Question About Overclocking


borgqueenx

Well-Known Member
Joined
May 21, 2008
Messages
1,887
Age
34
Location
The Netherlands, Overijssel
Website
gdteam.net
I have read this page:
CODE
http://www.arm.com/products/CPUs/ARM_Cortex-A8.html


i have read more pages about overclocking but i want to hear more about it from the pandora community.
I have heard that sometimes overclocking will break your proccesor, but some other people tell me that it is completly save. and more people tell me that overclocking can be done but will decrease the lifetime of a proccesor.
The page ive read about the cortex A8 says that it can reach 1ghz plus.

Maby there's more involved then overclocking? like the ram, or connectors?

Please tell me about it, so i can know if i can savely overclock the pandora so it can reach 1ghz. it may come in handy sometimes i guess to overclock it :)
 
Last edited by a moderator:
CODE
http://www.gp32x.de/board/index.php?showtopic=43293
http://www.gp32x.de/board/index.php?showtopic=42953&st=0&p=625364
 
Last edited by a moderator:
A lot of processors have safeguards in place to prevent any damage that things such as overclocking might cause. I believe the Pandora is no exception (but I'm not entirely sure there).
In my experience, overclocking too far will simply cause the machine to crash. Annoying, but fixable with a reboot. The life of the processor is shortened, but I have no idea by how much.
 
Last edited by a moderator:
To summarize what I recall:
Overclocking is possible on the OMAP. If you overclock too much, it causes instability as instructions fail, up to and including complete system crash. This does not cause any short term damage: if you've overclocked too much and it's crashed, simply restart at default clock speed and it'll be fine. Overclocking does cause long term damage, however, strictly speaking. It's not really so bad: the estimates were, running at standard clock and voltage, 24 hours a day, 365 days a year, you would get 20 years of use; running at overclocked high voltage for the same 24 hours a day, 365 days a year, you only get 5 years of use. I think I remember those being the figures we worked out.
Point was that you could overclock if it was really needed without much fear of permanent damage, but the hope was that we wouldn't have to ever overclock.
 
Last edited by a moderator:
'borgqueenx' said:
The page ive read about the cortex A8 says that it can reach 1ghz plus.

Please tell me about it, so i can know if i can savely overclock the pandora so it can reach 1ghz. it may come in handy sometimes i guess to overclock it :)
This figure gets thrown in a lot. Let me clarify it for everyone. ARM's core architecture can run up to 1Ghz that doesn't mean the OMAP can do this. Its like ARM provides a blueprint for the processor and gives it to TI. Its then TI's responsibility to implement all the features on the blueprint to get the processor to run at 1Ghz.

TI and often other companies will not include all the silicon features in order to conserve power, die space(cheaper to produce and test), and for better overall stability.

That being said TI has stated that all OMAP 3 are guaranteed to operate at 600Mhz while anything over that is the luck of the draw.
 
Last edited by a moderator:
'borgqueenx' said:
I have heard that sometimes overclocking will break your proccesor, but some other people tell me that it is completly save.
Before the main thrust of my argument, I have searched the net to find a few things that have been deemed to be "safe" or even "compeltely safe" by other people. These include:

Cocaine.

Cocaine mixed with ecstasy, LSD, mushrooms or other drugs provided that you do not have a heart condition or are in "bad shape".

The World's First Completely Safe For Work Porn Site

Banks... (savings, pensions, investments, shares in the banks themselves...)

The The Large Hadron Collider

Silicone breast implants

Cricket in India (honestly, these are all true. Search the web of "completely safe")

Bernard Matthews turkey-based foods (the article included a photograph of a BM worker cycling around the farm in protective clothing including a face mask)

Getting Your Eyeball Tattooed (I kid ye not!)

Now then. Ahem. On to overclocking. Overclocking can involve more - much more - than icnreasing the rate at which the CPU runs. You may have to adjust the speed of other items, such as RAM for example, and you may end up having to increase one or more voltages. There are all sorts of technical issues but to me these are all secondary to the plainest simplest perhaps too "common-sense" approach to the item.

If TI could reliably say that it's ARM package could run at 1GHz instead of 500MHz or 600MHz, would it not make sense for them to do so - and claim superior speed perhaps for an increased price? There is usually a reason that vendors claim a specific level of performance for an item, and that is generally because reliablity across all manufactured units is prone to variation and they set their limits within an area they consider "tolerably gauranteeable" or that using the product at a faster rate COULD cause damage. Possibly both. I can almost guarantee that if TI could legally say that their package could run at 900MHz without causing something bad - from damage to stability to battery life - then they would do so. It would be in their business interests to do so.

On the other hand, lots of people and equipment survive surprising abuse *and then state it is perfectly safe to do so*. From bungee jumping to nuclear reactors to taking cocaine to having breast implants the size of Montanna to having their eyeballs tattooed, survivors often claim complete safety because THEY didn't die. Those who DID die, of course, don't tend to say much because theya re dead. So you TEND to get one side of the story.

Completely safe? No. We've had "Pandora" motherboards fail overclocking for some uses at 750MHz IIRC, but only time will tell how safe overclocking is - how many Pandoras fail at what level of overclocking, and how many seem stable at whatever speeds.

Ta
 
Last edited by a moderator:
It's overrated. And so is underclocking, since it just stops when idle. It's fine with defaults
 
Last edited by a moderator:
Not totally correct Monk. One of the reasons for overclocking, is because it's cheaper to make a million indentical chips, then just put lower clocks on some and sell them at a lower price point than to create chips capable of different speeds. It's a very good buisness model. For that reason, one can save a lot of money by buying himself a cheaper chip and overclocking to the speed of a more expensive one.

Your argument about safety is flawed. Bad news is good news, we hear about stuff being un-safe all the time, but how often do you get somebody on the news saying that the latest extreme sport is totally safe. You're always going to get deaths, but you've still got a higher chance of dying in a car accident than of a cocaine overdose.
 
Last edited by a moderator:
'Butterman' said:
Not totally correct Monk. One of the reasons for overclocking, is because it's cheaper to make a million indentical chips, then just put lower clocks on some and sell them at a lower price point than to create chips capable of different speeds.
True, but you can NOT guarantee that you'll get a chip that's actually better than you've paid for - at any moment the CPU vendor could start knocking out lower quality chips, ones that only just pass the guaranteed spec - because that IS all that's guaranteed, and all one has paid for. It's a bonus when you get more or betetr than you paid for (and a nice one) :D

'Butterman' said:
Your argument about safety is flawed. Bad news is good news,
Oooh, I LIKE that phrase! However, let's not derail your post TOO much :)

'Butterman' said:
we hear about stuff being un-safe all the time, but how often do you get somebody on the news saying that the latest extreme sport is totally safe.
A single page of Google later:

CODE
http://www.kidsexercise.co.uk/ExtremeSports.html

QUOTE
While onlookers may gasp at the perceived dangers involved with extreme sports, most extreme athletes feel that their sports are relatively safe compared to the pleasure that they get from participating.


CODE
http://www.thanksdarling.com/experience-day-articles/Extreme-Sports-for-the-Not-So-Extreme.htm

QUOTE
There are several extreme sports that can give you an adrenaline rush like no other-but are also relatively safe and ideal for all skill levels.


CODE
http://www.extremesportscafe.com/main.html

QUOTE
White Water Rafting with Extreme Sports Cafe
White Water Rafting with Extreme Sports Cafe in New Zealand is the first of our White Water Rafting locations. Soon we will offer this exciting extreme sport with safe guidance as well as the opportunity to go and white water rafting by yourself inmore locations around the world.


No, it's not the Nine O'Clock news, but a quick sampling on the net will suggest that almost anything is safe - or dangerous, or sexy or boring or pretty much whatever you're hoping to find. IMHO. My single click one-page Google certainly suggests that "extreme sports" are a lot safer than *I* had imagined.

'Butterman' said:
You're always going to get deaths, but you've still got a higher chance of dying in a car accident than of a cocaine overdose.
Naturally, since I don't take cocaine yet I do frequent this country's roadways. However, that doesn't meant that I haven't known people who have died as a result of taking drugs.

At the end of the day, you pay your money and you take your chance. However, I find saying that overclocking, as a generic behaviour, is "completely safe" to be far less accurate than saying that it is not, across the board, completely safe. I do not find my argument to be flawed.
 
Last edited by a moderator:
Maybe after all the first batch Pandora's are shipped someone should make an app that tests the overclocking ability and gives everyone some uniform response so anyone willing to test their own chip can post the results here on the board...might be useful info to see what the average overclock speed would be.

I would be willing to run something for a few minutes or so to get the info even if I never intend to overclock for any other reason...but maybe this would be useful for dev's just in case they need just a bit more to make something work and wanted to make sure they were not past the lowest reported value.

Just a suggestion..I don't know how to code so I'm not sure how possible or useful this suggestion is...but I personally am curious enough to participate if something like this happens...as the old NBC public services announcements said..."The more you know..."
 
Last edited by a moderator:
'Butterman' said:
Not totally correct Monk. One of the reasons for overclocking, is because it's cheaper to make a million indentical chips, then just put lower clocks on some and sell them at a lower price point than to create chips capable of different speeds. It's a very good buisness model. For that reason, one can save a lot of money by buying himself a cheaper chip and overclocking to the speed of a more expensive one.
This is true if there is an higher speed model in current production, and even then if you buy the lower speed one you may get unlucky and find one of the bad ones that are actually limited.

however, AFAIK there is no higher speed OMAP with the same technology of the one in Pandora currently under production right now, so we have no chance to just get lucky and get an higher specced "for free"
 
Last edited by a moderator:
Monk said:
Completely safe? No. We've had "Pandora" motherboards fail overclocking for some uses at 750MHz IIRC, but only time will tell how safe overclocking is - how many Pandoras fail at what level of overclocking, and how many seem stable at whatever speeds.
No Pandora board that I know of has failed to achieve 900MHz when using the appropriate voltage level (that TI has a particular name for, and has particular statements on its safety). In fact, the more recent boards achieve higher overclocks w/o overvolting, indicating that headroom is increasing, not decreasing. Whether or not this remains true for every batch of Pandoras is anyone's guess, but what an educated guess would be that a high percentage of first batch Pandoras will be able to achieve this level of overclocking (please no one quote me on this, it's still just a reasonable GUESS).

I've never heard of a case of a CPU instantly frying due to overclocking, although it will probably happen if you overvolt too far. I doubt OMAP3530 will even let you overvolt it that far into such dangerous regions, and it probably has overvolt protection all over the chip (the power supply circuitry might as well). Overclocking can result in the temperature increasing a lot, but these CPUs are designed to operate in very high temperature ranges. What's more is that the "absolute maximum" and storage temperatures are even higher, meaning that even if they stop functioning by a certain point they won't be permanently damaged. That means if they shut down due to temperature being too high then they'll start cooling down before they can reach the dangerous zone.

It's commonly said that lifespan is decreased with overclocking, which may very well be true (probably more so with overvolting involved). But this is a vague and difficult thing to quantify. This is especially true when the chips are being sold well under-grade. Other OMAP3xxx chips are binned for higher clock speeds but are sold in much smaller quantities (almost definitely, look at devices using 35xx vs 34xx) so it's not unlikely that some from the higher grade bins are ending up sold as 3530s.

lulzfish said:
Overclocking is a waste of battery on portable devices, and you only need the extra power in edge cases.
Meaningless subjective statement. To someone willing to compromise battery life for CPU power to improve the useability of an application (which is a completely realistic scenario) it's not a waste. Why don't you let people decide these things for themselves?

hells_dark said:
Yes. Overclocking will drop the battery faster than make the power increase.
This statement is even sillier, as if battery consumption and improved power use the same units and are somehow comparable...

Besides, CPU is only one part of the whole system. If you increase the clock speed by 50% you won't end up with 66% the battery life.
 
Last edited by a moderator:
'Exophase' said:
hells_dark said:
Yes. Overclocking will drop the battery faster than make the power increase.
This statement is even sillier, as if battery consumption and improved power use the same units and are somehow comparable...
I disagree.. :unsure:
If you make a curve (power usage vs hum... what you gain in speed) , it will be exponential and not right..

well, i have difficulties expressing myself in english ..
 
Last edited by a moderator:
'Exophase' said:
This statement is even sillier, as if battery consumption and improved power use the same units and are somehow comparable...

They actually can use the same units. If you take (overclocked MHz) / (base MHz) you end up with a unitless value like 1.23, as dividing a unit by itself gives 1. Similarly (overclocked battery life) / (base battery life) is unitless. In the end you have percentage values for processor speed and lifespan, which you can graph against each other.

I do agree, however, that even demonstrable heavy losses in battery life might be worth it if the user wants to run a particularly power-hungry program. For example I might want to compile a largish application that isn't available as a native build and I have the power adapter and a power outlet available. If the adapter can keep up with the increased power draw my only cost would be a shortened CPU lifespan (assuming that calculation errors can be ruled out through prior testing at overclocked speeds).
 
Last edited by a moderator:
'Exophase' said:
'Monk' said:
Completely safe? No. We've had "Pandora" motherboards fail overclocking for some uses at 750MHz IIRC, but only time will tell how safe overclocking is - how many Pandoras fail at what level of overclocking, and how many seem stable at whatever speeds.
No Pandora board that I know of has failed to achieve 900MHz when using the appropriate voltage level

{turns research mode on}

I thought Pickle had failed to get over 750MHz reliably:

Pickle said:
this is the physical limit was able reach.
and
Pickle said:
the highest i could go was 750
But maybe I was reading Pickle wrong, or maybe Pickle doesn't know how to overclock? The latter still suggests that overclocking is not as simple as the OP was perhaps lead to believe...[/quote]
'Exophase' said:
Whether or not this remains true for every batch of Pandoras is anyone's guess, but what an educated guess would be that a high percentage of first batch Pandoras will be able to achieve this level of overclocking (please no one quote me on this, it's still just a reasonable GUESS).
Well, that basically confirms a bunch of what I've been saying - you CANNOT rely on Pandoras being able to meet a spcific over-spec state. The lower you set your aspirations the better your chances, for sure, but it is over specification.

'Exophase' said:
It's commonly said that lifespan is decreased with overclocking, which may very well be true (probably more so with overvolting involved). But this is a vague and difficult thing to quantify.
Absolutely correct, I believe. It's also been said that smoking can reduce a humans lifespan, but that is also difficult to quantify accurately because we're all different and treat our bodies slightly differently. The instances where someone dies the first time they try any "recreational" drug are a tiny, tiny percentage - I have never seen it. Doesn't meant that drugs are "completely safe". that's the question I was addressing - is overclocking, as has been said, "COMPLETELY safe"? I have very recently been told that a trason I have been feeling increasingly ill for the past long while is because of an allergy I was completely unaware of, to an innocuous everyday item. I've been having great trouble just breathing because I was being adversely affected (and indeed still am, but to a lesser degree now it's diagnosed and we're trying to address it) by something which is "completely safe".

If you want to argue that overclocking often has to detectable adverse effect then stop arguing - because I'm not arguing against that. I'm arguing that it's not COMPLETELY safe. It might in some cases cause catastrophic issues abruptly, but it's far more likely IMHO to cause damage over a long-term (possibly even a very long term) to the point where a user will not know whether overclocking was "the cause" or not. Still yet there may never be any issue, the Pandora may be abandoned and unloved as an old, low-power machine incapable of running modern software or the cost of replacement numbs over the years may cause the user to repalce the whole unit with something newer. Too many unfathomables. But is it completely safe? No - there are risks, generally considered to be minor, associated with running technology at a higher voltage/faster/hotter rate than it is specified for.

Ta,
 
Last edited by a moderator:
Monk said:
I thought Pickle had failed to get over 750MHz reliably:

Pickle said:
this is the physical limit was able reach.
and
Pickle said:
the highest i could go was 750
But maybe I was reading Pickle wrong, or maybe Pickle doesn't know how to overclock? The latter still suggests that overclocking is not as simple as the OP was perhaps lead to believe...
I can't find the quote now as I can't remember who exactly said it, but someone said they'd talked to Pickle and discovered he'd attempted to overclock at stock voltage. Once the voltage was increased, it reached the 900. That's what I remember, but I could be mistaken. Give me a bit and I'll try and track down the exact post.
 
Last edited by a moderator:
WizardStan said:
I can't find the quote now as I can't remember who exactly said it, but someone said they'd talked to Pickle and discovered he'd attempted to overclock at stock voltage. Once the voltage was increased, it reached the 900. That's what I remember, but I could be mistaken. Give me a bit and I'll try and track down the exact post.
If you can find the quote that's great, but I'm willing to go with your memory - as fallable as memory is, if you're reasonably sure that you remember it then there's a good chance it happened. Which is great :)

Edit for search results - all I can find is Exophase at:

CODE
http://www.gp32x.de/board/index.php?showtopic=47027&st=120


saying:

Everyone I know of who has overvolted to the "proper" voltage (TI "approved" amount, as in, they say it reduces overall lifetime but doesn't put the CPU in immediate necessary danger) has gotten 900+MHz. Pickle was running at normal volts - the clocks at that voltage have ranged from like 625MHz to 800MHz or so.

though I do not know from that quote if Exophase had actually spoken to Pickle about it or was just making assumptions. Still, I note the "[TI] say it reduces overall lifetime but doesn't put the CPU in immediate necessary danger"
 
Last edited by a moderator:
Back
Top