Frying An Arm Chip

[Magic Sam]

The default speed is 600MHz.

AFAIK, chips was 600 MHz "certified" (or class), but default speed will be set to 500 MHz. Sorry, I have read it more that once, but I haven't note any source.

Thanks guys for your quick replies
Now another question: what's the difference between "our" chips and the 720MHz ones TI is talking about ?
Bye and thanks again, Magic Sam

 

[mali]

^We don't know exactly if there is a physical difference. They might be binnned. The 720MHz version is more expensive, so the 2nd batch will also contain 600MHz chips.

 

[Magic Sam]

^We don't know exactly if there is a physical difference. They might be binnned. The 720MHz version is more expensive, so the 2nd batch will also contain 600MHz chips.

Thanks mali
Do you mean that "our" chips equal in quality the 720MHz chips, though they are not being certified to run at such speed by TI ?
Bye, Magic Sam
p.s: you're a great guy

 

[mali]

^Yes, that is possible. The 720MHz chips didn't exist at the time the first batch chips were bought. And thanks

 

[Magic Sam]

^you're welcome And your avatars always look HOT ! :wub: Who's that girl ?
I hope my girlfriend will never read that post
Bye, Magic Sam

 

[mali]

^It's a painting, so it's save

 

[kuru]

The only thing I ever overclocked was an old graphics card, I think an nvidia 4800 ti. The only thing it would do is display errors on too-high settings. There shouldn't be any major problems as long as your heat dissipation is fine and the chip isn't prone to failure by design.

If my Pandora can reach high mHz figures I'll happily overclock it whenever necessary. Five years of continuous use doesn't sound bad to me, since you probably won't run it OC'ed all the time. Furthermore I don't doubt the goodwill of the shops when it comes to warranty. I bet it'll all work out in the end. Hugs all over!

^It's a painting, so it's save

HAH! Women can and eventually will be jealous of virtually ANYTHING.

 

[Grench]

You don't have to actually fry a Pandora to see what conditions are within the processor's specs and what are not.

http://www.gp32x.de/board/index.php?/topic/52157-frying-an-arm-chip/page__view__findpost__p__818592

So, is there a temperature sensor in the RAM stack on top of the CPU or not? Is there one anywhere inside the case?

Examining the temperature curve of the CPU inside a MP case to see how hot it gets at what temperature - thats really the key to this isn't it? Overclocking to 800Mhz in an office/home environment without the battery in a charge cycle under it - sure. What about when it is 120*F (49*C) AND the battery is in an active charge cycle AND the CPU is overclocked? Heat kills. If OpenPandora is going to endorse (warranty) overclocking, they will need to understand the heat dissipation rate relative to the external environment of an OpenPandora console in it's highest amperage draw state (charging + pegged CPU).

Also, it is becoming clear that what I detailed as a "Jealousy RMA" is a real possibility. I.e. people considering their Pandora consoles to be defective if they can't overclock to 800Mhz. I think it should be made clear that although overclocking may be allowed, it is not guaranteed. A few of them have even voiced this opinion in this thread - i.e. requiring that they be able to run at 800Mhz just like the devs.

 

[Magic Sam]

Also, it is becoming clear that what I detailed as a "Jealousy RMA" is a real possibility. I.e. people considering their Pandora consoles to be defective if they can't overclock to 800Mhz. I think it should be made clear that although overclocking may be allowed, it is not guaranteed. A few of them have even voiced this opinion in this thread - i.e. requiring that they be able to run at 800Mhz just like the devs.

Let's have a look at the GP2X: mine overclocks and is stable @ 260MHz max (when using only one core), I know some GP2X will overclock more, some will overclock less...this is pure luck, and it is not going to change with the Pandora !
@ Craig: IIRC you advertised the Pandora as being able to reach speeds higher than 600MHz in the past, in my mind it was like a feature, one of the many reasons I pre-ordered for, and I would be terribly disappointed if you decided afterwards that "regular" overclocking (not pushing the Pandora to 2GHz, overvolting it, then crying like a baby because it no longer works) voids the warranty
Though I use to disagree more and more with you these days, thank you anyway for this fabulous project, all your hard work, and the money you invested from your own pocket
Bye, Magic Sam

 

[mali]

Also, it is becoming clear that what I detailed as a "Jealousy RMA" is a real possibility. I.e. people considering their Pandora consoles to be defective if they can't overclock to 800Mhz. I think it should be made clear that although overclocking may be allowed, it is not guaranteed. A few of them have even voiced this opinion in this thread - i.e. requiring that they be able to run at 800Mhz just like the devs.

MWeston had become pretty nasty once, because people _assumed_ that it's guarateed to be able to overclock to 900MHz. Since then it has always been a possibility and no guaranteed feature.

 

[chaosfx]

I think the only way to test it would be :

Take a pandora overclock it to 900(?)Mhz and let it run 24/7 (maybe 2 or 3 pandoras).

Noone out there will use it that way so these pandoras should fry long before customers pandoras could die.

I'm not shure how long the Coretx A8 is used but i don't believe there are any longtimeoverclocked devices out there for.

Most people i know overclocking x86 cpus don't use them for more than 2 years before getting a new one so even there it's very hard to get exact numbers.

Only TI could know anything about frying the Cortex
I'm shure they have some units permanent overclocked so they can rise the Chiprating after operating for example the 100000 hours stated.

P.S. I know my english isn't the best but i hope you understand my point.

 

[mali]

Like Laurent and others said, you have to see a device as a whole. The CPU is just one component sensitive to heat. There are many more heat producers in the Pandora, which are encapsuled inside a case sharing the same little bit of space. There is no active ventilation, so it adds up, eventually. If the dissipation through the case is even more reduced because of a high ambient temperature, it gets even worse.
The only way to test this is by experimentation. Even TI couldn't help, because Pandora is unique.

 

[puppydee]

Over clocking wont kill these chips within a year so I don't know what all the fuss is about. I'm not really fussed either way about the warranty I don't see me having any problems with my pandora (I HAVE FAITH IN MWESTONS DESIGN) I'm more worried about getting the cases finished than the electronics although I'm sure the chinese will pull it off. I just hope it's sooner rather than later.

How many dev boards have died due to overclocking? Non? See

 

[ldesnogu]

Only TI could know anything about frying the Cortex
I'm shure they have some units permanent overclocked so they can rise the Chiprating after operating for example the 100000 hours stated.

I will let you compute how many years 100k hours represent. Then you'll take a look at when OMAP3 was announced

All this noise around overclocking makes me feel very uneasy. I have been saying for a very long time people here shouldn't take for granted anything about it, and that talking about the 900 MHz was unwise. It's so bad it deteriorated the way it did

 

[Awakening]

I don't think anyone here will care about a drop from 100'000 hours to 44'000 hours. I doubt anyone will run it for even 3'000 hours per year and that won't be at full speed all the time anyway. Most users probably want to use OPP5 when they need the extra power.

I wonder what kinds of tests really are interesting. If you max out a CPU running at 900 MHz for 10 hours running inside a Pandora case and compare the temp inside the case and the CPU itself to one running at 600 MHz you should be able to estimate what kind of effect that will have. The room temp should probably be at least 30C (maybe more) for both tests to be accurate. If the difference is small you should be certain that OC/OV does not increase the temperature enough to cause a problem.

 

[darkblu]

i hate to sound like somebody who has to bring common sense to a thread, but:

the clocking of VLSIs has been about realiability. or IOW, your bits meeting expectations, doing exactly what they're support to do while traversing the circuit. i grin each time somebody comes up with 'what's the worst that can happen when i up the clock (but leave the voltages) - a crash at most' - i don't know about you, but i don't look at random crashes lightly. and as a software guy i would direct any stability reports about my software, coming from overclocked systems, rigth down to /dev/null.

btw, i'm a software guy, but i happen to work at a company that makes embedded devices (custom PCBs, custom enclosures, etc). a lot of time and effort is spent here to ensure TDPs are rigidly met, and the consumer can run the device the advertised way (i.e. 24/7, in unfavorable ambient t environments).

i know the prevailing mentality on these boards is 'hey, as long as i can run emu X at fps Y for limited time before it crashes, i'm good', but i don't share that mentality - for me the pandora is a linuxbox that will be treated as any other - it will have to be able to run tasks reliablty for the whole duration of the battery charge. and then again on the next day, and on the next, etc, while travelling with me to locations of diverse t conditions. if TI say that OPP3 goes as high as 600MHz (reliably) for pandora's omap, and that's the recommended OPP for everyday usage and normal lifespan, then that's what my pandora will be run at, and what my sw will be targeted at.

on the whole frying the cpu matter:

totally frying a chip is not the worst that can happen to a system. back in the 486 days a pc here could not compile its kernel anymore after its cpu fan got stuck and that fact was noticed only later after it started crashing repeatedly. the fan was fixed, but that pc could never again run a successful kernel build. otherwise everything was seemengly ok with it from the outside. but i'd hate to be the person dealing with the potential 'bug reports' from that site. in this regard, i never understoof OP's liberal attitude toward overclocking - don't they have some software to maintain too?

 

[atomicthumbs]

So, is there a temperature sensor in the RAM stack on top of the CPU or not?  Is there one anywhere inside the case?  

the OMAP3530 temperature sensor is accessible using the Band Gap Voltage and Temperature Sensor module, which also provides capabilities for current and voltage monitoring

The temperature sensor feature is used to convert the temperature of the device into a decimal value
coded on 7 bits. An internal Analogic to Digital Converter (ADC, realized with a resistor array) is used for
conversion. The recommended operating temperatures have to be in the -40 to 125°C for standard
conversion, and can go up to 160°C when using the thermal shutdown comparator output (TSHUT signal).
This signal indicates a too high temperature of the device. This signal is internally connected to a GPIO
pin.

see page 707 of this PDF

 

[hch]

frying some chips

hmmm

 

[Exophase]

i hate to sound like somebody who has to bring common sense to a thread, but:

the clocking of VLSIs has been about realiability. or IOW, your bits meeting expectations, doing exactly what they're support to do while traversing the circuit. i grin each time somebody comes up with 'what's the worst that can happen when i up the clock (but leave the voltages) - a crash at most' - i don't know about you, but i don't look at random crashes lightly. and as a software guy i would direct any stability reports about my software, coming from overclocked systems, rigth down to /dev/null.

btw, i'm a software guy, but i happen to work at a company that makes embedded devices (custom PCBs, custom enclosures, etc). a lot of time and effort is spent here to ensure TDPs are rigidly met, and the consumer can run the device the advertised way (i.e. 24/7, in unfavorable ambient t environments).

i know the prevailing mentality on these boards is 'hey, as long as i can run emu X at fps Y for limited time before it crashes, i'm good', but i don't share that mentality - for me the pandora is a linuxbox that will be treated as any other - it will have to be able to run tasks reliablty for the whole duration of the battery charge. and then again on the next day, and on the next, etc, while travelling with me to locations of diverse t conditions. if TI say that OPP3 goes as high as 600MHz (reliably) for pandora's omap, and that's the recommended OPP for everyday usage and normal lifespan, then that's what my pandora will be run at, and what my sw will be targeted at.

This is all your preference, and that's fine, but it's not relevant to the question at hand. While you may view a crash as unacceptable, it's not something that should be capable of bricking a Pandora and therefore isn't something that should bring warranty into question. Otherwise the warranty issue has just got a whole lot worse.

By the way, TI only specifies 500MHz at OPP3. 600MHz is only rated at OPP5. TI doesn't make any official allowance for > 600MHz. So if you consider OPP5 overvolting then even the advertised clock speed officially requires overvolting. On the other hand, OMAP3530s have been shown to run fine at 600MHz/OPP3.

 

[TaG]

How many dev boards have died due to overclocking? Non? See

But dev boards aren't in the cases, so are not reflective of production units.

How long before we're told the next delay is while there is extensive testing pre-release? I do hope not.

Genuine questions:
How can a comany say "this will take 10 years of use in these circumstances, or five years in these circumstances, before it will break" if they haven't run it for that long, which they can't with cutting edge tech. eg a car shows no sign of rust for eight years, then within 6 months it looks like it's got measels.

Statistically, how many units do you have to test to destruction to know that the result is represenative?