why not? Could it be also a FCC demand? "NO OVERCLOCK" available to the users , it's not ASUS!!!!! The funny part here is that new models CPUs are the same ones overclocked, and a serial cable proofs it, the CPU goes from 1GHZ to 1.4GHZ. Also, why forcing users to use an older FW version so they can overclock their routers? Why not also on latest versions? Is it because new models came out? Sure it is.

Marketing is just great isn't it? Locking up old models so new ones can sell better? Does this sounds right? IMO it doesn't.

Tip: Set any old router model (RT-AC56U, RT-AC68U/P, RT-AC87U, RT-AC3200) at 1.4GHZ and run on SSH/Telnet terminal:

# openssl speed aes-256-cbc

Now do the same on new released models (RT-AC88U, RT-AC3100 or RT-AC5300), after the result compare you will have your answer. Same score / performance.

I cannot tell, i would need to test the latest RMerlin / Stock FW to know, for reference:

BogoMIPS : 1595.80 (800MHZ)
BogoMIPS : 1992.29 (1GHZ)
BogoMIPS : 2398.61 (1.2GHZ)
BogoMIPS : 2798.38 (1.4GHZ)
BogoMIPS : 3182.65 (1.6GHZ)

Remember every single router has different OC capabilities, it's a lottery depending on the CPU VID, the lower the better.

 

[FTC]

Hi, thanks for your message.. anything to try to overcome this new limitation ?

 

[hggomes]

I will give it a try in the next hour, i will let you know.

 

[hggomes]

No more overclocking, is seems the code really changed on 380.58 and now the clkfreq is not saved, you can always downgrade or use another FW.

 

[snakebite3]

No more overclocking, is seems the code really changed on 380.58 and now the clkfreq is not saved, you can always downgrade or use another FW.

What if you do it manually? I can still OC on my AC68u with 380.58.
Commit the setting, then reboot using the web browser not from whatever client you used.

 

[hggomes]

Yes, i was testing it once again and it's a model specific thing, RT-AC3200 can't overclock for some reason on any FW version.

Heres the proof:

admin@RT-AC3200-FC10:/tmp/home/root# nvram get clkfreq
99999,800
admin@RT-AC3200-FC10:/tmp/home/root# cat /proc/cpuinfo
Processor : ARMv7 Processor rev 0 (v7l)
processor : 0
BogoMIPS : 1998.84

processor : 1
BogoMIPS : 1998.84

Features : swp half thumb fastmult edsp
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc09
CPU revision : 0

Hardware : Northstar Prototype
Revision : 0000
Serial : 0000000000000000

It will accept any higher clock frequency it doesnt do anything with it, but it can downclock:

admin@RT-AC3200-FC10:/tmp/home/root# nvram get clkfreq
800,800
admin@RT-AC3200-FC10:/tmp/home/root# cat /proc/cpuinfo
Processor : ARMv7 Processor rev 0 (v7l)
processor : 0
BogoMIPS : 1599.07

processor : 1
BogoMIPS : 1599.07

Features : swp half thumb fastmult edsp
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc09
CPU revision : 0

Hardware : Northstar Prototype
Revision : 0000
Serial : 0000000000000000

Theorically the only way to overclock is set the frequency on the bootloader but that would be an insane step, risking on a permanent router brick if it doesn't accept for some reason the new frequency.

It seems to me a protection not allowing to setup only a higher clock frequency, downclock is allowed.

Thanks for reporting this, i found why it's happening. Overclock is back, i will test it and let you know.

 

[sfx2000]

BogoMIPS : 1595.80 (800MHZ)
BogoMIPS : 1992.29 (1GHZ)
BogoMIPS : 2398.61 (1.2GHZ)
BogoMIPS : 2798.38 (1.4GHZ)
BogoMIPS : 3182.65 (1.6GHZ)

These are just timing loops, so be careful comparing things across different chipsets/architectures - MIPS/ARM/ARC/x86...

Even within the same arch - ARMv7, different chips on Bogomips give different numbers...

J1900 running on a KVM instance on debian... and we all know that a Intel Silvermont @ 2GHz can well outperform an ARMv7 Cortex-A9 dual-core@1.6Ghz...

$ lscpu
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                2
On-line CPU(s) list:   0,1
Thread(s) per core:    1
Core(s) per socket:    2
Socket(s):             1
NUMA node(s):          1
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 55
Stepping:              8
CPU MHz:               1999.999
BogoMIPS:              3999.99
Hypervisor vendor:     KVM
Virtualization type:   full
L1d cache:             24K
L1i cache:             32K
L2 cache:              1024K
NUMA node0 CPU(s):     0,1

 

[kvic]

For some losing ability to overclock, the cause might be this:

http://www.snbforums.com/threads/te...ds-are-now-available.30648/page-4#post-241555.

which in turn might be caused by this:

http://www.snbforums.com/threads/sbin-preinit-memory-leak-in-380-series-fw.31622/#post-252144

 

[FTC]

Thanks to all... so you are saying that doing the reboot from the GUI instead of with the reboot command may work ?
I'll test in a minute...

 

[snakebite3]

Thanks to all... so you are saying that doing the reboot from the GUI instead of with the reboot command may work ?
I'll test in a minute...

How did it go?

 

[FTC]

How did it go?

Well, it didn't work. Overclock still does not stick.

 

[FTC]

Thanks for reporting this, i found why it's happening. Overclock is back, i will test it and let you know.

hggomes, it seems you are our only hope now.. any news/tests to do that you can share ? Thanks in any case for your support...

 

[snakebite3]

Well, it didn't work. Overclock still does not stick.

It seems like only the 3200 can't OC.

 

[hggomes]

Yes, new models overclock fine, i will need to investigate it further.

DDR Clock: 800 MHz
Info: DDR frequency set from clkfreq=1600,*800*
et2: Broadcom BCM47XX 10/100/1000 Mbps Ethernet Controller 7.14.131.1608 (r589038)
CPU type 0x0: 1600MHz
Tot mem: 1048576 KBytes

 

[kvic]

Thanks to all... so you are saying that doing the reboot from the GUI instead of with the reboot command may work ?
I'll test in a minute...

/jffs/scripts/services-stop is not called from GUI reboot either.

The only time I saw services properly stopped is when auto reboot after firmware upgrade.

Regarding this issue on services-stop, we need @RMerlin 's help.



Overclocking itself.. No problem for RT-AC56U on 380.58

 

[hggomes]

It seems like only the 3200 can't OC.

Indeed, nice finding @FTC.

Bootloader overwrites the clkfreq (but not on nvram) after setting it higher than 1000, that seems to be the problem.

 

[FTC]

/jffs/scripts/services-stop is not called from GUI reboot either.

The only time I saw services properly stopped is when auto reboot after firmware upgrade.

That must not be the only case... I placed yesterday a 'touch' sentence to a file in a flash disk and effectively up on reboot services-stop was NOT called, but immediately I changed the touch destination file to /jffs/ dir and the following reboot did call services-stop -although I am not sure that time if I rebooted through command line or GUI.

When I saw your message with a thread link where it was stated that a GUI reboot may be the one that worked correctly I though that this would have been the cause of my inconsistent tests.

So I think we are dealing with two problems here :
1. Services_stop is not being called some times (we will have to dig on to why.. possibly with Merlin ?)
2. In the RT-AC3200 at least and as hggomes stated the clkfreq may even if set be overwritten by bootloader... What seems very strange is that you can overclock to a *smaller* value and then bootloader does not overwrite the value. Seems like done on purpose, but then .. why only in one model of router ?

So still waiting for some ideas here.. in the line of (just some thoughts) :

- Finding another 'place' during boot or shudown were we could attempt the overclock
- Trying to overclock by directly writing to the PLL ?? is this even possible ??
- Finding which module in the kernel does the clock setting and calling to it somehow with the new values set.. or forcing a restart without passing through bootloader.. is it even possible ?
- Finding why bootloader overwrites the clock frequencies and try to avoid it ? (maybe this is caused under some conditions that can be set somehow)
....

 

[hggomes]

Serial cable is required to debug this.

 

[kvic]

@FTC

Only certain versions of CFE allow O/C.

Check out the big thread about CFE under ASUS Wireless.

So it's actually your first attempt (?)...not that O/C stops in 380.58. Didn't know where I got the wrong presumption.