Overclocking the Pandora

[2bit]

Hey avid question asker 2bit here anyone know if overclocking the pandora has any significant benifet if so what are its risks thanks .

 

[DaMummy]

benefits are better n64 emulation, the norm clock is 800, no matter what official papers say, the risks are that Pandora can shut off, and reduce life expectancy from your lifetime, to your children's lifetime

 

[Linux-SWAT]

Qemu is probably the app that benefits the most from OC.


No real risk apart crashing the OS and eventually corrupt the SD cards.


The lifetime of the cpu is also shortened, but it's not a big deal if you don't run your OP 24/7.

 

[WizardStan]

To expand some on the CPU thing, the expected running time when clocked at stock voltage and clock speed running at 100% usage 24 hours a day every day is 24 years or so (200'000 hours)


Increasing the voltage and overclocking it, same situation of running it non-stop at 100% usage is only 5 years (40'000 hours).


It's a substantial loss, but one I don't think anyone should be concerned about.

 

[fusion_power]

This brings me to the point where I wonder, if with the new Super Zaxxon Kernel, OC is more battery friendly than before because of the improved powersaving features? So when I OC my Pandora from 600 to 800MHz and lets say, just use xfce Desktop stuff, is the battery time still identical to 600MHz, when the Processor is OC but not at full load? I can imagine thet with the new powersaving technics the MHz is an top value that can be reached but not have to be reached when the load is still low.


Or does it not work that way?

 

[diligentcircle]

benefits are better n64 emulation

That's wrong. Overclocking makes no difference for N64 emulation; the bottleneck is something else (GPU, I think).

 

[WizardStan]

Or does it not work that way?

It depends more on the OPP value. It's always been that an idle CPU with a max clock of 800Mhz used about the same power as an idle CPU with a max clock of 600Mhz, that idle power usage is now just significantly lower. That assumes that you set the OPP value to be the same for both, typically 3. If you overclock further, setting it to 5 to reach 1Ghz, then the higher voltage means more power even when idle.


Also busy loops (a program that isn't idle but isn't actually doing anything either, ie a loop which checks for keypresses as fast as it can) will consume 100% of the CPU to do nothing, however, and in such a case the 800Mhz will use a lot more power doing nothing, but such a loop is generally considered bad practice and shouldn't appear that often.

 

[pmprog]

Or does it not work that way?

It depends more on the OPP value. It's always been that an idle CPU with a max clock of 800Mhz used about the same power as an idle CPU with a max clock of 600Mhz, that idle power usage is now just significantly lower. That assumes that you set the OPP value to be the same for both, typically 3. If you overclock further, setting it to 5 to reach 1Ghz, then the higher voltage means more power even when idle.


Also busy loops (a program that isn't idle but isn't actually doing anything either, ie a loop which checks for keypresses as fast as it can) will consume 100% of the CPU to do nothing, however, and in such a case the 800Mhz will use a lot more power doing nothing, but such a loop is generally considered bad practice and shouldn't appear that often.

Did I read that if you set the OPP to 5, it'll drop down to 3/4 when running at 600MHz? Or should I manually switch it back down when not needing to overclock?

 

[WizardStan]

Did I read that if you set the OPP to 5, it'll drop down to 3/4 when running at 600MHz? Or should I manually switch it back down when not needing to overclock?

Now that you mention it, I remember a similar conversation, but I honestly can't recall if it was about it now doing this (because it didn't before) or if it still needs to be done.

 

[fusion_power]

Or does it not work that way?

It depends more on the OPP value. It's always been that an idle CPU with a max clock of 800Mhz used about the same power as an idle CPU with a max clock of 600Mhz, that idle power usage is now just significantly lower. That assumes that you set the OPP value to be the same for both, typically 3. If you overclock further, setting it to 5 to reach 1Ghz, then the higher voltage means more power even when idle.


Also busy loops (a program that isn't idle but isn't actually doing anything either, ie a loop which checks for keypresses as fast as it can) will consume 100% of the CPU to do nothing, however, and in such a case the 800Mhz will use a lot more power doing nothing, but such a loop is generally considered bad practice and shouldn't appear that often.

Thanx for the explanation. No, I don't want to torture my Pandora with OPP changing, or even use 800MHz as standard. But some programs like Android benefit from OC and I just searched for the best compromise in between speed and Battery time. So I guess I stay at my current method, just OC when it is needed fo certain programs and always stay at OPP3.


Thanks to MiniMenu it is an easy task to OC individual programs.

 

[_wb_]

To expand some on the CPU thing, the expected running time when clocked at stock voltage and clock speed running at 100% usage 24 hours a day every day is 24 years or so (200'000 hours)


Increasing the voltage and overclocking it, same situation of running it non-stop at 100% usage is only 5 years (40'000 hours).


It's a substantial loss, but one I don't think anyone should be concerned about.

This is interesting information. I am concerned about lifetime, and while 5 years at 100% usage sure looks like a lot, I guess it doesn't scale if you have less usage (e.g. 10 years at 50% usage is probably not true), and these are only _expected_ lifetimes, so you can always be unlucky and get a lower-than-average lifetime. Just like you can drown in a lake that is on average only 1 meter deep.


Any idea what those numbers would be if you don't overvolt but you do overclock? E.g. I can get 900MHz at 3 volt, does that decrease the lifetime too? By how much? And how reliable are these statistics? Obviously nobody actually took 1000 CPUs, ran them for a few decades at 100% usage until they exploded, and counted the average lifespan...

 

[Glissa_Sunseeker]

IIRC overclocking should not reduce the lifetime, cause it creates no/nearly no additional stress for the materials used in the CPU. So OCing should be fine for everyone.


I don't fully understand what you meant with your first question, but I'll answer the two variants I can read out of it.


If overvolted to OPP5 the CPU usage shold have little effect on the lifespan. I assume you would get average 5 years out of it, regardless if 50% load or 100%.


In terms of usage-time, when you use the overvolted Pandy only half the time, you naturally would get an average lifetime of 10 years.


(40'000h/24h a day-->about 1'666 days lifetime; 40'000/12h a day-->about 3'333 days lifetime)


Feel free to correct me if I'm wrong

 

[Linux-SWAT]

It's guaranteed by TI.


http://www.ti.com/product/omap3530

 

[_wb_]

If overvolted to OPP5 the CPU usage shold have little effect on the lifespan. I assume you would get average 5 years out of it, regardless if 50% load or 100%.


In terms of usage-time, when you use the overvolted Pandy only half the time, you naturally would get an average lifetime of 10 years.


(40'000h/24h a day-->about 1'666 days lifetime; 40'000/12h a day-->about 3'333 days lifetime)

The thing is: I never do a full power off, and I rarely even suspend. Usually I just close the lid and that's low-power enough for me. The thing is, that means I am using the CPU 24h a day, although of course usually it has close to 0% load. So I don't feel comfortable overvolting to OPP5, since I want this Pandora to last more than 5 years - at some point I'll want to give it to my kid and buy a new 1GHz one or perhaps a Pandora 2 for myself

 

[Glissa_Sunseeker]

Oh, ok. In this case thats bad, of cause...


If I'm not mistaken there should be a way to automatically ajust the OPP level. But since I ain't got no Pandora till now, I'm not certain about that. I just seem to remember something like that.


Result of a quick search:


A part of the official Ti documentation, full one can be found here:

Recommended Operating Conditions


OMAP3530/25 Applications Processor


www.ti.com SPRS507F – FEBRUARY 2008 – REVISED OCTOBER 2009


All OMAP3530/25 modules are used under the operating conditions contained in Table 3-3.


Note:


To avoid significant device degradation for commercial temperature OMAP3530/OMAP3525 devices (0°C


≤ T


j


≤ 90°C), the device power-on hours (POH) must be limited to one of the following:


• 100K total POH when operating across all OPPs and keeping the time spent at OPP5-OPP6 to less


than 23K POH.


• 50K total POH when operating at OPP5 - OPP6.


• 44K total POH with no restrictions to the proportion of these POH at operating points OPP1 - OPP6.


To avoid significant device degradation for extended temperature OMAP3530A/OMAP3525A devices


(-40°C ≤ T


j


≤ 105°C), the following restrictions apply:


• OPP5 and OPP6 are not supported.*


• The total device POH must be limited to less than 50K.*


*If an extended temperature device is operated such that T


j


never exceeds 90C (-40°C ≤T


j


≤ 90°C) then


the OPP POH limits for commercial devices indicated above apply.

And about the autoadjust:

The TI OMAP processors have a power saving functionality that they call "SmartReflex" which is basically a dynamic clock and voltage scaling mechanism... as far as I understand it, the processor will (by default) automatically downscale the processor clock speed (MHz) and voltage levels (OPP settings) depending on the demand at any given time. In other words, even if you were to "overclock" to 1Ghz @ OPP5 and playing a game that only demanded ~600MHz => OPP3 the processor would automatically downscale to the fit the demand. Essentially the "overclock" settings only set the maximum levels that the processor is allowed to go up to but will only use what it needs.


What is not clear to me is how often and rapid the processor assess and changes the clock speed and voltage levels.


The bottom line is..."overclocking" and "overvolting" should not dramatically reduce the lifespan of the processor even when done to the extreme due to the above feature. To quote Douglas Adams... its "Mostly Harmless"...

 

[WizardStan]

This is interesting information. I am concerned about lifetime, and while 5 years at 100% usage sure looks like a lot, I guess it doesn't scale if you have less usage (e.g. 10 years at 50% usage is probably not true)

Not exactly a 1:1 scale like that, probably, but I'd be surprised if idle operation counted very much towards the consumption. I wouldn't worry about it at all. I forget the exact specifics but it definitely said these were the figures under full load. No load had a much, much longer lifetime, values that I can't remember off the top of my head right now, but so far into the future that I would be very surprised if there weren't other things in your Pandora failing first.

Any idea what those numbers would be if you don't overvolt but you do overclock? E.g. I can get 900MHz at 3 volt, does that decrease the lifetime too? By how much? And how reliable are these statistics? Obviously nobody actually took 1000 CPUs, ran them for a few decades at 100% usage until they exploded, and counted the average lifespan...

Not that I can recall. There may have been something about different clock speeds but all I remember is the different voltage levels.

 

[Linux-SWAT]

OPP switching is reported by dmesg so if someone want to make some tests...

 

[2bit]

Well , when the Pandora 2 comes out I'll sell the current one which I should have overclocked by then .