Binary patched kernels for Pipo models

[A64]

A bit off: dou you guys think, android 5.x will be available for these PiPO tablets ? RAM is enough, CPU as well, .. what do you think ? I bought the M9Pro 3G for my mother so I'll need one for myself.. I pretty much like the this one now.. shall I consider another brand if I'd like to be a bit future-proof or is the M9Pro pretty ok now ?

[rrileypm]

Binary patched kernels for Pipo models

It depends on if Rockchip puts out an SDK for it on the 3188. I haven't heard if they plan to or not.


Randy

[Balrog]

which is your antutu result?
What about temperatures?
Why did you use ram Voltage to 1.25V? Is it not stable to you at 1.2V?

Antutu scores with 18565, Quadrant with 5467 and I get 37,1 FPS from Epic Citadell at High Quality with full Resolution.
The ram voltage value is from the stock Kernel.

So far I am satisfied with the values. Battery life and the further behavior in terms of stability I will see in the next few days when I work with the tablet.

[rrileypm]

FYI, I was able to patch some kernels for the M8 Pro, myself with the pearl script...using your comments in this thread as a resource.

Thanks!!!!!!

[spectra9]

Is there an app that can show the GPU and RAM clock/voltage being used, kinda similar to CPU-Z for CPU?

[giouncino]

Hmm maybe you didn't understand my question - or I questioned the wrong way.
My question was if you can feel the difference. Synthetic benchmark (AnTuTu) and pure numbers don't help me much, scrolling, browsing, playing film, music, responsiveness, whatever.. - I need to know, but HONESTLY .. if you feel the tablet behaving more smooth due to overclocking or don't.

Regarding overclocking I'm not a beginner either, doing it since 486 times, best fun was the DX4-100 @ 120MHz, online switchable with the PC case lock . Same for 5x86-133 (and 160) Then the Pentiums etc..

But I think I'll leave my M9Pro as is. Consumption raises exponentially as we increase voltage and frequencies, so 10% overclock gain will kick your battery's ass by more than 10% consumption. And my stock-clocked tablet with RileyROM 2.0 and Nova Launcher (and lots of apps) already does 18.100-18.500 in AnTuTu.

What would be more interesting is to spare some more power: CPU/GPU untouched but less voltage. Same for RAM-s: stock frequency but then let it have less voltage like 1.1V or so, even 1.0 - who knows.. Have you tried this direction ?

I assume the screen is consuming the most, but for sure lowering voltages on stock clocks would bring couple of extra minutes/hours during everyday use as well.

If you read my post #54, you can see that I decrease almost EVERY voltage and at the same time I've increased maximum frequency.
For this tablet the best performance shall be a trade off between power consumption and pure performance.

About overclocked performances: In my office we've bought 7 pipo m6pro so I can make a REAL life comparison side by side.
I can tell you that with my latest kernel, with respect to the stock almost every app is starting (and ready to be used) in almost 1 second less than the stock kernel.
I'm sure that also battery consumption has decreased a lot: I tested yesterday a lot with benchmarking and I feel the tablet run cool, so I can assume less heat = less power consumption.

Chinese people give to there tablet too much voltages to have more tablet usable, but I think that the most part of our tablet could perfectly run with a lot less voltages for CPU and I hope also RAM and GPU.

The RAM of this tablet seems too much underclocked. I can do more and more. I'm interested in finding the real limit.

Now that I'm confident that my last kernel is working stable on my CPU I'll going to check the frequency wall of our RAM. They should be DDR3 and because of that could run really fast in my opinion.
The RAM is shared between CPU and GPU and because of that are crucial for the overall performance of the tablet itself.

[giouncino]

Antutu scores with 18565, Quadrant with 5467 and I get 37,1 FPS from Epic Citadell at High Quality with full Resolution.
The ram voltage value is from the stock Kernel.

So far I am satisfied with the values. Battery life and the further behavior in terms of stability I will see in the next few days when I work with the tablet.

I suggest you to use my frequency parameters that can give you better performance test results (read post #54).
As far as I understand you can use ram to 720Mhz/1.2V or 800Mhz/1.3V stable. Try them!

[claro966]

I suggest you to use my frequency parameters that can give you better performance test results (read post #54).
As far as I understand you can use ram to 720Mhz/1.2V or 800Mhz/1.3V stable. Try them!

Hello giouncino,

have you already released (or worked on) a kernel for m6pro such this:

CPU 1.7GHz
GPU 660MHz
RAM 800MHz

whatever the related voltage would be?

Can you share it in case so to be tested?

Thanks in advance for your time and job

Claudio

[Balrog]

I suggest you to use my frequency parameters that can give you better performance test results (read post #54).
As far as I understand you can use ram to 720Mhz/1.2V or 800Mhz/1.3V stable. Try them!

I made a kernel for my M6 max 3G with the following values:

CPU (in MHz): 312, 504, 816, 1008, 1200, 1416, 1704
GPU (in MHz): 133, 200, 266, 400, 600, 660
RAM (in MHz): 720

It is very similar compared to the version before in terms of max. speed in Benchmarks (Antutu 18775, Quadrant 5610 and 36,8 FPS in Epic Citadell (HQ with full resolution)) but CPU Spy shows a smoother usage of each step. The version before shows lesser usage of the second highest step (1608 MHz).

For now I stick with this as I do not want go any higher with the RAM.
For those who want to try it I append it here:
kernel_cpu1704_gpu660_ddr720_v4.zip

[A64]

If you read my post #54, you can see that I decrease almost EVERY voltage and at the same time I've increased maximum frequency.
For this tablet the best performance shall be a trade off between power consumption and pure performance.

About overclocked performances: In my office we've bought 7 pipo m6pro so I can make a REAL life comparison side by side.
I can tell you that with my latest kernel, with respect to the stock almost every app is starting (and ready to be used) in almost 1 second less than the stock kernel.
I'm sure that also battery consumption has decreased a lot: I tested yesterday a lot with benchmarking and I feel the tablet run cool, so I can assume less heat = less power consumption.

Chinese people give to there tablet too much voltages to have more tablet usable, but I think that the most part of our tablet could perfectly run with a lot less voltages for CPU and I hope also RAM and GPU.

The RAM of this tablet seems too much underclocked. I can do more and more. I'm interested in finding the real limit.

Now that I'm confident that my last kernel is working stable on my CPU I'll going to check the frequency wall of our RAM. They should be DDR3 and because of that could run really fast in my opinion.
The RAM is shared between CPU and GPU and because of that are crucial for the overall performance of the tablet itself.

That's what I wanted as an answer, thx.
So it seems that overall consumption is still better than original, which means that undervoltage brings more efficiency regarding consumption than at the same time overclocking the chips. Thx !

Pls keep us followed on your work, I'm really curious where this tablet maxes out but don't have the "spare part" to sacrifice my mother's gift now - so I can't test. Nevertheless, thanks again

[giouncino]

I made a kernel for my M6 max 3G with the following values:

CPU (in MHz): 312, 504, 816, 1008, 1416, 1608, 1704
GPU (in MHz): 133, 200, 266, 400, 600, 660
RAM (in MHz): 720

It is very similar compared to the version before in terms of max. speed in Benchmarks (Antutu 18775, Quadrant 5610 and 36,8 FPS in Epic Citadell (HQ with full resolution)) but CPU Spy shows a smoother usage of each step. The version before shows lesser usage of the second highest step (1608 MHz).

For now I stick with this as I do not want go any higher with the RAM.
For those who want to try it I append it here:
[ATTACH]3953[/ATTACH]

I suggest you to replace 1416 with 1200 and 1608 with 1440. You will have better overall results using less power.
1608 and 1704 are too similar in frequency (have you seen my excel graph?)

[giouncino]

Hello giouncino,

have you already released (or worked on) a kernel for m6pro such this:

CPU 1.7GHz
GPU 660MHz
RAM 800MHz

whatever the related voltage would be?

Can you share it in case so to be tested?

Thanks in advance for your time and job

Claudio

I'll prepare it for you as soon as I'll arrive at home this evening,
stay tuned!

[Balrog]

I suggest you to replace 1416 with 1200 and 1608 with 1440. You will have better overall results using less power.
1608 and 1704 are too similar in frequency (have you seen my excel graph?)

There is a typo...
The correct used values in MHz for CPU are: 312, 504, 816, 1008, 1200, 1416, 1704

I correct the post in a minute.

[rceccleston]

Building upon the success found within this thread, the components within my M6 are now running the manufacture's ratings, rather than Pipo's underclocked ratings.

CPU - Rockchip RK3188 -- 1.6Ghz quad-core >> 1.8Ghz

GPU - Mali 400 MP -- 600Mhz quad-core >> 798Mhz

RAM - Hynix H5TQ4G83AFR-PBC 4X 512MB -- DDR528 >> DDR800

Antutu has rated these tweaks at 20874



All tweaks have passed Anutu's stability tests.

   "cpu1800h" =>
    [
      [900000, 312000],
      [975000, 816000],
      [1025000, 1008000],
      [1100000, 1200000],
      [1200000, 1416000],
      [1250000, 1608000],
      [1325000, 1800000],
     ],

  "gpu798h" =>
    [
      [975000, 133000],
      [1000000, 200000],
      [1050000, 300000],
      [1100000, 400000],
      [1250000, 600000],
      [1350000, 798000],
   ],

  "ddr800" =>
    [
      [1200000, 800001],
    ],

A link to my kernel can be found below.

[Pipo2000]

Building upon the success found within this thread, the components within my M6 are now running the manufacture's ratings, rather than Pipo's underclocked ratings.

CPU - Rockchip RK3188 -- 1.6Ghz quad-core >> 1.8Ghz

GPU - Mali 400 MP -- 600Mhz quad-core >> 798Mhz

RAM - Hynix H5TQ4G83AFR-PBC 4X 512MB -- DDR528 >> DDR800

Antutu has rated these tweaks at 20874

[ATTACH=CONFIG]3954[/ATTACH]

All tweaks have passed Anutu's stability tests.

[ATTACH=CONFIG]3958[/ATTACH]

   "cpu1800h" =>
    [
      [900000, 312000],
      [975000, 816000],
      [1025000, 1008000],
      [1100000, 1200000],
      [1200000, 1416000],
      [1250000, 1608000],
      [1325000, 1800000],
     ],

  "gpu798h" =>
    [
      [975000, 133000],
      [1000000, 200000],
      [1050000, 300000],
      [1100000, 400000],
      [1250000, 600000],
      [1350000, 798000],
   ],

  "ddr800" =>
    [
      [1200000, 800001],
    ],

A link to my kernel can be found below.

I'm also a guy that tries to find the limit of components but have my thoughts by the 1.8Ghz cpu and 798Mhz GPU.
Where did you find the information that says that Pipo 'underclocks' the RK3188 ?

This is a piece of the specifications that Rockchip (the MAKERS of the RK3188) do publish :

Quad-core Cortex-A9 processor with up to 1.6GHz performance

28nm HKMG process with low leakage and high performance

Quad-core Mali-400 GPU, supporting OpenGL ES 1.1/2.0 and OpenVG 1.1, up to 600Mhz

High performance dedicated 2D processor

Full memory support, including DDR3, DDR3L and LPDDR2

Now for CPU : UP to 1.6Ghz

For GPU : UP to 600Mhz
Source : http://www.rock-chips.com/a/en/News/.../0620/227.html

Now I know that some do publish that the RK3188 is a 1.8Ghz CPU, but they don't tell us if that can be done in every situation (i mean without proper cooling)

So i'm wondering, althow the 1.8Ghz and 798Mhz are reachable, if the RK3188 doesnt get too hot in the process and for some of us, maybe kill the CPU/GPU due to excess of heat ?