@szomi - no issues with directly flashing Padavan from openWRT? I've read somewhere that the partitions are different so better to go via stock firmware
For me it has worked good without stock firmware. But you can put it back if you want.
You have to remove the R27 resistor, next to the SoC, under the heatsink. Just use the thinnest iron tip and a steady hand.
Here is what I've done (no thin tip and also being retarded with the iron).
EDIT: placebo. U-boot/bootloader mod also needed.
Thanks, the pictures are sufficient...
Probably i will try it on my second mir3g next time...
From where do you get this information?
Are you really sure that the clock-frequency property in the dts alter the real cpu clock and this is no illusion?
That's what I'm testing right now, I'm rolling a custom build with a bunch of benchies, gonna run them as soon as I get home. Stability wise it's just fine.
The fellas from 4pda figured that out, I'm just trying the same under openwrt.
EDIT: Compiling the image again, there was something fishy with that build. WAN would disconnect for no apparent reason.
1 Like
Please report back your results, i´m very interested in modding my mir3g too...
@szomi - I'm stuck at the same point as guilhem.
After creating the firmware with prometheus at the time of flashing (after backup which shows errors as it's not on stock firmware) - 4 firmware and again 4 in submenu for flush a firmware - I'm seeing the same screen and have no idea what to choose - kernel ?
I found somewhere else 2 options:
- revert to stock firmware first before installing padavan and then use prometheus again
- install Breed and flash it via Breed
No way that simply editing the DTS file overclocks it. That's just going to cause time drift.
The Breed bootloader should support overclocking this.
Why are you so sure?
Should? Breed doesn´t support oc on mt7621 currenty...
Because I've done enough DTS file manipulation to know. Changing the clock in the DTS file means you're lying about the speed it's actually running at.
I thought I read somewhere Breed could with the 3G. Maybe I was mistaken.
I´m also very skeptical that modifying the device tree property will really change the real clock.
On 4pad i´ve read that they modify the u-boot source to change the real clock.
I´ve read nothing about modifying the clock property in the device tree...
@gr1m Can you give me a link to the post on 4pda according to the dts modification?
I think you have to change the dts file if you overclock the mt7621 with the modified bootloader to get no time drifts because the clock could not detected in linux on mt7621...
I've double checked and lede > padavan is not working. So I've put back the stock and after the padavan. Now it is OK.
Yeah, I did that also. Now on padavan and no disk errors in dmesg and also wifi seems to be more stable but that's just one hour so I need to monitor it deeper.
Strange - under padavan I run into file limit issue on samba (~4GB size), even using samba 4.8.0-1 with SMB3 enabled. Via FTP it works fine, I can copy large files. Any idea ?
With new stable build 18.0.1 everything o use seems ok.
For me too, switched today from snapshot to 18.06.1 and everything is fine. Also 2.4g wifi until now...
Also @neheb.
It didn't time drift or really OC at all, besides displaying much higher bogomips. My results were a bit higher than using your builds @juppin, the ones you uploaded to your github account, but mostly due to compiling optimizations.
Turned out it was funky because of using anything other than -O2 when compiling with GCC8.1, it was constantly hanging and crashing, just like when OCing the hell of the WDR4320.
https://4pda.ru/forum/index.php?showtopic=837667&st=780#entry64676189 is the post with instructions in russian.
@neheb, BREED as of some months ago didn't support OCing the MIR3G, I'm updating it right now to see if anything has changed.
1 Like
Meh. Nothing new.
Maybe hex editing the eeprom?
I´ve read this post with google tanslate...
They are modifying the u-boot source of padvan to oc the soc.
For me this is to risky until i haven´t tested a second boot source from a spi flash or a ram booted u-boot...
They mention in this thread also a possibility t boot the bootloader from spi. Probably i will try this and then build my own oc bl.
The reason I say that is because on the GnuBee devices, the bootloader sets the CPU to run at 900MHz by default whereas the DTS file used to say 880MHz. This was causing time drift until I figured out the problem.
I can't understand that very well. Are they saying memory clocks are crippled on the 3G?