Overclock AC68U

[Guibs]

that site is presented to you in french because it knows your locale

I just assumed that hggomes might speak french since he seem close to Merlin and his Beta build (he gave me access to BETA 2 even before Merlin officially released it today) and Since Merlin is french and live about 30 minutes from where I am, I possibly assumed hggomes might be from where I live as well.

 

[hggomes]

Ahaaha yeah, Sinshiva is right

You will loose warranty if you opened it yes, the 4 screws are inside on the heatsink.

 

[Guibs]

Ahaaha yeah, Sinshiva is right

You will loose warranty if you opened it yes, the 4 screws are inside on the heatsink.

Yeah I googled pictures of the board and notice the 4 screws on the heatsink. I think I'll just go and replace it today with a new one and hope that new one has the thermal pad in there. Since its the same store and probably the same manufacturing lot, it might be missing as well but I have a 30 day period to return it so I'll switch it tonight and hope to get lower temperature. If my router is downcloacking because of temperature, I'm not going to keep it.

 

[RMerlin]

I just assumed that hggomes might speak french since he seem close to Merlin and his Beta build (he gave me access to BETA 2 even before Merlin officially released it today) and Since Merlin is french and live about 30 minutes from where I am, I possibly assumed hggomes might be from where I live as well.

I uploaded the beta builds to a shared directory on Mediafire and shared the link on IRC with the folks there before publishing them. That's where he got it (along with a dozen other folks who hang out there).

 

[KKLD]

Yeah I googled pictures of the board and notice the 4 screws on the heatsink. I think I'll just go and replace it today with a new one and hope that new one has the thermal pad in there. Since its the same store and probably the same manufacturing lot, it might be missing as well but I have a 30 day period to return it so I'll switch it tonight and hope to get lower temperature. If my router is downcloacking because of temperature, I'm not going to keep it.

Are you sure it's downclocking?

Please do the these commands via telnet/SSH and post the results here along with the current temp of your AC68U:

nvram get clkfreq
openssl speed aes-128-cbc

 

[Guibs]

Are you sure it's downclocking?

Please do the these commands via telnet/SSH and post the results here along with the current temp of your AC68U:

nvram get clkfreq
openssl speed aes-128-cbc

Will do tonight when I get home. I went ahead and replaced the router during lunch time today. With the other crashes / lookup I had earlier this week, it just didn'T feel like I had a stable unit hardware wise. I will setup it up tonight, update the firmware and give you the information on it. Hopefully the temperature will be lower on it but we'll see.

 

[Guibs]

Just want to report that I've exchanged my router and my new 68U does run a lot cooler. Current temperature seems to hold between 78C-82C now even when overclocked at 1000,666.

Will report back on stability after a few days.

 

[philmiami]

IT's FASTER to use 1100,667<<<<

I originally set my clkfreq to 1100,800
here are TWO different setting
one is at 1100,800 and one is at 1100,667<----- which is FASTER and gives more done

ASUSWRT-Merlin RT-AC68U_3.0.0.4 Sat Jun 7 16:32:17 UTC 2014
admin@RT-AC68U:/tmp/home/root# cd //
admin@RT-AC68U:/# nvram get clkfreq
1100,800
admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 7121022 aes-128 cbc's in 2.98s
Doing aes-128 cbc for 3s on 64 size blocks: 1955805 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 256 size blocks: 506737 aes-128 cbc's in 2.99s
Doing aes-128 cbc for 3s on 1024 size blocks: 124562 aes-128 cbc's in 2.91s
Doing aes-128 cbc for 3s on 8192 size blocks: 15601 aes-128 cbc's in 2.90s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 38233.68k 41723.84k 43386.18k 43832.13k 44070.14k
admin@RT-AC68U:/# nvram set clkfreq=1100,667
admin@RT-AC68U:/# nvram commit
admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 6928999 aes-128 cbc's in 2.83s
Doing aes-128 cbc for 3s on 64 size blocks: 1900702 aes-128 cbc's in 2.95s
Doing aes-128 cbc for 3s on 256 size blocks: 491572 aes-128 cbc's in 2.86s
Doing aes-128 cbc for 3s on 1024 size blocks: 124061 aes-128 cbc's in 2.88s
Doing aes-128 cbc for 3s on 8192 size blocks: 15718 aes-128 cbc's in 2.95s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 39174.55k 41235.57k 44000.85k 44110.58k 43648.09k
admin@RT-AC68U:/#

So who came up with 1100,800
Mine is faster at 1100,667
just a little less done on 8192 bytes

 

[Guibs]

I have no idea. If I clock my ram at 800 my ac68u does not even boot enymore and require a full reset to work. Even when I was overclocking to 1000,666 at first I kept trying 1000,667 but it would not retain the value after a reboot. It only stick when I went for 1000,666.

My router default value was 800,533

 

[hggomes]

You need to reboot the router after cpufreq change, otherway will not make efect.

Note: Openssl results can change a little bit from test to test even at the same frequency.

I originally set my clkfreq to 1100,800
here are TWO different setting
one is at 1100,800 and one is at 1100,667<----- which is FASTER and gives more done

ASUSWRT-Merlin RT-AC68U_3.0.0.4 Sat Jun 7 16:32:17 UTC 2014
admin@RT-AC68U:/tmp/home/root# cd //
admin@RT-AC68U:/# nvram get clkfreq
1100,800
admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 7121022 aes-128 cbc's in 2.98s
Doing aes-128 cbc for 3s on 64 size blocks: 1955805 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 256 size blocks: 506737 aes-128 cbc's in 2.99s
Doing aes-128 cbc for 3s on 1024 size blocks: 124562 aes-128 cbc's in 2.91s
Doing aes-128 cbc for 3s on 8192 size blocks: 15601 aes-128 cbc's in 2.90s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 38233.68k 41723.84k 43386.18k 43832.13k 44070.14k
admin@RT-AC68U:/# nvram set clkfreq=1100,667
admin@RT-AC68U:/# nvram commit
admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 6928999 aes-128 cbc's in 2.83s
Doing aes-128 cbc for 3s on 64 size blocks: 1900702 aes-128 cbc's in 2.95s
Doing aes-128 cbc for 3s on 256 size blocks: 491572 aes-128 cbc's in 2.86s
Doing aes-128 cbc for 3s on 1024 size blocks: 124061 aes-128 cbc's in 2.88s
Doing aes-128 cbc for 3s on 8192 size blocks: 15718 aes-128 cbc's in 2.95s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 39174.55k 41235.57k 44000.85k 44110.58k 43648.09k
admin@RT-AC68U:/#

So who came up with 1100,800
Mine is faster at 1100,667
just a little less done on 8192 bytes

 

[Guibs]

I don't know how you guys can even clock your RAM to 800, mine refuse to boot back if I do. Is there something I am missing?

 

[hggomes]

That's a CFE limitation, you can upgrade and take it off but thats something that has some risks, if something goes wrong you will need a replacement.

http://forums.smallnetbuilder.com/showthread.php?t=17793

I don't know how you guys can even clock your RAM to 800, mine refuse to boot back if I do. Is there something I am missing?

 

[Guibs]

Ah thanks, that explains it. I don't really see a need right now to do so. Just having 1000,666 might be enough for me. Where would I notice an improvement if I was to upgrade further? Faster WIFI speed with concurrent clients/streaming?

 

[hggomes]

Use 1200,666 its a safe OC and you will have alot better performance, not in wireless but in USB speed, OpenVPN, FTP, Samba, etc.

 

[philmiami]

You need to reboot the router after cpufreq change, otherway will not make efect.

Note: Openssl results can change a little bit from test to test even at the same frequency.

ok. this is after a reboot, running at 1100,666

Connection to 192.168.0.1 closed.
phil-lan-imac:~ philmiami$ ssh admin@192.168.0.1
admin@192.168.0.1's password:


ASUSWRT-Merlin RT-AC68U_3.0.0.4 Sat Jun 7 16:32:17 UTC 2014
admin@RT-AC68U:/tmp/home/root# cd //
admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 6930386 aes-128 cbc's in 2.88s
Doing aes-128 cbc for 3s on 64 size blocks: 1969670 aes-128 cbc's in 2.99s
Doing aes-128 cbc for 3s on 256 size blocks: 510377 aes-128 cbc's in 3.01s
Doing aes-128 cbc for 3s on 1024 size blocks: 128699 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 8192 size blocks: 15645 aes-128 cbc's in 2.89s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 38502.14k 42160.16k 43407.48k 43929.26k 44347.35k
admin@RT-AC68U:/#

 

[Guibs]

Use 1200,666 its a safe OC and you will have alot better performance, not in wireless but in USB speed, OpenVPN, FTP, Samba, etc.

Will probably try this after one more day at 1000,666 to make sure its stable at current overclock.

Do we know if we are void our warranty with Asus if we overclock our router?

 

[SLK350]

Overclocking my AC68U, currently @ 1200/666. Any idea to try push for 1300 or 1400? Will I gain anything?

 

[hggomes]

Offcourse you will, but check if its stable at those speeds.

 

[philmiami]

Offcourse you will, but check if its stable at those speeds.

I made a post in the regular Asus wireless
mine went FUBAR @ 1600,666

1200,775 is working right now
I read the posts and the outside website postings
1400 not stable
1100 yes
1200 yes
1400 no
1600 NO!


admin@RT-AC68U:/# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 7128555 aes-128 cbc's in 2.97s
Doing aes-128 cbc for 3s on 64 size blocks: 1963745 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 256 size blocks: 509833 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 1024 size blocks: 128676 aes-128 cbc's in 2.99s
Doing aes-128 cbc for 3s on 8192 size blocks: 16139 aes-128 cbc's in 3.01s
OpenSSL 1.0.0m 5 Jun 2014
built on: Sat Jun 7 12:38:55 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-128 cbc 38402.99k 41893.23k 43505.75k 44068.30k 43923.82k
admin@RT-AC68U:/#

no difference between 1200,775 and 1000,800 as far as I can see
I don't really care
I think 1000,800 or maybe 1100,666 is fine

 

[System Error Message]

I managed to get mine at 1400mhz and having a usb fan blowing air out. Max temperature is 66 degrees and ever since i used the usb fan i have not had any wireless drops. After more than a day of stress testing the CPU it is totally stable.

Perhaps you should try checking performance using openssl speed aes-128-cbc -multi 2.

openssl speed aes-128-cbc -multi 2
Forked child 0
Forked child 1
+DT:aes-128 cbc:3:16
+DT:aes-128 cbc:3:16
+R:8348338:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:64
+R:8375709:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:64
+R:2353621:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:256
+R:2355297:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:256
+R:613130:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:1024
+R:612976:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:1024
+R:154569:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:8192
+R:154664:aes-128 cbc:3.000000
+DT:aes-128 cbc:3:8192
+R:19311:aes-128 cbc:3.000000
+R:19352:aes-128 cbc:3.000000
Got: +H:16:64:256:1024:8192 from 0
Got: +F:16:aes-128 cbc:44670448.00:50246336.00:52307285.33:52791978.67:52843861.33 from 0
Got: +H:16:64:256:1024:8192 from 1
Got: +F:16:aes-128 cbc:44524469.33:50210581.33:52320426.67:52759552.00:52731904.00 from 1
OpenSSL 1.0.0n 6 Aug 2014
built on: Sun Aug 17 16:15:01 EDT 2014
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: arm-brcm-linux-uclibcgnueabi-gcc -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -ffunction-sections -fdata-sections -DTERMIO -O3 -Wall -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
aes-128 cbc      89194.92k   100456.92k   104627.71k   105551.53k   105575.77k

amazing speeds when using 2 threads. If you want to stress test the CPU aside from iperf is by looping the openssl benchmark.

simply connect via SSH and run

while :
do
openssl speed aes-256-cbc -multi 2
done

I tried overclocking further but it wouldnt boot. Has anyone gotten past 1.4Ghz yet? I havent tested samba yet but i hope this is enough to get 150MB/s from it. I could overclock further if it was possible to overvolt. I literally taped a black dual usb fan to the router which made it felt bulky but doesnt drop wireless like it used to.

I would really like to ask merlin to make the features in router mode available in other modes. I leave the handling of internet, QoS, VPN, firewall and much more to a much much faster router but router mode has many features i would like to use but router mode is not possible for my setup. I noticed that there is a lack of VLAN and port bonding capability in LAN (especially in AP mode) which i need in order to reduce the mess of cabling and make it possible to connect to 2 networks simultaneously via VLAN tagging. My ISP doesnt allow the use of NAT so i cannot use router mode since there is no bridging mode for WAN. Although this can be overcome with some deep packet processing but this is not possible on consumer routers even with 3rd party firmware. My ISP actively blocks computers behind a NAT.