Root Overclocking Ally

[getitnowmarketing]

There is a stable 1GHz Pre kernel.

Please read 2 posts above yours. WE DO NOT HAVE THE SAME PROCESSOR AS THE PRE. The pixi has the same processor as ours which the palm guys have yet to overclock.

 

[Mitchell4500]

Please read 2 posts above yours. WE DO NOT HAVE THE SAME PROCESSOR AS THE PRE. The pixi has the same processor as ours which the palm guys have yet to overclock.

Basically it means drellisdee pwned and was able to do it before those guys and considering the Ally has been out less time (i think...) is pretty impressive.

 

[flick159]

So is there anyway we could try it out?

Would love to give this one a shot also.

 

[getitnowmarketing]

Would love to give this one a shot also.

If all goes well for testing I'll release in a few days. Follow here if interested in testing GetItNowMarketing (getitnowmarket) on Twitter

 

[lolbug]

Our processor & Palm Pixi MSM7627

Our device spec sheet VS740

Palm Pre has OMAP 3430

Here is the freq table I'm using currently.
The 6th column is the AHB bus which has a max khz of 200000 on msm7x27 models. The 7th column is the ahb divider which is (x + 1) which the highest value allowed is (3 + 1) =4 so its Clock speed 806400/4 = 201600. So the highest clock speed without pushing the ahb over its max is 787200.

Also a good read is Overclocking 101 Read the part about the wafers on the chips construction. That is why the max speed your processor can handle is different amongst the same models as its due to construction quality.

So 787 is the maximum value we can use, huh? Oh well. Still better than stock. You oc'd with a 2.2-compatible kernel, I'm guessing. Otherwise you might have to do this all over again.

 

[badandy2021]

There is actually a *stable* 1GHz Pre kernel.



Ah, I see. Interesting...

Where is it? I'd love to be able to validate this claim

 

[getitnowmarketing]

So 787 is the maximum value we can use, huh? Oh well. Still better than stock. You oc'd with a 2.2-compatible kernel, I'm guessing. Otherwise you might have to do this all over again.

Source to kernel 2.6.32 for froyo is not out. This is on kernel 2.6.29 and will have to be reworked for 2.6.32.

 

[badandy2021]

regardless props to drellisdee. although the extent of my overclocking has reached into gaming consoles and a few other arbitrary devices, android is something totally new to me and i can definitely appreciate the hard work that goes into working on a project like this.

 

[gregskeels]

If all goes well for testing I'll release in a few days. Follow here if interested in testing GetItNowMarketing (getitnowmarket) on Twitter

That's a better ballpark date then we got for the 2.2 update lol but I b so happy for the oc

 

[darkxsun]

Please read 2 posts above yours. WE DO NOT HAVE THE SAME PROCESSOR AS THE PRE. The pixi has the same processor as ours which the palm guys have yet to overclock.

I *did* read it... I didn't say it was the same processor, just that there was a 1GHz stable kernel, meaning 800MHz wasn't the stable ceiling as the comment I responded to claimed...

Where is it? I'd love to be able to validate this claim

Easy one
Application:UberKernel - WebOS Internals

 

[badandy2021]

its too bad that this kernel is only for the palm pre and pixi
it barely works on the pixi (ie no overclocking) and the Ally and Pixi share the same
CPU or similar. I'm not saying it won't ever be OC'ed over 800mhz, I'm
saying that this processor will NOT be stable at significant speeds over that. it just won't. I can apply voltages and kernel multipliers to my old P4,
that still doesn't mean it won't experience some sort of china syndrome at 4 ghz.

 

[SuperMerg]

Flashed OC kernel. OCing with OCWidget (only has benchmark for 768, I requested with devs to included 787).

Runs well for about 15 mins, then kernel panic. Pretty consistent. Droid EmuLite runs better now though!

I <3 drellisdee!

 

[darkxsun]

Flashed OC kernel. OCing with OCWidget (only has benchmark for 768, I requested with devs to included 787).

Runs well for about 15 mins, then kernel panic. Pretty consistent. Droid EmuLite runs better now though!

I <3 drellisdee!

Is this possibly caused by the 768/787 discrepancy between the kernel and the widget? Also, is the kernel set to scale and thus need this widget, or is it set to always run at only 787 automatically? Just some debugging questions.

Congrats to drellisdee for all that's been done so far -- these are big, challenging steps even if we're not at a releasable product yet.

 

[getitnowmarketing]

Ok for testing OC kernel you need to use Setcpu ONLY!!!! I have a link for a free download of it in the read me from the developer of it so you won't have to buy it just to test. Its so worth the $1.99 though.

Not all devices can handle all the same speeds. So don't ask why yours can't run at 787mhz when some one else's can. All the QC tests on these processors require is it to run at 600mhz and some can run higher due to better quality construction so each case is unique as to what you can run at.

On boot the kernel is set to performance for faster booting as per LG. The default min that I built into kernel is 245mhz. I set the max to 600mhz on boot to prevent it trying to boot at frequencies higher than your ally may be able to handle.

Once booted and after every reboot use setcpu to change it from performance to ondemand which must be used for my overclocking (if you leave it on performance your battery life will suffer alot and its no gain in performance as I think its worse at freq above 600mhz). Try frequencies for a period of time in order of 729, 748, 768, 787 to see what your device can handle.

Do not use set on boot stuff in setcpu or you could wind up with a non booting device and have to boot into recovery and wipe data. You have been warned!!!

Also can we keep this thread on topic as I don't really care about devices with other processors like palm pre and moto droid as they are irrelevant to our development here.

For feedback I'd like a post like:

Rom:
Purchase/Manufacture date:
Max stable speed for your device:
Max observed cpu temp:
Comments:

Download Beta#1

 

[badandy2021]

awesome, nice job : D

 

[badandy2021]

Velocity v0.1.1
July 2010
787 Mhz
96.8 F
Stress test at 787 Mhz for 340 sec no errors

 

[savoxis]

Velocity v0.2 Beta 8
April 2010
787 Mhz
78.8 F
Ran stress test for 280 seconds, no errors

 

[badandy2021]

so how is that 806 working out?

 

[dots]

Velocity v0.1.1
May 2010
729mhz
102.2 F
Stress test time was 178 seconds with no errors.

 

[getitnowmarketing]

Velocity v0.1.1
May 2010
729mhz
102.2 F
Stress test time was 178 seconds with no errors.

Looks like he has as lesser die quality. What were your results over 729mhz?

 

[lolbug]

Velocity 0.1.1
August 2010
787.2MHz
~31C after 163 seconds w/ no errors

Looks like I'm lucky.

 

[gregskeels]

If I go over 768mhz mine crashes and mine is a may baby

 

[dots]

Looks like he has as lesser die quality. What were your results over 729mhz?

I tried it at 787mhz and about 15 minutes later my phone restarted itself

 

[lolbug]

Velocity 0.1.1
August 2010
787.2MHz
~31C after 163 seconds w/ no errors

Looks like I'm lucky.

I take that back. Kernel panic after about 1hr. Next time it happens, how do I go about getting the logfile?

 

[badandy2021]

Been running at 787 all day. My arm6 is revision 5, don't know if all allys are same revision. Temp has dropped significantly as well, down to low 80s F. Set to oc at boot, I'm comfortable with the near 200 mhz performance increase thanks to drellisdee : D