[KERNEL][Omegamoon][B926+][ICS]Overclock to max 1.3GHz (Updated 2012/08/24)

[Guest Cyda]

This kernel breaks camera for me in infusion. Where is genokolar kernel?

CWM flashable can be found here:

[Guest timfimjim]

camera works fine for me with infusion b3

[Guest Frankish]

Pretty sure Geno's kernel will be the same since Geno supplied the camera fix

<br /><br />Then why on a clean install does omega kernel break camera yet genos is working fine for me Edited September 11, 2012 by Frankish

[Guest Potatoes]

<br /><br />Then why on a clean install does omega kernel break camera yet genos is working fine for me

I don't know for sure, the camera on geno's worked for me as does omega's (at least since the 2012/08/19 update). Tbh I'm just speculating since I don't have access to geno's kernel source.

[Guest bladebuddy]

If you read back a good few posts you will see different cameras have been used on the same phone . ZTE also used different parts.

[Guest Potatoes]

Yeah I think that's probably it bladebuddy, I forgot about that. I did notice when I was compiling mine that the new B936 .config has more sensors enabled than omegas, omega's literally only has MT9P017 whereas the official rom has a whole bunch which I assume accounts for all the variations.

B936

CONFIG_HUAWEI_CAMERA=y

CONFIG_HUAWEI_CAMERA_SENSOR_S5K5CA=y

CONFIG_HUAWEI_CAMERA_SENSOR_MT9T113=y

# CONFIG_MT9T013 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9D113 is not set

CONFIG_HUAWEI_CAMERA_SENSOR_MT9V113=y

# CONFIG_MT9D112 is not set

# CONFIG_HUAWEI_SENSOR_HI701 is not set

# CONFIG_HUAWEI_SENSOR_OV7690 is not set

# CONFIG_HUAWEI_SENSOR_OV7736 is not set

# CONFIG_HUAWEI_SENSOR_HIMAX0356 is not set

# CONFIG_HUAWEI_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_SENSOR_MT9V114 is not set

CONFIG_HUAWEI_SENSOR_MT9P017=y

Omegamoon

CONFIG_HUAWEI_CAMERA=y

# CONFIG_HUAWEI_CAMERA_SENSOR_S5K5CA is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9T113 is not set

# CONFIG_MT9T013 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9V113 is not set

# CONFIG_MT9D112 is not set

# CONFIG_HUAWEI_SENSOR_HI701 is not set

# CONFIG_HUAWEI_SENSOR_OV7690 is not set

# CONFIG_HUAWEI_SENSOR_OV7736 is not set

# CONFIG_HUAWEI_SENSOR_HIMAX0356 is not set

# CONFIG_HUAWEI_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_SENSOR_MT9V114 is not set

CONFIG_HUAWEI_SENSOR_MT9P017=y

Edited September 11, 2012 by Potatoes

[Guest omegamoon]

is possible overclock the GPU? thanks

Read this thread over at XDA. Bottom line is that GPU overclocking is impossible but "global overclocking that affects 1GHz CPU clock will affect GPU too"

[Guest omegamoon]

Hi Omegamoon, just getting round to adding some zRAM to my g300 and got a question:

1. On your website you give instructions to add 100kb - isn't that 100mb?

2. I got stuck with the last line - turning the swap on. Screen shot attached.

3. For people who add zRAM but don't know if it's working, type 'busybox free' into terminal emulator and it'llshow you the status of the swap on the last line :)

You're right, the proposed size is too high. I'm looking into this "short write" problem. Don't know why it occurs, it sometimes does. After that I have to reboot to give it another try.

[Guest Dazzozo]

Yeah I think that's probably it bladebuddy, I forgot about that. I did notice when I was compiling mine that the new B936 .config has more sensors enabled than omegas, omega's literally only has MT9P017 whereas the official rom has a whole bunch which I assume accounts for all the variations.

B936

CONFIG_HUAWEI_CAMERA=y

CONFIG_HUAWEI_CAMERA_SENSOR_S5K5CA=y

CONFIG_HUAWEI_CAMERA_SENSOR_MT9T113=y

# CONFIG_MT9T013 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9D113 is not set

CONFIG_HUAWEI_CAMERA_SENSOR_MT9V113=y

# CONFIG_MT9D112 is not set

# CONFIG_HUAWEI_SENSOR_HI701 is not set

# CONFIG_HUAWEI_SENSOR_OV7690 is not set

# CONFIG_HUAWEI_SENSOR_OV7736 is not set

# CONFIG_HUAWEI_SENSOR_HIMAX0356 is not set

# CONFIG_HUAWEI_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_SENSOR_MT9V114 is not set

CONFIG_HUAWEI_SENSOR_MT9P017=y[/code]




Omegamoon

[CODE]CONFIG_HUAWEI_CAMERA=y

# CONFIG_HUAWEI_CAMERA_SENSOR_S5K5CA is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9T113 is not set

# CONFIG_MT9T013 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_CAMERA_SENSOR_MT9V113 is not set

# CONFIG_MT9D112 is not set

# CONFIG_HUAWEI_SENSOR_HI701 is not set

# CONFIG_HUAWEI_SENSOR_OV7690 is not set

# CONFIG_HUAWEI_SENSOR_OV7736 is not set

# CONFIG_HUAWEI_SENSOR_HIMAX0356 is not set

# CONFIG_HUAWEI_SENSOR_MT9D113 is not set

# CONFIG_HUAWEI_SENSOR_MT9V114 is not set

CONFIG_HUAWEI_SENSOR_MT9P017=y

I just made some changes based on these findings to my own kernel configuration, but mine wasn't that too few cameras were enabled, mine was too many.

Huawei provides a standard msm7x27a configuration which they must use for more devices, as the one used on the U8815 is a little different.

[Guest omegamoon]

Omegamoon kernel - OC 1228Mhz (performance, sio), Infusion B03 ICS ROM, Chainfire3d - PowerVR plugin, charging

PNY SDHC card, class 10,16GB

a score of 3849 is very high, nice :) ...

Just gave it a try, using stock B936 ICS. I scored 3810 (see screenshot below). This is at 1.31GHz. Bare in mind though that the score is just a rough indication of system performance, and it's influenced by a lot of things. I did another test just after this one, and the score was 3785.

Edited September 12, 2012 by omegamoon

[Guest omegamoon]

@omegamoon,

great work, thanks for this. It works great on the newest Infusion build. Like greased lightning now B)

For me its stable to the highest setting (in the Infusion processor settings) which is 1.325. I imagine No Frills may provide higher values?

Anyway, its probably not the best to keep it at full wack is it? I mean for wear and tear? Just wondering what people's thoughts are on that.

And on the lower side, I find 122 doesn't wake up very nice. Anyone else experience this as well?

All the best,

Bill

Could you try running a AnTuTu benchmark at 1.325? If I do that it reboots at the 3d performance test. I use 122 as min freq and I don't see any problems waking up the device

[Guest omegamoon]

Nice, you're lucky to be able to hit 1.325. It can be compiled to go faster. Omega has put in the values all the way up to 1.5 but they're commented out

// Users report that CPU is stable in between 1.2 and 1.3GHz

{ 1, 1270000,	ACPU_PLL_4,  6,  0, 158750,   3,  7, 200000 },

{ 1, 1280000,	ACPU_PLL_4,  6,  0, 160000,   3,  7, 200000 }, // PLL4 max of 66.7MHz

{ 1, 1290000,	ACPU_PLL_4,  6,  0, 161250,   3,  7, 200000 },

{ 1, 1300000,	ACPU_PLL_4,  6,  0, 162500,   3,  7, 200000 },

// Up to 1.3GHz is stable but... Let's go bananas!!

{ 1, 1310720,	ACPU_PLL_4,  6,  0, 163840,   3,  7, 200000 },

{ 1, 1324800,	ACPU_PLL_4,  6,  0, 165600,   3,  7, 200000 },

/*

{ 1, 1344000,	ACPU_PLL_4,  6,  0, 168000,   3,  7, 200000 },

{ 1, 1382400,	ACPU_PLL_4,  6,  0, 172800,   3,  7, 200000 },

{ 1, 1420800,	ACPU_PLL_4,  6,  0, 177600,   3,  7, 200000 },

{ 1, 1459200,	ACPU_PLL_4,  6,  0, 182400,   3,  7, 200000 },

{ 1, 1497600,	ACPU_PLL_4,  6,  0, 187200,   3,  7, 200000 },

{ 1, 1536000,	ACPU_PLL_4,  6,  0, 192000,   3,  7, 200000 },

*/

I commented the higher frequencies out since the device becomes unstable or doesn't even boot. I could add smaller steps to see what the ultimate stable value is. I'll try that later

[Guest Potatoes]

Those sdcard card scores are nice :D is that using the VR scheduler? On the subject of that, whenever you next update github can you chuck in the cpu and io schedulers. I've managed to add them myself but for the life of me can't get VR compiling, always bombs out with

  CC      block/vr-iosched.o

block/vr-iosched.c:417: error: unknown field 'elevator_queue_empty_fn' specified in initializer

block/vr-iosched.c:417: warning: initialization from incompatible pointer type

error, forbidden warning: vr-iosched.c:417

make[1]: *** [block/vr-iosched.o] Error 1

make: *** [block] Error 2

But no rush :)

[Guest omegamoon]

I have wifi toggle freeze too. I thought it was a B936 related issue and reading some posts here I'm not the only one suffering. Switching to geno.kolar kernel that has never failed... for the time being :)

During the AnTuTu tests I did yesterday I witnessed the same wifi toggle freeze a few times. Too bad that every time this happened I didn't have the device connected to my pc, so I still don't have a debug log :(

[Guest omegamoon]

cheers, I'm going to try the llama app for that. Staying steady for 24 hours @ 1.3ghz, smartassv2 and vr scheduler. Going back up to 480mhz as minimum. Seems like anything less is slow to wake up.

Slow to wake up? I really don't know what you mean... ah, wait, I just set the minimum to 480Mhz myself, just to see if I see any difference while waking up the device ;)

[Guest omegamoon]

This kernel breaks camera for me in infusion. Where is genokolar kernel?

I hope you haven't switched yet... could you provide me a dmesg log right after you did a reboot? I just want to check what camera sensor your device has.

[Guest Potatoes]

I commented the higher frequencies out since the device becomes unstable or doesn't even boot. I could add smaller steps to see what the ultimate stable value is. I'll try that later

Ooh glad your here omega so I can pick your brain :P adding new frequencies I figured you just divide the frequency by 8 to get the AHB-KHz value so like 1210000 would be 151250 and so on. But your comment

{ 1, 1280000,    ACPU_PLL_4,  6,  0, 160000,   3,  7, 200000 }, // PLL4 max of 66.7MHz

made me think whether there's a link between the cpu frequency and frequencies around the rest of the phone like in old school overclocking where overclocking the cpu would also overclock say the pci until you hit certain dividers. So I didn't add any new ones for myself on the safe side.

Also, do you know whether the voltage table here works? Reducing them did seem to make my phone unstable so I think they probably don't but I have no idea for sure.

[Guest omegamoon]

If you read back a good few posts you will see different cameras have been used on the same phone . ZTE also used different parts.

Currently there are two camera sensors supported, since only those two are spotted in the wild. I can add support for additional sensors of course. Just don't tell me you've got a Hasselblad H5D sensor :D

[Guest omegamoon]

Those sdcard card scores are nice :D is that using the VR scheduler? On the subject of that, whenever you next update github can you chuck in the cpu and io schedulers. I've managed to add them myself but for the life of me can't get VR compiling, always bombs out with

  CC	  block/vr-iosched.o

block/vr-iosched.c:417: error: unknown field 'elevator_queue_empty_fn' specified in initializer

block/vr-iosched.c:417: warning: initialization from incompatible pointer type

error, forbidden warning: vr-iosched.c:417

make[1]: *** [block/vr-iosched.o] Error 1

make: *** [block] Error 2

But no rush :)

Sure, I will add them later

[Guest timfimjim]

You're right, the proposed size is too high. I'm looking into this "short write" problem. Don't know why it occurs, it sometimes does. After that I have to reboot to give it another try.

Well on my nexus 7 I've dabbled in zRAM and it would seem 128mb is approaching optimum so maybe it's not too high. But seeing as though the N7 has double the RAM do you think it might be good to try 64mb of zRAM on the g300? I'd try it but I always just get the same short write error. Might try again today though.

Currently there are two camera sensors supported, since only those two are spotted in the wild. I can add support for additional sensors of course. Just don't tell me you've got a Hasselblad H5D sensor :D

haha!!

[Guest omegamoon]

Ooh glad your here omega so I can pick your brain :P adding new frequencies I figured you just divide the frequency by 8 to get the AHB-KHz value so like 1210000 would be 151250 and so on. But your comment

{ 1, 1280000,	ACPU_PLL_4,  6,  0, 160000,   3,  7, 200000 }, // PLL4 max of 66.7MHz

made me think whether there's a link between the cpu frequency and frequencies around the rest of the phone like in old school overclocking where overclocking the cpu would also overclock say the pci until you hit certain dividers. So I didn't add any new ones for myself on the safe side. Also, do you know whether the voltage table here works? Reducing them did seem to make my phone unstable so I think they probably don't but I have no idea for sure.

that's right, dividing the frequency by 8 will get you the AHB-KHz value. I searched for specs and found the PLL4 max of 66.7MHz mentioned somewhere. I tried and that seems indeed to be the case. If you look a bit further down in the code, you'll find that I change the PLL4 speed if going over 1GHz by dividing the frequency by 19200

// Change the speed of PLL4

writel(hunt_s->a11clk_khz/19200,PLL4_L_VAL);

[/code]

If you divide 1280000 though 19200 this will give you the PLL4 frequency of exactly 66.667MHz. My guess is that this is the max according to the books. Going any higher you will have to see for yourself if the device remains stable or not. There is a relation between the CPU frequency and other parts, for instance the GPU, as mentioned before. I didn't touch the voltage table because, as you found out yourself, it makes the device very unstable. It could work though, if you have enough patience tweaking it ;)

[Guest Potatoes]

that's right, dividing the frequency by 8 will get you the AHB-KHz value. I searched for specs and found the PLL4 max of 66.7MHz mentioned somewhere. I tried and that seems indeed to be the case. If you look a bit further down in the code, you'll find that I change the PLL4 speed if going over 1GHz by dividing the frequency by 19200

// Change the speed of PLL4

writel(hunt_s->a11clk_khz/19200,PLL4_L_VAL);

If you divide 1280000 though 19200 this will give you the PLL4 frequency of exactly 66.667MHz. My guess is that this is the max according to the books. Going any higher you will have to see for yourself if the device remains stable or not. There is a relation between the CPU frequency and other parts, for instance the GPU, as mentioned before.

Ahh I see, thanks for the explanation, that's awesome.

I didn't touch the voltage table because, as you found out yourself, it makes the device very unstable. It could work though, if you have enough patience tweaking it ;)

That's the thing, I ended up compiling quite a few kernels reducing the value in the table a notch and it didn't seem to have any effect of stability. So I tested one build where I changed all the values to "1" expecting the phone to just fail on boot, but it didn't so I not sure whether the voltages are taken from the table. Wished there was some way to measure voltage.

[Guest Dazzozo]

During the AnTuTu tests I did yesterday I witnessed the same wifi toggle freeze a few times. Too bad that every time this happened I didn't have the device connected to my pc, so I still don't have a debug log :(

I don't understand how they can apparently compile the same source without this issue.

[Guest Colossae3.23]

Could you try running a AnTuTu benchmark at 1.325? If I do that it reboots at the 3d performance test. I use 122 as min freq and I don't see any problems waking up the device

Tried it @ 1.325 and 1st time it froze and only a battery pull brought it back, then tried @ 1.3 & 1.311 and got 3569 & 3672 respectively. Then tried full wack again and it rebooted just where you said.

Re: the slow wake. It came to life OK, but I thought touch response was sluggish for 20-30 seconds. I've gone back to 122 and I'll try it again for whole day.

[Guest Davidoff59]

antutu ran fine for me at Max. Got around 3500 on B936 with cheap SD card. Had a reboot when I went on opera to post this though so running one step down from Max and its very fast phone now