Overclocking Test On An Cortex A8 Processor

[mali]

I didn't check the datasheet in a while and don't have time right now. Is at OPP5 the highest voltage available already, or is there something higher? If it's the highest, the problem stands as above, if it's not the highest, it's a different thing.

 

[ari64]

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).

I think it's more of an issue with the power supply being unstable, which is a well-known problem with the beagleboard. Increasing the core voltage makes it worse.

 

[Dimacus]

The datasheet specifies:

	(min)		(nominal)	(max)
OPP5:	1.2825v		1,35v		1,4175v
OPP4: 	1,2065v		1,27v		1,3335v
OPP3:	1,14v		1,2v		1,26v
OPP2:	1,007v		1,06v		1,113v
OPP1:	0,93575v	0,985v		1,03425v

It's also interesting to note that the absolute maximum ratings table states a maximum voltage of 1.6v

[Speculation]
So unless I am horribly mistaken (and I have a feeling that I am), a voltage of, let's say, 1.5v should not instatly kill the soc, even if it might cause sever instability and shorten it's lifetime.
[/Speculation]

 

[Trip]

Interesting thread wuwtk but I hope you don't fry your Droid

 

[wuwtk]

Interesting thread wuwtk but I hope you don't fry your Droid

Well later on when i get home from work im gonna do a test for at least 5 hours and do an overall more thorough test. As for frying it I have overclocked every phone/pda Ive had and although im not using them for the long term use ,e.g. more than 5 years, i have never seen or heard anything overheating and frying the cpu. I have seen them run very unstable, but that is to be expected; actually i have never seen anything run as stable while doing this as android, I dont know if its the hardware or software that is letting this run this stable though.

 

[Yoyobuae]

The datasheet specifies:

	(min)		(nominal)	(max)
OPP5:	1.2825v		1,35v		1,4175v
OPP4: 	1,2065v		1,27v		1,3335v
OPP3:	1,14v		1,2v		1,26v
OPP2:	1,007v		1,06v		1,113v
OPP1:	0,93575v	0,985v		1,03425v

It's also interesting to note that the absolute maximum ratings table states a maximum voltage of 1.6v

[Speculation]
So unless I am horribly mistaken (and I have a feeling that I am), a voltage of, let's say, 1.5v should not instatly kill the soc, even if it might cause sever instability and shorten it's lifetime.
[/Speculation]

You do realize the meaning of "absolute maximum"? Reaching it probably means instant death of the OMAP. You really don't even want to get close to it. What if the power supply hiccups and temporarilly increases by 0.1V? With 1.35V there is more than enough margin of error, with 1.5V there's not.

And who knows what would be the long term effects of going so high. Maybe it doesn't insta-kills it, but it takes a week of use instead.

 

[Dimacus]

You do realize the meaning of "absolute maximum"? Reaching it probably means instant death of the OMAP

Yes, I know what that means and that is why I took 1.5v it as an extreme example, in showing that it's way outside the recommended voltage, but still won't cause instant death.

What if the power supply hiccups and temporarilly increases by 0.1V? With 1.35V there is more than enough margin of error, with 1.5V there's not.

I was not in any way trying to imply that it would be ok to do such a thing.
In fact, I was trying to point out that exact thing; the large difference between the recommended voltage of OPP5 and the absolute maximum.

I don't think the power management chip should/can (undamaged) fluctuate that much though, it would probably cause extreme instability, even under normal usage.

And who knows what would be the long term effects of going so high. Maybe it doesn't insta-kills it, but it takes a week of use instead.

Probably quiet severe, and it's possible it would not work at that voltage at all.

I was not in any way trying to suggest that running at 1.5v would be ok, but if that is the way it came out, I apologise, it was not what I ment at all.

 

[mali]

I wonder why these chips need such a high voltage. My su3500 runs at 1.7GHz 0.875V. Must be some kind of ARM/x86 architecture difference, I suppose.

 

[fischju2000]

A couple of the devboards have been run at OPP5, that's how they got to the 900mhz+ mark. They were stable as far as I remember. Overclocks around 700-800mhz on devboards seem to be on the default OPP3. OPP5 @ 1.4175v is a little close to the max, but it's within TI spec, and as far as I know you can't increase it more than that (at least in software).

 

[WizardStan]

A couple of the devboards have been run at OPP5, that's how they got to the 900mhz+ mark. They were stable as far as I remember. Overclocks around 700-800mhz on devboards seem to be on the default OPP3. OPP5 @ 1.4175v is a little close to the max, but it's within TI spec, and as far as I know you can't increase it more than that (at least in software).

The dev boards are also built with engineer samples, if I understand correctly. The engineer samples are likely to have been hand picked as the best of the best at the time.

 

[Exophase]

The dev boards are also built with engineer samples, if I understand correctly. The engineer samples are likely to have been hand picked as the best of the best at the time.

Why would they go through the trouble to bin engineering samples by clock speed?

 

[amortvigil]

so is it capable to clock 1ghz? without problems?

 

[kingoddball]

They said 900mhz at one stage - They are now saying "don't".
So.. I'd say no. I have never seen or heard anyone with a Pandora hitting 1ghz.

 

[WizardStan]

Why would they go through the trouble to bin engineering samples by clock speed?

Not necessarily by clock speed, but if you're going to send samples that you know are going to be abused to customers you are selling large quantities to, wouldn't you want them to be the most durable? Able to operate at the upper levels of heat and voltage? I could be off base, though I do think I recall MWeston saying something similar, that engineering samples tend to be more durable, so may be clocking higher than average. To be fair, I recall a lot of things, whether they happened or not.

 

[Tokiopop]

They said 900mhz at one stage - They are now saying "don't".
So.. I'd say no. I have never seen or heard anyone with a Pandora hitting 1ghz.

I think even 900MHz is a little too much, 800 would be plenty IMO.

But we really should be allowed to OC to some degree since when I paid for my Pandora, it was advertised (and still is, I believe) as 600+MHz.

If any OC voids the warranty, then that's false advertising.

 

[renwils]

I'll be overclocking regardless of the warranty, since it seems to be relatively safe.

 

[alucart1984]

I'll be overclocking regardless of the warranty, since it seems to be relatively safe.

What do you know your a Lunatic?! I will OC also once I get a Pandora.

 

[mali]

You are all cancelled.

 

[WizardStan]

I'll be overclocking regardless of the warranty, since it seems to be relatively safe.

In the event that it breaks, man up and admit you overclocked, at least.

 

[renwils]

I'll be overclocking regardless of the warranty, since it seems to be relatively safe.

In the event that it breaks, man up and admit you overclocked, at least.

Absolutely; I just hope that if there is some other hardware defect on the device that it will still be covered. Even if it isn't though, I'll take the risk.