thing is NO ONE from nokia or pretty much any other company would say yeah its "safe" to overclock or mod their devices.

but everything produced has to have operational tolerances. if this chip couldn't run over 600 mhz it would not have been released as a 600.....

it also seems most of the anticlockers also choose to ignore the voltage drop in the newer kernel/settings and the gains in life this would give at comparable freqs.....

I was criticizing the claims that overclocking is safe because device feels cooler. and replied to matan who seemed to claim that single end user reports are better facts than opinion of hw engineer who has been designing the device.

hw engineer who unfortunately is an employee of Nokia.
If I was him, I would also say it's not safe to overclock. Legal reasons.

btw, I never said device feels cooler. Just going by technical theory.
and I'm not new to overclocking cpus and was an electrical engineer.

Underclocking isn't bad for the CPU. The stock speed is 600MHz.

you could use the dbus-scripts to change the minimum freq. to 500 when
the screen is unlocked and back to 125 when it's locked.
As other people have already mentioned, you could also stick to 500 with a lower voltage kernel.

Perhaps the thread could benefit from an overclocker's touch.

First, let me clear a few things up, because while I read most, I didn't read all the replies, but it seems like you people have the strangest views on overclocking.

a) Overclocking does little to a CPU. What a CPU does, overall, is transform its electricity bill into heat (mostly) and EM (very little). As a result, the stress on the device is directly related to power consumption (and power dissipation).

b) It is not the frequency that does damage to a CPU unless you clocked it so far it doesn't start. Frequency is a component in stress because it is a component in power/heat.

c) Power consumption is: P = V^2 * F * C. That is, voltage squared, multiplied by frequency and capacitance. C is constant and specific to each CPU, so what you get is a linear increase in power (and heat) with frequency and an exponential increase with voltage. Overvolting matters more.

The CPU that rung the machine I'm posting from has a 33% in frequency, from 3 GHz to 4 GHz. It's a Q9650. As a result, the power increased (I had to ante the voltage) from 90W average to 225W average (there's a pump involved - and an analog amp meter hooked to the power system).

This is why the difference from 800 to 900 to 1000 MHz matters so.

d) As long as you can dissipate the heat and the voltage stays below the reversal/frying of components, there is nothing wrong with OC on a realistic scale.

While it is well known OC will destroy components, with decent voltage, a decent clock and ACTUALLY APPROPRIATE COOLING, you can run a component up to most of its useful life.

With no perishable components inside, a CPU like ARM can most likely run until materials age. I'm gonna call it 10 years. As components heat, aging is accelerated. By how much is directly dependent on material and heat, bust most components have a threshold where life shortens quickly. Under that threshold, an OC system will probably live 9 years.

For example, a Core 2 will last a lifetime (whatever) at 55 degrees, a decent time at 60 and a short while at 65. At 70, you should probably be on your way to the store. Actually, you should already have a backup on your desk.

This is when you have a sensor inside the damned chip. When you feel it through 1 cm of plastic and isolating materials, the difference is almost not noticeable. As a result, you could fry it and not even know.

That said, enemy number one in a compact device isn't CPU heat as overall device heat. If the GPU is working like nuts, along with battery, wifi, etc, the CPU can't dissipate from 50 to 25 degrees, but from 50 to 45.

As a result, its heat transfer capacity drops 5-fold, meaning you only have 20% of the power available. That sound very little until you look up at the formula and realize that 20 hear is likely more than 50% speed, if you throttle voltage to match.

It is unlikely that an N900 is unable to dissipate the heat it generates at 600 MHz, so it safely runs. However, it assumes that it runs hand-held, or on a table, or solo. When things get out of hand, bad stuff happens.

So, before I wear out the keyboard, the real issue is this:

A compact machine that is N900-like is able and willing to run at max designed speed if cooled appropriately. Cut any corners in cooling (you know, like building world's most crammed phone), and you drop at 800. Add the fact that other components also heat up. Add the fact that the idi... sorry, user will forget the phone on the dashboard and it will heat up, then pick it up and set it to navigate with GPS, online, and rotate/zoom the map in real time.

So there are safeties that insure survival. I'm pretty sure you can run 900+ if you take care of the CPU to run alone, or less stressed than pinned at full and not while using a lot. Also, keep it cool.

It is also true that another user is able to fry the thing by letting it out in the sun, running some ghastly game that pushes GPU, CPU, WIFI, in turn making battery hot and so on.

IPhone has thermal protection, google it up. So do most if not all PDAs. So do most PNAs. I'm guessing we have one too. Don't rely on it though.

Oh, and one last thought. Unless you drop it, CPUs don't fry in a week. There hasn't been enough time for the OCd N900 to start dropping like flies. I'm thinking 3-6 more months before the first ones start to act up. And by that I mean stuff like stability, inability to boot, randomly missing hardware, stuff like that, way before the device gives out. And it never comes back.

[pheew]

Sure. And that was an exact reason why Nokia (!???) locks N900 at 600MHz during phone call from start until your party answers...

It was discovered that running at HIGHER frequency consumes LESS power PER CPU CYCLE. So, for finite load it is much more efficient.

Of course, it is not for stupid scripts like while true; do; true; done...

I can add something interesting here - I also ran that test and monitored battery and heat. I should say that it is not very big actually (I have SmartReflex switched ON). Actually, the back cover is not heated at all.

But I definitely see a HUGE power consumption during some video processing even with thumbnailer tracker in background. It is definitely not CPU (and I don't use overclocked kernel) but it is on CPU chip. I suspect it is SGX or OneNAND R/W - some people reported a very hot device and battery exhausted fast with a moderate CPU usage.

So, I don't think the CPU heat is an issue for this specific ARM CPU. Nokia doesn't rush to publish PR1.2 which should fix at least some of video related issues.

This formula doesn't work well for this CPU. Look this - http://talk.maemo.org/showpost.php?p...postcount=3286

It is clear that frequency increase 4 times (from 125 to 500) increases CPU energy consumption not to 4 times but 30% or (in worst case) - twice.