Overclocking Test On An Cortex A8 Processor

[wuwtk]

Since the other threads are becoming more offtopic and cumbersome ive decided to start a new thread for this.
So I decided to do some test on my Droid as for heating and the processor. I know that it uses a 600mhz cortex a8 processor underclocked to 550mhz, which I believe is extremely similar to the Pandora. So I have my phone rooted and using setcpu app for overclocking and tempmonitor to keep an eye on the temperatures ive decided to see how hot the cpu can get so at least we can get a basic idea of what overclocking can do to the processor. Recently Quake 3 was ported to the Droid so i picked this as my test. And also sorry if this does not help at all when it comes to the overclocking issue in any way. Just doing anything i can do to at least try and help.

At normal temperatures at 600mhz it was at steady 31C.
After playing Quake 3 for about 10 minutes it at 36C.

Overclocked to 800mhz it was hovering around 31-32C idle.
Overclocked to 800mhz while playing Quake 3 for about 10 mins was at 37C.

Overclocked to 1000mhz it was at 33-35C idle.
Overclocked to 1000mhz while playing quake 3 it was at 40C.

At 1000mhz everything ran perfectly fine with no crashing, although only running for about 10 mins, but the keyboard and the back started to become very warm.

I know that this is a very informal test and may not help this subject at all, but i figured id try. If anybody wants me to test anything else for longer periods of time or different things in general, please tell me and i will try within reason.

 

[Sphinxter]

I salute you and your noble sacrifice.

 

[Username]

I salute you and your noble sacrifice.

Agreed.

 

[rabidpoobear]

I salute you and your noble sacrifice.

Agreed.

Definitely. Would like to see some long-term numbers though. Eg. 800mhz for 5 hrs or something, as that's probably what's gonna happen (esp. if I have to OC to 800mhz to get ocarina of time to play smoothly ) Not tellin' you to do it obviously, just askin'.

 

[wuwtk]

I salute you and your noble sacrifice.

Agreed.

Definitely. Would like to see some long-term numbers though. Eg. 800mhz for 5 hrs or something, as that's probably what's gonna happen (esp. if I have to OC to 800mhz to get ocarina of time to play smoothly ) Not tellin' you to do it obviously, just askin'.

I am currently running a match with bots killing me and running around shooting each other at 800mhz and im gonna leave it up for at least 2 hours to see what that does.

 

[socket]

Sweet. Thanks!

 

[wuwtk]

At 800mhz playing quake 3 for 2 hours my cpu went to about 45c. The keyboard and back got very warm but nothing of what felt dangerous for the cpu itself. And now im even more excited for the Pandora after trying to play quake on a touch screen...

 

[bustaballs]

Maybe after some tests there could be specific warranty statements like:

"Overclocking past xxx Mhz will void the warranty. Overclocking the voltage at all will void the warranty. This warranty is subject to change at any point in time by the discretion of OpenPandora Ltd."

 

[rabidpoobear]

At 800mhz playing quake 3 my cpu went to about 45c. The keyboard and back got very warm but nothing of what felt dangerous for the cpu itself. And now im even more excited for the Pandora after trying to play quake on a touch screen...

Awesome, 45c is not even close to the max operating temp of the A8. Let's hope the Pandora's not worse at dissipating heat than the Droid.

 

[grydholt]

Maybe after some tests there could be specific warranty statements like:

"Overclocking past xxx Mhz will void the warranty. Overclocking the voltage at all will void the warranty. This warranty is subject to change at any point in time by the discretion of OpenPandora Ltd."

IANAL, but I do not think you are allowed to change a warranty in the E.U., and I would assume it is the same for the rest of the civilised world.

 

[ari64]

My beagleboard will not reliably run at more than 750MHz. I didn't try to change the voltage. (The TPS65950 outputs 1.2 volts by default.)

I wonder what voltage the Droid CPU is at.

 

[wuwtk]

My beagleboard will not reliably run at more than 750MHz. I didn't try to change the voltage. (The TPS65950 outputs 1.2 volts by default.)

I wonder what voltage the Droid CPU is at.

I tried to look it up but i couldn't find out what the voltage is at. And today i had it all day at 1000mhz and everything was stable except for one widget that crashed twice. Playing random games for 10-15 mins at a time and web browsing was no problem.

 

[kingoddball]

Thats some good news.
Odd about the beagle board though. It's case free design should be better for not holding heat. I guess its dependent on the individual chip (Which HAS been said all along - I know).

 

[Awakening]

All this test shows is that the temp increases over time and that it increases faster with a higher clock. For this test to be useful you need to do a longer test where you measure the temperature more often. For example 5 hours where you measure after each hour. If the temp is still rising you should consider running a longer test. Right now i think 800 MHz is the most interesting speed to test.

 

[ldesnogu]

At last some interesting numbers, thanks

Recently Quake 3 was ported to the Droid so i picked this as my test. And also sorry if this does not help at all when it comes to the overclocking issue in any way. Just doing anything i can do to at least try and help.

Can you say what parts of the CPU/SoC this port of Q3 uses?

- NEON? (I guess it does)
- SGX? (it surely does)
- DSP? (I guess it doesn't)

If I'm right, the only missing thing is the DSP. But your test is definitely better than anything I've seen (at least anything publicly available) at stressing OMAP3.

 

[wuwtk]

All this test shows is that the temp increases over time and that it increases faster with a higher clock. For this test to be useful you need to do a longer test where you measure the temperature more often. For example 5 hours where you measure after each hour. If the temp is still rising you should consider running a longer test. Right now i think 800 MHz is the most interesting speed to test.

Well tomorrow when i should have some free time ill be able to do a test like that. ^_^

At last some interesting numbers, thanks

Recently Quake 3 was ported to the Droid so i picked this as my test. And also sorry if this does not help at all when it comes to the overclocking issue in any way. Just doing anything i can do to at least try and help.

Can you say what parts of the CPU/SoC this port of Q3 uses?

- NEON? (I guess it does)
- SGX? (it surely does)
- DSP? (I guess it doesn't)

If I'm right, the only missing thing is the DSP. But your test is definitely better than anything I've seen (at least anything publicly available) at stressing OMAP3.

Well reading from the project website http://code.google.com/p/kwaak3/ it says its using Neon. Beyond that i truly don't know enough about this, i merely just picked this because it stressed my phone more than anything else Ive tried.

 

[cosurgi]

Maybe after some tests there could be specific warranty statements like:

"Overclocking past xxx Mhz will void the warranty. Overclocking the voltage at all will void the warranty. This warranty is subject to change at any point in time by the discretion of OpenPandora Ltd."

I like that - great idea!

Perhaps they cannot change the warranty - for any given, sold unit. But then can change the warranty for next units being sold. You buy the unit, you get the warrany with it.

So if they have too much returns, they change the XXX for units sold next year. Or in the opposite direction.

 

[mali]

^ Exophase explained the problem of that approach above(edit: in a differnt thread). OPP3 is guaranteed to work for 500MHz and the advertised 600MHz _might_ not work at OPP3. TI says you should use OPP5 at 600MHz, which is considered overvolting already, as I understand it.

 

[Awakening]

All this test shows is that the temp increases over time and that it increases faster with a higher clock. For this test to be useful you need to do a longer test where you measure the temperature more often. For example 5 hours where you measure after each hour. If the temp is still rising you should consider running a longer test. Right now i think 800 MHz is the most interesting speed to test.

Well tomorrow when i should have some free time ill be able to do a test like that. ^_^

I realized I did sound a bit ungrateful in my post. Thanks for spending your time and risking your hardware for these tests

 

[Dimacus]

^ Exophase explained the problem of that approach above(edit: in a differnt thread). OPP3 is guaranteed to work for 500MHz and the advertised 600MHz _might_ not work at OPP3. TI says you should use OPP5 at 600MHz, which is considered overvolting already, as I understand it.

But should it be considered overvolting?

The way I interpret the datasheet (and I could very well be totally wrong), the different modes are not really related to each other (i.e. OPP5 is Not OPP3 on steroids), they are more like the gears of a car and unless you throw some nitro
(change the clock or voltage beyond the specified nominal values for each mode) into the fuel they are all working within their specified parameters.

From the warranty question perspective, if TI guarantees that the soc will function 50k POH on continues OPP5 mode, shouldn't that mean that if a device fails before the 50k POH, OP has the right to some sort of compensation?