• After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.

Root FINALLY A WORKING JELLY BEAN ROM!! Public release! Update from ICS!

Do you want more dev for this phone or is it time to let it go?

  • Total voters
    53
Buzbee said:
It probably displayed the original recovery. I have never seen any bootloader for Force. The bootloader you described sound like an HTC ...
Click to expand...
That's what I thought I remembered, but I may be wrong. I did not back up recovery, but it seemed that its behavior changed. Thanks for pointing me to /etc. I see a description of (apparently successful) overclocking at http://androidforums.com/threads/kernel-kernel-1-8ghz-wip-module-method-also.742386/ but no flashable kernel or module -- I don't understand the code in the first post, which refers to moving a copy of thermald.conf into /data/local, but the only thermald.conf I see on the Force is already in /etc.
 
Ponow said:
I'm happily using sawbones999's JB on ZTE Force from Freedompop. Thanks!

I'm most concerned about battery life. It seems that, except for wifi, most of the battery is not used by the screen brightness nor even supposedly resource hungry apps, but rather the sensors in "Android System".

Using the Google Play store app "Battery Monitor Widget", on the statistics tab it lists two sensors as using the overwhelming majority of battery (since previous charge):
30.67% out of 55.61% for Freescale MMA845X Accelerometer
18.39% out of 55.61% for Light Sensor TSL2771

At one point during my experiments I was able to get rid of the accelerometer power usage (I think I by disabling the auto screen rotate). I have tried to disable the automatic brightness detection, but that didn't seem to affect the light sensor battery usage. So at this point I'm most concerned about the light sensor.

I also tried using the "Hardware Disabler" app (I can't find it on the Google Play store now, & I can't recall where I found it), but none of the devices listed were the light sensor (I determined this by doing a web search for each of the listed device model numbers).

Running "lsmod" on the "Terminal Emulator" app only lists wlan.

I did a web search for "linux tsl2771" and found http://lwn.net/Articles/439746/, which shows that this chip also provides the proximity sensor. While I'd like to just disable this chip (maybe there's a kernel boot parameter that can be used?), I'd also still like to use this device as a phone, which means that we need the proximity sensor, so disabling just the light sensor functionality may be what's required to reduce the battery drain. How can this be achieved?
Click to expand...

Robert Chase said:
Thanks for your labors. The ZTE bootloader is a blank screen, but before I flashed it, my Freedompop Force displayed a menu like that on my Freedompop Evo Design 4G offering HBOOT, fastboot, recovery, and system as options -- how should this component be described and do you know how to restore it? How can the boot-up sound be disabled? It didn't play on ICS and I don't see a setting under "sound". Any progress on overclocking?
Click to expand...

Would both of you mind trying out my latest kernel? It should save on battery by default, but will also let you clock up to 2 GHz. I get some throttling but just added voltage controls, and if I undervolt I can get a stable OC around 1.7GHz. Just flash on top of any of the 4.1 ROMs.

http://www.mediafire.com/download/t28oa3qvapxqkgr/force_volt.zip

If you want to go back, this undoes my changes

http://www.mediafire.com/download/il15ymm4bcr7or6/force_sawbones_kernel.zip
 
Ponow said:
I'm happily using sawbones999's JB on ZTE Force from Freedompop. Thanks!

I'm most concerned about battery life. It seems that, except for wifi, most of the battery is not used by the screen brightness nor even supposedly resource hungry apps, but rather the sensors in "Android System".

Using the Google Play store app "Battery Monitor Widget", on the statistics tab it lists two sensors as using the overwhelming majority of battery (since previous charge):
30.67% out of 55.61% for Freescale MMA845X Accelerometer
18.39% out of 55.61% for Light Sensor TSL2771

At one point during my experiments I was able to get rid of the accelerometer power usage (I think I by disabling the auto screen rotate). I have tried to disable the automatic brightness detection, but that didn't seem to affect the light sensor battery usage. So at this point I'm most concerned about the light sensor.

I also tried using the "Hardware Disabler" app (I can't find it on the Google Play store now, & I can't recall where I found it), but none of the devices listed were the light sensor (I determined this by doing a web search for each of the listed device model numbers).

Running "lsmod" on the "Terminal Emulator" app only lists wlan.

I did a web search for "linux tsl2771" and found http://lwn.net/Articles/439746/, which shows that this chip also provides the proximity sensor. While I'd like to just disable this chip (maybe there's a kernel boot parameter that can be used?), I'd also still like to use this device as a phone, which means that we need the proximity sensor, so disabling just the light sensor functionality may be what's required to reduce the battery drain. How can this be achieved?
Click to expand...

I'm wondering if there's some problem with the reporting from the battery monitor app.
I was seeing large % battery usage from the items you mentioned as well. The results were from at least 15 charge cycles.
I decided to clear history and start over to get clean analysis.
The current run I'm no longer seeing the large % drain anymore.
I can see that cell standby is the largest drain on battery which completely makes sense.

I would suggest clearing history at the beginning of every cycle to get a clean look.

It might really just be a reporting error from the app as it accumulate data from cycle to cycle.
 
spock1104 said:
Would both of you mind trying out my latest kernel? It should save on battery by default, but will also let you clock up to 2 GHz. I get some throttling but just added voltage controls, and if I undervolt I can get a stable OC around 1.7GHz. Just flash on top of any of the 4.1 ROMs.

http://www.mediafire.com/download/t28oa3qvapxqkgr/force_volt.zip

If you want to go back, this undoes my changes

http://www.mediafire.com/download/il15ymm4bcr7or6/force_sawbones_kernel.zip
Click to expand...

I tried it, but Antutu scores went from almost 21,000 to below 12,000; I gather that the CPU is being throttled -- by thermald.conf?
 
No, thermald is not used by my kernel. I stuck to using thermal monitoring integrated in the kernel so that there's less system resources being used to monitor the CPU. The controls are accessible through 3C Tuner, but the throttling I've been seeing isn't in line with the settings and so I'm still working out the full functionality. I was able to undervolt the higher frequencies and avoid the throttle.

I use Cool Tool to monitor my CPU as part of the kernel testing. It'll show your frequency and can also show core loads and temperatures, so you can tell if you're getting throttled.

I've also found that Antutu is doing some odd things with CPU usage, only using one core even during the multithreaded tests. I recommend also comparing Quadrant scores, it's been much less finicky for me. To help me narrow things down though, which individual test(s) performed lower?

Thanks for trying it out!
 
In Antutu:
(Multitask, Runtime, CPU Integer, CPU floating point, Single-thread integer, S.T. floating point, RAM Operation, Ram Speed, 2D Graphics, 3D graphics, Storage I/O, Database I/O
Sawbones JB: 3604, 1324, 1101, 1034, 1638, 1104, 590, 1171, 911, 7181, 522, 640
overclocked (1.73GHz): 1099, 423, 317, 302, 969, 639, 172, 750, 914, 6435, 428, 440

The only measure not significantly decreased is 2D graphics.
 
That's a huge help, thanks! I'm going to put in showp1984's re-implementation of thermal monitoring, will report back when complete.
 
spock1104 said:
I've replaced the thermal monitoring and tweaked the voltages. Good luck.

http://www.mediafire.com/download/ab82sibpq555op9/force_tweak.zip

If you want to test the raw performance I recommend shutting off mpdecision and switching to the performance governor
Click to expand...
OK; I'll try it. I hope your tweaking still leaves safety margins for the chip. As for using other governors than the default (ondemand), whenever I do that, the OS resets it. I'll try changing governors if benchmarks reflect increased clock frequencies.
 
I lowered the voltages, so if anything there's some more margin now. Which app are you using to set the governor?
 
1728MHz seems to be my upper limit.

qKKDKyg.png
eSgiTrj.png
 
That's an impressive benchmark, but my Force with the tweak installed would not run under most combinations of governor and frequency; the only combination which could complete a benchmark was 1728MHz under SmartassV2, but it again got an Antutu score below 12,000. I've been using Antutu CPU Master Free to set these.
 
Spock must have got a mutant CPU in his Force. With the kernel tweaks installed I have never been able to complete an AnTuTu benchmark at any frequency. Remove the tweaks and I get 21,700 or so every time.
 
Hodr said:
Spock must have got a mutant CPU in his Force. With the kernel tweaks installed I have never been able to complete an AnTuTu benchmark at any frequency. Remove the tweaks and I get 21,700 or so every time.
Click to expand...
How do you get ~1,000 more than me? I realize that the scores vary from run to run, but I've never broken 21,000. Have you applied adjustments to improve performance?
 
Robert Chase said:
That's an impressive benchmark, but my Force with the tweak installed would not run under most combinations of governor and frequency; the only combination which could complete a benchmark was 1728MHz under SmartassV2, but it again got an Antutu score below 12,000. I've been using Antutu CPU Master Free to set these.
Click to expand...
Did antutu crash or just quit? If it crashed, try the default clocks. If it's still crashing, the /proc/last_kmsg file should tell me why if you can grab it. Hodr and I both saw a crash at the 3d test that I haven't been able to repeat.

For changing the kernel settings I recommend 3c toolbox or kernel tuner 2014.
 
I re-installed force-tweak (after reverting) and then tried to use Kernel Tuner 2014 to set the clock to 1728MHz and the governor to Interactive -- now I can't even start Antutu without the Force rebooting. It stayed up long enough to reset the clock to 1512MHz in Kernel Tuner and then seemed to be running normally, but after turning on the phone this morning it rebooted sometime around ~45% through an Antutu run; I uploaded last_kmsg prior to the reboot and last_kmsg2 after. I am reverting to my last known good copy.
 

Attachments

P.S. With force-tweak installed, the clock set at 1512 MHz, and the CPU governor set to smartassV2, ondemand, and interactive in Kernel Tuner, running Linpack resulted in 345Mflops, 825Mflops, and a reboot, respectively. After reverting I removed Antutu CPU Master Free (which had set my CPU governor to performance), installed Kernel Tuner, and got a Linpack of 824. I changed the governor to ondemand and interactive and got Linpacks of 821 and 829, respectively.
 
Thanks for having another try and grabbing logs. Unfortunately the first one just shows the bootloader, the phone must have wiped the log on reboot. The second one though shows an error in your L1 cache, which would certainly explain the instability. I'll have a chance to look in more detail later tonight. Hopefully there's a kernel patch for this, or maybe it's the toolchain.
 
an L1 issue is usually hardware but can be induced by doing something stupid in the kernel, either a voltage being off or a clock being too high. BUT, since this was at the stock speed and voltage, I have to look for something more extreme. I've been compiling with linaro 4.9.3 2015.03 but I'm getting a bad vibe between this and the memory error I saw in the video driver. Switching the toolchain is easy, so here's one done with 4.7.4 2014.06. I know that one to be reliable. If we still see issues I have two other options: some more tuning of clocks and voltages (also done by qualcomm and in the franco kernel) or removing the L1 error from being detected (also done in the franco kernel). I figure try the least drastic and go from there.

http://www.mediafire.com/download/3ruzf9y9adf1oqu/force_tweak47.zip

MD5 hash is B0DF7A659BA5898F9F75D61E775D5FED so we can rule out a bad download

I think I actually like it better like this. Not bragging, trying to show that I'm testing...

ZyauWyi.png
ib5z7gz.png
aWHWvXb.png
R1nOKYy.png
H1MaEel.png
 
Still no go. I attach two more last_kmsg files after crashes of Antutu; one of these was at 1.68GHz (I believe). I verified the checksum of force_tweak47. My phone rebooted once during Antutu with Sawbones' 4.1 installed -- perhaps I have a marginal (or damaged) CPU.
 

Attachments

Same thing both times, L1/TLB error. I'll remove the L1 kernel panic tonight for kicks, we'll see if it's somehow stable.
 
I just wanna say thanks for making this! my freedompop phone works great with this rom. I just have a question for anyone who might now, maybe sawbones. Does anyone know which file is loaded when the charger is plugged in when the phone is off? I want to make it automatically turn on the phone when plugged in if it is off instead of going to the battery charging screen.
Thanks!!
 
I withdraw my support for Kernel Tuner 2014, it's only applying settings to cpu0 correctly. Stick with 3c System Tuner, 3c Toolbox (my preferred), or Trickster Mod. I produced an L1 issue on my phone after using KT2014...
 
Last edited:
spock1104 said:
I withdraw my support for Kernel Tuner 2014, it's only applying settings to cpu0 correctly. Stick with 3c System Tuner, 3c Toolbox (my preferred), or Trickster Mod. I produced an L1 issue on my phone after using KT2014...
Click to expand...
Interesting -- it goes against the grain to have to use a large app with ads to set such parameters as CPU maximum frequency, governor, and read-ahead cache size, but if that's the available tool that works, I suppose I'll have to use it. Do you know anything about the injector script (http://forum.xda-developers.com/showthread.php?t=2256647)? Can it be used to add init.d support to this ROM? I re-installed the ROM, looked at all the recommended apps, and settled on System Tuner -- nothing I could do, including replacing Busybox with Busybox on Rails, would placate Trickster Mod, which apparently displays the Busybox version as a cryptic error message in lieu of the current scheduler, CPU governor, and other data. Having finally verified another good copy and backed it up, I re-installed force_tweak47, but as I raised the clock frequency with the slider in System Tuner, the Force crashed. I again could not clock it any higher than stock without crashing during benchmarks -- my Antutu score dropped from ~21,000 without the tweak installed to ~17,000 with it installed and the CPU at stock frequency. I'll induce another crash and attach a last_kmsg tomorrow.
 
You must log in or register to reply here.
Back
Top Bottom