Some Components of the N900 will lose much of their estimated Lifetime through overclocking. Everything gets hotter. This may not be the case if you overclock & undervolt at maximum efficiency. Maybe.

But anyway, the components loosing lifetime through overclocking still wont be the first to die, probably. Except for the Battery. Batteries dont like running hot and stuff.

Please don't say overclocking is safe. Just because your cpu still works doesn't mean OP is as lucky.
http://talk.maemo.org/showthread.php?t=68795
http://talk.maemo.org/showthread.php?t=71107

Running CPU @ 500MHz it lasts 100.000h, @ 600MHz it lasts ~20.000h (at nominal temperature!!) - that's why 600MHz sometimes is called "official OC".

Running CPU @800MHz may mean it gets permanently irrecoverably fried after as little as 2000h total of actually running at that clock speed (see my prev posters about dynamic clock speed and idle)

Generally OC is *strongly deprecated* unless you definitely know what's the deal - most posters before exposed a scary ignorance regarding that, stating only FUD and handwaving on a level of "WFM so it's safe"

Nokia definitely did invest lots of work to tune the device to run as fast as safely possible, while still ensuring a reasonable product lifespan. No smartass can outperform this on a "no clue why but it works" basis.

regards
Joerg (EE)

@joerg_rw - what is missing in your equation is voltage. So if you undervolt while overclock, thus keeping total energy (i.e. average power multiplied by time) in sane limits (near stock) you are safe. And your explanation may leave someone with the impression that overclocking at 800MHz means locking CPU at that speed - which you perfectly know is not true.

Re great Nokia job - recently i have played a lot with kernel code in both n900 and n950, and TBH I am under impression that Nokia just copy/cat what comes from TI. And what comes from TI is a set of parameters which are safe for TI evaluation boards, thus no clock/voltage optimizations specific to n900 exists in n900 kernel. What I totally agree on is that the job Nokia HW engineers have done to dissipate the heat from the SOC is absolutely great.

sorry I have to disagree, for two reasons: first your equation is missing the true nature of the problem which is electromigration and not voltage or power or energy. And second it's missing SmartReflex(R) set of measures in OMAP, which mostly defeats the only possible purpose of your undervolting which was to reduce current density, the true cause of electromigration. Higher clock rates cause steeper edges in level changes, cause faster charging of parasitary capacitors, cause higher current density surges, cause more electromigration.
Actually SmartReflex function blocks inside OMAP, which are taking care about virtually every single gate and transistor and adjust their individual working points (basically the quiescent current) accordingly to match the clock frequency might even cause your undervolting to result in a worse situation regarding electromigration. I for sure don't know enough of the particular details of what's going on inside the chip's gates to rule it out. If you actually know more... I'm listening.
So no, you're NOT safe.

Please don't first turn my words into their opposite to accuse me being a liar then! :-( What of >>2000h total of actually running at that clock speed (see my prev posters about dynamic clock speed and idle)<< makes you think I told something about CPU being permanently locked at max clockspeed?

Sure the parameters come from TI, as not even Nokia can afford the needed tests or had the mandatory insight in and knowledge about the chip's internals, nor the tools like electron microscopes etc to examine chips suffering EM after test runs. It's however a weird idea to think of those parameters as "optimized for TI eval-boards". And it's silly to think you could optimize those for N900, as it's just the chip and only the chip that's relevant here and that is determining those parameters. Any design particulars like heat dissipation of N900 are absolutely irrelevant for that, they only where relevant if the main problem of OC was overheating which it definitely is not.
All this shows again why there's so much nonsense around OC, everybody is starting with arbitrary random assumptions (like OC problem was heat) and then gets involved in sophistic developments and theories based on those false assumptions (here e.g. undervolting, maybe even dynamic, based on ambient temperature, eh?)

Sorry if the above maybe sounds a bit harsh, but it really annoys me since almost 2 years now, and all the info has been given over and over and over again, it's all there for everybody to read and understand. But no, OC is cool, and WFM, and of course every EE that tells something different is just a fool, no matter if it's Igor of Nokia, or me, or the guy writing the OMAP3430 datasheet. N900 community is so smart they know best, no doubt.

cheers
jOERG

Have been running mine on 850mhz for the last 18months and have widgets set up to jump to 1000mhz for emu's when I wish.

I have seen no ill effect at all.

*SIGH*
Didn't you get it that all that info of yours is mere rubbish, as long as you can't tell how many hours your CPU ACTUALLY SPENT IN 850Mhz, and how many hours in 1000MHz?
It's exactly that type of statement that makes everybody fail to understand what OC is *really* all about. The deal is NOT "18 months with no issues".

But don't worry, Texas Instruments has already evaluated those values that really matter:
100kh@500MHz
23kh@600MHz
will result in no more than 50 of 1000 CPUs dead after that time.
For the normal CPU, the MIL CPU is a *bit* worse:
50kh@500MHz
10kh@600MHz will result in 70 of 1000 dead

you know what's an exponential graph? look at 50@500, 10@600, and tell me what you guess will be a reasonable value to expect for lifespan at 800!
Your CPU won't age when device is sitting there idle for 18 months, it *will* during you watching youtube videos though.
http://mg.pov.lt/maemo-irclog/%23mae...03-11T03:04:11


/j

so it is dangerous sir?

yes, exactly like smoking is dangerous to your health and reduces your statistic lifespan.
Just with smoking you can stop and 3 or 4 years later your statistic lifespan is back to average of non-smokers.
Stopping OC won't heal the CPU fro EM damage done during OC. And OC to 1000MHz is more like smoking crack than like smoking cigarettes.

/j

i am using batterypatch and changing the default as default but on idle i think its underclock...