[daniele tabanella]

i've used this program today and i've found that my pmu was set to 533mhz

now i have it set to 867mhz (as it should be) and i've found that in xbench i can gain 2 points on 80.

btw this is what the author say on his blog:

Explanation of sysctl hw.cpufrequency and what CPUfrequency does..._

My friend, Jay Savage, who understands the BSD underbelly of OS X quite a bit better than I, explains what sysctl hw.cpufrequency does:
The poster at version tracker is correct that sysctl doesn't return a real number. It returns a constant passed to the system at system start-up (and possibly wake-up, I'm not sure) by the PMU. The kernel and PMU will modify the actual speed on-the-fly based on load numbers and energy-saver setting in the pref pane, and sysctl will not return that number, you need to actually pool the processor to find out what it's doing at any given time...this is what the bogomips calculated a Linux start-up are all about.


For your purposes, though, the real numbers don't matter, because the bug in question involves the PMU essentially forgetting what the real speed is and setting the max Mhz artificially low. Effective clock speed may well often be below the number reported by sysctl (a kind of variable clock speed is what makes mobile chips possible: why waste power if your load numbers are below 1?), but it will never be above the number reported by sysctl. So if sysctl is reporting 400Mhz for a 800Mhz processor, you know something is seriously wrong. Even under full load, the PMU will not allow the CPU speed up past 400Mhz, which means you're never getting more than 50% out of your hardware. Get that back up to 867, and you can theoretically get 867Mhz out of your chip. You won't get that all the time, especially on battery power, but better and actual 400 out of 800 than an actual 200 out of 400.[bold emphasis mine]

So, the CPUfrequency scripts work fine for what they are meant to do, which is simply to report the maximum possible number (speed in hertz) your computer currently thinks in the processor is capable of achieving. If it reports back 5333333334 and you have an 867mhz chip, as happened with me earlier this week, then your PMU needs to be reset, so that the computer will realize that there is more processing power available there, should it be needed...

this is the link to download:

<http://3650anda12inch.blogspot.com/>

[cambro]

Originally posted by daniele tabanella:
i've used this program today and i've found that my pmu was set to 533mhz

now i have it set to 867mhz (as it should be) and i've found that in xbench i can gain 2 points on 80.

This is not a significant result. Reunning X-bench gives a frequency distribution of speed scores. Reunning X-bench and getting a score that differs by 2 points is WELL WITHIN the range of error for any given estimate. Thus, your X-bench score did not change significantly after doing whatever it is that you think you did.

If you don't believe this, rerun X-bench a bunch of time and keep track of your scores. There will be a mean score and a bunch that fall within +/- 2 points of that mean.

EDIT:: Read this other thread.

[daniele tabanella]

i believe you, and i think that 2 points (or more) can't be significant, even with over 200 mhz of difference.

the fact is that this guy claims that:"even under full load, the PMU will not allow the CPU speed up past 400Mhz, which means you're never getting more than 50% out of your hardware.."

and "that the computer will realize that there is more processing power available there.."

so... if this is true.. where is the proof?

thanks..

Daniele