[beta] Run your Hero at 660mhz!

[Guest Lox-]

Beta testing of a new kernel that allows over clocking up to 660mhz. I have no time to do a lot of testing. So far MCR rom 3.2 is stable and fast with it. So I am calling for testers. The kernels are using ramzswap.

Download

Note: It will not work on 2.x rom as it is a 2.6.27 kernel.

[Guest tristan202]

Beta testing of a new kernel that allows over clocking up to 660mhz. I have no time to do a lot of testing. So far MCR rom 3.2 is stable and fast with it. So I am calling for testers. The kernels are using ramzswap.

Download

Note: It will not work on 2.x rom as it is a 2.6.27 kernel.

Can't wait to try it out.

[Guest String Theory]

I would love to test it, but how big is the chance of overheating my Hero?

[Guest Paul]

Awesome Lox-! ;)

P

[Guest majczu]

Don`t work. Write this commend on ADB: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq

BogoMips is a not of Freq CPU.

Don`t download this.

[Guest String Theory]

Don`t work. Write this commend on ADB: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq

BogoMips is a not of Freq CPU.

Don`t download this.

BogoMips is 657,81 after installing Lox's kernel on MCR 3.2.

Edited February 9, 2010 by String Theory

[Guest majczu]

Write this commend on ADB: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq

Freq screen turns off: 245 screen turns on: 528. Don`t work!

BogoMips is a not of Freq CPU.

[Guest grassiekuik]

I would love to test it, but how big is the chance of overheating my Hero?

I second this

[Guest String Theory]

My phone becomes very slow... SetCPU doesn't work well, can't set it to keep 660 as it's max freq. Everytime I put the phone to sleep it puts itself at 254 and when it wakes up it doesn't scale up. For the time being I'm going back to Teknologists kernel.

[Guest commodoor]

Ok let say it worked, Whats the battery life?

[Guest majczu]

BogoMips (from "bogus" and MIPS) is an unscientific measurement of CPU speed made by the Linux kernel when it boots, to calibrate an internal busy-loop.[1] An oft-quoted definition of the term is "the number of million times per second a processor can do absolutely nothing."[2][3]

BogoMips can be used to see whether it is in the proper range for the particular processor, its clock frequency, and the potentially present CPU cache. It is not usable for performance comparison between different CPUs

This is not a CPU Frequency!!!!

[Guest String Theory]

There was another problem I had with Lox's kernel: the screen wasn't as bright as it should be. The screen had a blueish shade.

[Guest stevenz]

Sometimes bogomips DOES = cpu frequency just by nature of the CPU architecture and kernel config. In the case of the Hero it aligns pretty closely, and certainly if it goes from approx 528000 to approx 660000 then it's obviously working.

In the single-core, non-hyperthreaded CPU days Bogo's were generally a good match for the actual CPU clock, or were a fairly exact multiple (usually 2x) of the clock. While it's not an accurate measurement, it's often a good representation.

[Guest ippoP]

First of all, the word "overclocking" has a specific meaninng, which to me is this one: http://en.wikipedia.org/wiki/Overclocking.

Second. if at any time a CPU clock changes, under Linux 2.6 kernel you'll see the new clock rate in /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq.

Third, the BogoMips (http://en.wikipedia.org/wiki/BogoMips) is calculated only once at the boot and then never again.

Fourth, it's normal when changing kernel (version) to get better (or worse) BogoMips scores, because of a number of reasons, among which I would number:

- quality of CPU tuning

- amount of cache available

- quality of code generated by the assembler/compiler used.

So what I'd say is that Lox's kernel can run (at least) the BogoMips loop faster (more efficiently) than the "normal" kernels.

If this kernel is faster in (almost) all other stuff, then Lox work is a great breakthrough, no question on this and kudos to him from all of us.

Otherwise is just a "false positive", which can happen to anyone of us.

So what?

Simply, run as many benchmark tools as possible before and after the Lox kernel is installed and booted.

Possibly in the same way (external power supply, no CPU frequency tool running, long screen blanking delays and so on).

Then compare the numbers.

Anything else looks to me worthless discussing.

This is my EUR 0.01 contribution.

[Guest MacCarron]

First of all, the word "overclocking" has a specific meaninng, which to me is this one: http://en.wikipedia.org/wiki/Overclocking.

Second. if at any time a CPU clock changes, under Linux 2.6 kernel you'll see the new clock rate in /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq.

Third, the BogoMips (http://en.wikipedia.org/wiki/BogoMips) is calculated only once at the boot and then never again.

Fourth, it's normal when changing kernel (version) to get better (or worse) BogoMips scores, because of a number of reasons, among which I would number:

- quality of CPU tuning

- amount of cache available

- quality of code generated by the assembler/compiler used.

So what I'd say is that Lox's kernel can run (at least) the BogoMips loop faster (more efficiently) than the "normal" kernels.

If this kernel is faster in (almost) all other stuff, then Lox work is a great breakthrough, no question on this and kudos to him from all of us.

Otherwise is just a "false positive", which can happen to anyone of us.

So what?

Simply, run as many benchmark tools as possible before and after the Lox kernel is installed and booted.

Possibly in the same way (external power supply, no CPU frequency tool running, long screen blanking delays and so on).

Then compare the numbers.

Anything else looks to me worthless discussing.

This is my EUR 0.01 contribution.

Agreed. If i can make some time i will start benching with the "standard" kernel this afternoon, then switch to Lox's and compare the results. I think that 50 bench iterations on each kernel should provide statistically acceptable results.

[Guest Greg_C]

Note: It will not work on 2.x rom as it is a 2.6.27 kernel.

Any plans for a version that will work with 2.X Roms? Would love to give this a try but really like the 2.1ness of my Hero at present.

[Guest ippoP]

@Lox_Dev

Could you explain the kind of work you did? Not the details, just the outlook of the whole thing.

I'm still convinced that you've found a way to run code faster on the CPU, more than overclock it.

Could anyone running Lox_Dev's kernel post the contents of /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq ?

[Guest Paul]

Any plans for a version that will work with 2.X Roms? Would love to give this a try but really like the 2.1ness of my Hero at present.

I'm sure this will happen WHEN the 2.6.29 kernel sources are released (and not before)...

P

[Guest grassiekuik]

@Lox_Dev

Could you explain the kind of work you did? Not the details, just the outlook of the whole thing.

I'm still convinced that you've found a way to run code faster on the CPU, more than overclock it.

Could anyone running Lox_Dev's kernel post the contents of /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq ?

if i did it correctly it sill says 528000

[Guest ippoP]

if i did it correctly it sill says 528000

If you read that pseudo-file with the Lox_Dev kernel running, that's the current clock speed.

Which is the hardware maximum (AFAIK) on the Hero and clearly shows it's not an overclocking thing at all.

And, grassiekuik, thanks for the feedback, so Paul won't loose any precious time on it ...

Edited February 10, 2010 by ippoP

[Guest grassiekuik]

If you read that pseudo-file with the Lox_Dev kernel running, that's the current clock speed.

Which is the hardware maximum (AFAIK) on the Hero and clearly shows it's not an overclocking thing at all.

I did it with lox' kernel running Strange however that overclockwidget says it's running @ 660 as well as setcpu. I'm not much of a techie to know all that. But maybe lox can enlighten us. And it's still in beta though

[Guest ippoP]

I did it with lox' kernel running Strange however that overclockwidget says it's running @ 660 as well as setcpu. I'm not much of a techie to know all that. But maybe lox can enlighten us. And it's still in beta though

That's strange, for sure... but it's hard to get lies from the kernel ...

[Guest grassiekuik]

That's strange, for sure... but it's hard to get lies from the kernel ...

Lol that's true. Here's a screenshot from setcpu

[Guest ippoP]

Lol that's true. Here's a screenshot from setcpu

That's the content of /proc/cpuinfo and, as said, the bogomips are computed only once at boot in a tight loop.

The faster you make that loop, the higher the bogomips.

If only Lox could shed some light on his approach ...

[Guest Trekster2000]

That's the content of /proc/cpuinfo and, as said, the bogomips are computed only once at boot in a tight loop.

The faster you make that loop, the higher the bogomips.

If only Lox could shed some light on his approach ...

We have to assume that if it infact was an "overlcock" of the CPU such benchmarking tools as Linpack would show a significant increase in MFLOP/s throughput, as this is not the case(i haven't seen any) i believe you are correct in assuming that this infact does nothing to overclock the CPU.