Frying An Arm Chip

[craigix]

Does anyone have any evidence that this can be done on the current generation? That is frying an ARM chip via overclocking?

I don't mean 'he said she said' stories, I mean some solid proof of it being achieved before the chip crashed.

Let's start here before deciding the overclocking policy.

 

[Tokiopop]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

Edit: ED seems pretty sure OC wouldn't do much (any) harm, but OV would.

 

[mcobit]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

+1

Edit: Or do it the Appleway:
http://support.apple.com/kb/HT2101

 

[craigix]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

 

[Rathum]

Make sure you're clear that you're not talking about overvolting it. Overvolting will definitely kill the chip :lol:

 

[atomicthumbs]

We should ask around on Beagleboard and N900 forums

 

[Mutilator]

Just out of interest Craig, how many gp2xs have you had returned which have been fryed by overclocking? (I know you send them to GPH to test why they failed)

 

[Farox]

We could also ask to TI ...the producer of the Chip maybe have some more experience on it ...no ?

 

[craigix]

Just out of interest Craig, how many gp2xs have you had returned which have been fryed by overclocking? (I know you send them to GPH to test why they failed)

I'm not aware of any, I don't think the chip can get hot enough. I guess if you started to fuck about with the voltage by hand (ie doing wacky hardware mods) you could kill it.

That said it was a pretty common GP32 mod.

 

[mali]

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

It was a first generation xScale, so old stuff compared to Cortex. You could either make a stress test by going to the highest possible stable voltage+clock combination and let it run till it breaks(or not) or just lock the voltage(or just don't expose it to the user, like it is now already) and let the user overclock without overvolting. This should be save enough to be covered under warranty, I suppose. And this would also be totally in 'community spirit'.

 

[Jumpman]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

I doubt you get any evidence, it is all he said she said! Like 90% of what's said out in the world! I think you would have just as much chance of burning out a DS or PSP, by running pirated software on them, it's just not going to happen! My 2 cents! B)


Chris

 

[dpoarch]

Craig said that it doesn't give off much more heat on an overclock. The only way you are going to fry it is if you get past its operating temperature specified by TI. Really the only way you are going to fry it is if you try to bump up the voltage to the core so it will run more stable at higher clock speeds. Since this is a micro I believe the internal voltages are somewhat self controlled so it may not even be possible to fry it without doing some kind of hardware hack to adjust the voltage. I think that overclocking this chip will just make it become unstable at some point and then it will crash. Running at a higher speed may reduce the life of processor by making the transistors switch faster/more often but without increasing the heat significantly I don't see it breaking.

edit: How hot is this thing getting? I mean if its under 70 degrees Celsius, the logic is under 1.35v, and the core is under 1.15v I think it should be fine. MWeston would probably know a lot more about this than any of us since he built it.

page 123
http://focus.ti.com/lit/ds/symlink/omap3530.pdf

 

[craigix]

We could also ask to TI ...the producer of the Chip maybe have some more experience on it ...no ?

I don't just mean the OMAP, I mean any ARM chip killed via software via frying.

We need to know if this can happen.

People might remember a similar thread years and years ago about the GP32.

 

[craigix]

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

It was a first generation xScale, so old stuff compared to Cortex. You could either make a stress test by going to the highest possible stable voltage+clock combination and let it run till it breaks(or not) or just lock the voltage(or just don't expose it to the user, like it is now already) and let the user overclock without overvolting. This should be save enough to be covered under warranty, I suppose. And this would also be totally in 'community spirit'.

Can you give exact details, we want to buy this chip/device, clock it and measure the temp.

What did you do to kill it exactly? This seems to be an area we need to document. ARM overclocking.

 

[aq21]

I thought you made 105 boards for testing. Are there not a few of them running since then at the maximum frequency they can sustain?

 

[craigix]

I thought you made 105 boards for testing. Are there not a few of them running since then at the maximum frequency they can sustain?

Yes, it gets hot but not much hotter, certainly not hot enough to fry it. Crash it yes.

Until yesterday I'd never heard of an ARM chip getting hot enough to die. I think a lot of people on this board thought the same too - but maybe the new generation can?

 

[Poem58]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

Craig, I think it may have been misinterpreted. If I understood (Mail's?) comment about the killing of the Arm chip, I thought it was a PDA one like in my Dell Axim, you know the old Xscale CPU's. Granted I haven't gone back to that thread to check. I think those old ones could be fried basically because they could be overclocked but I don't think it was ever intended for any reason. I've read about people killing their PDA cpu's all the time pushing them too far, which is why I never tried, I actually liked my PDA! I know the software was always at your own risk and I think they overvolted too not sure, and they all had warnings that were very dangerous to try. Some people fried theirs in one attempt. Which you already know hasn't happened on a Pandora yet.I've never heard anyone ever say a cortex was fried. I think if we don't find anything online to prove it then you'd have to do a test. Which is best for your interests anyway as it would be a real world test in a real situation.

Quick google showed this on first page
http://blog.7touchgroup.com/tag/overclocking-the-arm-cortex-a8-a-550mhz/ no mention of it being fired

 

[MDave]

Isn't xScale an Intel design? Might explain a lot

 

[craigix]

I think we need someone to actually get a Pandora and overclock it and just leave it doing something and see what happens after x amount of time until it either crashes or breaks.

Nothing happens, the temp does not rise much. But the claims on the other thread about people managing it mean it must be re-investigated. I want some evidence.

Craig, I think it may have been misinterpreted. If I understood mail's comment about the killing of the Arm chip, I thought it was a PDA one like in my Dell Axim, you know the old Xscale CPU's. Granted I haven't gone back to that thread to check. I think those old ones could be fried basically because they could be overclocked but I don't think it was ever intended for any reason. I've read about people killing their PDA cpu's all the time pushing them too far, which is why I never tried, I actually liked my PDA! I know the software was always at you own risk, and they all had warnings that were very dangerous to try. Some people fried theirs in one attempt. Which you already know hasn't happened on a Pandora yet.I've never heard anyone ever say a cortex was fried. I think if we don't find anything online to prove it then you'd have to do a test. Which is best for your interests anyway as it would be a real world test in a real situation.

Does it fry it by overheating it? Or is something else happening? Can you give me a scale of the overclock which kills it? I'd like to measure how hot the chip gets.

 

[Butterman]

Make sure you're clear that you're not talking about overvolting it. Overvolting will definitely kill the chip :lol:

Proof? You can't prove that a reasonable overvolt does squat. It generally will speed up electromigration and cause the chip to heat up faster. The exact effect varies between chips, but taking it too far™ can exceed the voltage specification for certain components on the die or even components off the die.

We could also ask to TI ...the producer of the Chip maybe have some more experience on it ...no ?

This is a good idea, but remember some chip makers are cautious about overclocking, it allows you to save money by just OC'ing their lower binned chips.

Craigix: We've got the facts and evidence. Pickle posted it in the last thread. It's a beautiful architecture and allowing overclocking in the warranty is generally considered a feature. If the chips do fail, (from this evidence) we can almost certainly conclude the chip was faulty and was going to inevitably fail. The fact that it failed before the warranty is good for both parties.