Overclocking Test On An Cortex A8 Processor

[sold]

In regards to the engineering samples, why would they be any different? I thought a sample is just that. A small amount of the very thing you are buying. Why would they make em any different? If the samples get abused during testing and development, it may reduce the lifespan of the hardware, but 50000 POH is like, 5+ years, so you wouldn't notice any negative effects immediately. Unless you damage/kill it somehow, which I guess would be immediately apparent and junk...

 

[gahndalf]

In regards to the engineering samples, why would they be any different? I thought a sample is just that. A small amount of the very thing you are buying. Why would they make em any different? If the samples get abused during testing and development, it may reduce the lifespan of the hardware, but 50000 POH is like, 5+ years, so you wouldn't notice any negative effects immediately. Unless you damage/kill it somehow, which I guess would be immediately apparent and junk...

engineering samples are small batches made to exacting standards. not "mass produced". Mass produced chips are with a little less care and attention. just running tests too make sure they are still in spec. so engineering samples are usually a little better quality.

 

[Alpha2]

All this heart ache and pain and people trying to burn out poor little devices. Why cant the warranty simply say:
"As an open and freedom based platform the Pandora is capable of overclocking beyond it's rated clockspeed via software. Openpandora is not liable for any damage due to Overclocking, You do so at your own risk."
And the OS warns you if you switch the speed.

This way the people scared to do it simply wont, those that do can't complain if gets busted and they have to buy and new one. Those that bust their's will simply be more careful with the next one they have to buy to replace it. Back in the GP32 days you knew there might be a risk of some sort that it'd break your system and that it was unlikely that you'd be able to get one replaced if it broke because you tried to do it. So you either took the risk and exhaled heavily if it worked or just didn't do it.

 

[WizardStan]

All this heart ache and pain and people trying to burn out poor little devices. Why cant the warranty simply say:
"As an open and freedom based platform the Pandora is capable of overclocking beyond it's rated clockspeed via software. Openpandora is not liable for any damage due to Overclocking, You do so at your own risk."
And the OS warns you if you switch the speed.

1) They're investigating it. If overclocking doesn't actually do any damage, why shouldn't they replace a faulty chip that broke for other reasons? Overclocking could also damage things unrelated to the OMAP in ways we might not be considering that they have to take into account. I'm sure that they'll come up with something perfectly fair that protects both the user from faulty parts and them from user error.
2) As has been brought to attention, 600 mhz is only guaranteed under OPP5. Running at 600 mhz in OPP3 is technically overclocking. If it doesn't break warranty to set at a clock speed in OPP5, does it break warranty to have it at the same clock speed, even if it is perfectly stable as far as anyone can tell (right up until some unfortunate failure) in OPP3 which is a lower voltage, because it is technically overclocked then?
3999 people put their Pandoras into OPP5 and start running them at 900 mhz and have no problems. One person doesn't want to risk overvolting it so he keeps it at 600 mhz in OPP3, but as luck would have it, he got the one faulty OMAP that dies after only a 10 hours. Technically an overclock, so warranty is void. Pity.

 

[Alpha2]

So wait then, that means the Pandora is OCing no matter what you do? (assuming the power saver isnt powering down certain needed parts when they're not being used.)

 

[Mofokubik]

I had a long talk last night with a former TI engineer. I asked him about overclocking and there wasn't much he told me that wasn't already discussed on these boards somewhere. From what I understand, you can generally push the cpu 30% past it's limit as long as you keep it cool enough. Not all cpus may have a die capable of handling 800mhz, this very limit of 600mhz is to have a consistent product since maybe 1 out of 500 cpus may fail in the process of overclocking. So if I understand correctly, they set the speed limit based on the overall quality of the product, so that it can be reliable enough to manufacture 1000s at a time and still back them up with a warranty. Going past this limit may not damage your product, but if it does it's your fault.

Also, i'm not sure if this is useful, but I found it interesting none-the-less:

http://focus.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=sprz172f&fileType=pdf

under "important notice"

TI warrants performance of it's hardware products to the specifications applicable at the time of sale in accordance with TI's standard warranty. Testing and other quality control techiques are used to the extent TI deems necessary to support this warranty. Except where mandated by government requirements, testing of all parameters of each products is not necessarily performed. TI assumes no liability for applications assistance or customer product design. Customers are responsible for their products and applications using TI components... etc..

 

[Patrick R Ludvigsen]

Well, the thing was talked about very much like everyone would get to 900mhz easily, etc :\

It'd be pretty cool to wait a year and a half after ordering, finally get your Pandora, watch everyone overclock to play Dreamcast, follow along, and be one of the unlucky 1/500 "OOPS, SORRY YOU LOSE"

 

[darkblu]

wuwtk, nice initiative there. just a remark, though: until you know how & from where the temperature is reported on the droid, don't try any bold tests.

the t ranges given by TI in the spec sheet are all measured 'at junction' - that means by this analytical procedure - your droid's t report could be coming from arbitrary far from anything you'd call a junction in the SoC. for all you know the droid readings may show 50C when the t@ junction has already reached the critical 90C. so be careful there.

actually, while typing this post, i found this interesting post from the droid boards which practically confirms my warning to you.

 

[wuwtk]

Well I was testing it a little more thorough and this time i had it charging while i did this because the battery would of probably died without it. I had Quake 3 running with multiple bots and i tested to see the temperature every hour and this is all at 800mhz.

Hour 1 it was at 49C
Hour 2 it was at 51C
Hour 3 it was at 55C

I was planning on doing it for 5 hours but at 55C it was very hot to the touch and that started to worry me so i decided to stop it. Oddly enough it still ran perfectly fine and stable. It seemed much hotter while it was charging so im assuming there was a lot more at work than just the processor to generate that heat. And about 2 minutes after stopping the game and overclock it was back down to 35C. Dont exactly know if any of this truly helps but Ive never seen any handheld get this hot before.

 

[mali]

^Like I said(in this or a different thread), Looping a Quake 3 timedemo, while streaming something(best would be something that can be decoded via DSP) from USB HDD with a flat battery that is charging, at a relatively high ambient temperature, should give a proper worst case scenario to test a Pandora.

 

[Awakening]

Well I was testing it a little more thorough and this time i had it charging while i did this because the battery would of probably died without it. I had Quake 3 running with multiple bots and i tested to see the temperature every hour and this is all at 800mhz.

Hour 1 it was at 49C
Hour 2 it was at 51C
Hour 3 it was at 55C

I was planning on doing it for 5 hours but at 55C it was very hot to the touch and that started to worry me so i decided to stop it. Oddly enough it still ran perfectly fine and stable. It seemed much hotter while it was charging so im assuming there was a lot more at work than just the processor to generate that heat. And about 2 minutes after stopping the game and overclock it was back down to 35C. Dont exactly know if any of this truly helps but Ive never seen any handheld get this hot before.

That is kind of scary, I would have aborted too at that point. I think the jump from 51 to 55 is interesting. I guess this is because the inside of the Droid is getting warm enough that the cooling is not working properly, the battery temperature is probably constantly rising giving a dual effect. The Pandora's case is larger so it should be less of a problem though. It would be interesting if you could do the exact same 3h test test on say 600 MHz as a comparison.

This of course does not give us an answer to the question: Will it be possible to fry the CPU or otherwise damage your Pandora from overclocking?

My old laptop gets really hot in one spot from playing 3D games even with a cooling pad under it that it actually shuts itself off. It's not overclocked at all. It's not something I would want to hold in my hands but it seems perfectly safe for the computer, except that I bet it slowly damages it over time.

 

[MrPoletski]

Voltage adjustment?

Does the pandora include the ability to change the supply voltage to the processor? (and GFX chip?)

If so, that will mean a lot more to overclocking. If people want to go really nuts then I don't know about the chip positioning, but installing your own phat heatsink (would have to be seriously custom) would definately help results. Basically, is the omap chip on the bottom side of the board, close to the bottom face of the case? if it is, great, cut a hole and add a sink!

 

[mali]

^Battery is in the way.

 

[Yoyobuae]

Even if the processor can tolerate overclocking and the temperatures that come with it, you're still frying your battery. :rolleyes:

 

[SirisC]

^Battery is in the way.

Maybe a flat copper plate to act as a heat spreader.

 

[mali]

^Actually there must be a post floating around from 2008, where I suggested to design a copper heatspreader and connect it to the metal cages of the SD Card slots

 

[droneb]

^Actually there must be a post floating around from 2008, where I suggested to design a copper heatspreader and connect it to the metal cages of the SD Card slots

yep i remember it, its something similar i did with my 600m optical drive slot. (at much more larger scale XD)
but instead of the flat copper i was thinking on corrugated near the front and rear spots and if so maybe modding/drilling the case for more air flow.

you need some to get air in and some to get air out.

it was also disscused on the plugin mod. to get a socket that would be locked on the battery port and battery adjusted over the add on.
so even more airflow available

 

[wuwtk]

Well I was testing it a little more thorough and this time i had it charging while i did this because the battery would of probably died without it. I had Quake 3 running with multiple bots and i tested to see the temperature every hour and this is all at 800mhz.

Hour 1 it was at 49C
Hour 2 it was at 51C
Hour 3 it was at 55C

I was planning on doing it for 5 hours but at 55C it was very hot to the touch and that started to worry me so i decided to stop it. Oddly enough it still ran perfectly fine and stable. It seemed much hotter while it was charging so im assuming there was a lot more at work than just the processor to generate that heat. And about 2 minutes after stopping the game and overclock it was back down to 35C. Dont exactly know if any of this truly helps but Ive never seen any handheld get this hot before.

That is kind of scary, I would have aborted too at that point. I think the jump from 51 to 55 is interesting. I guess this is because the inside of the Droid is getting warm enough that the cooling is not working properly, the battery temperature is probably constantly rising giving a dual effect. The Pandora's case is larger so it should be less of a problem though. It would be interesting if you could do the exact same 3h test test on say 600 MHz as a comparison.

This of course does not give us an answer to the question: Will it be possible to fry the CPU or otherwise damage your Pandora from overclocking?

My old laptop gets really hot in one spot from playing 3D games even with a cooling pad under it that it actually shuts itself off. It's not overclocked at all. It's not something I would want to hold in my hands but it seems perfectly safe for the computer, except that I bet it slowly damages it over time.

The Droids casing is extremely small considering that it is a slider so that could have a big influence. And as for the older laptop; x86 processors, especially older Intel ones, were idle at 50C easily. I just thought this was interesting because I never saw and arm processor get this hot. It was to my understanding arm processors shouldn't of gotten that hot. And if somebody is playing a heavy stressing game on the Pandora with it plugged in and overclocked to 800mhz+ i honestly now think there is a possibility of doing some kind of damage. I don't have any real backing and I do understand that they should crash before they get to hot but that is just my two cents.

 

[Awakening]

The Droids casing is extremely small considering that it is a slider so that could have a big influence. And as for the older laptop; x86 processors, especially older Intel ones, were idle at 50C easily. I just thought this was interesting because I never saw and arm processor get this hot. It was to my understanding arm processors shouldn't of gotten that hot. And if somebody is playing a heavy stressing game on the Pandora with it plugged in and overclocked to 800mhz+ i honestly now think there is a possibility of doing some kind of damage. I don't have any real backing and I do understand that they should crash before they get to hot but that is just my two cents.

Well, if you want to run the Pandora overclocked and plugged in you could remove the battery, and even leave the battery compartment open for possibly some extra cooling.

 

[SNESFAN]

The Droids casing is extremely small considering that it is a slider so that could have a big influence. And as for the older laptop; x86 processors, especially older Intel ones, were idle at 50C easily. I just thought this was interesting because I never saw and arm processor get this hot. It was to my understanding arm processors shouldn't of gotten that hot. And if somebody is playing a heavy stressing game on the Pandora with it plugged in and overclocked to 800mhz+ i honestly now think there is a possibility of doing some kind of damage. I don't have any real backing and I do understand that they should crash before they get to hot but that is just my two cents.

I'd almost be willing to say you would blow up your battery before you blow up your proc

http://en.wikipedia.org/wiki/Lithium-ion_battery#Safety