Overclocking Poses Risks to PDAs

[Jason Dunn]

<div class='os_post_top_link'><a href='http://www.pcworld.com/news/article/0,aid,109556,tk,dn022603X,00.asp' target='_blank'>http://www.pcworld.com/news/article...n022603X,00.asp</a><br /><br /></div>"Several software companies have developed small programs that allow handheld users who crave pure performance to run their processor's clock speed faster than advertised. Just because something can be done, however, doesn't mean it should, according to analysts. For a long time, PC enthusiasts have been tweaking their systems--a practice known as overclocking--to get better-than-advertised performance out of their CPUs, motherboards, and other components. But such efforts involved tinkering with the motherboard and BIOS, activities beyond the comfort level of average users. Now several companies are offering software downloads that enable users to increase the speed of their handheld's processor with just a few clicks and for just a few bucks."<br /><br />Tom Krazit must read sites like this - I'm surprised mainstream journalists know about overclocking. It's a pity he didn't interview any users that overclock their devices. He would have gotten an interesting answer: W?BIC! (or Y?BIC! depending how hooked on phonics you are). There's also an interesting psychological aspect here - people don't overclock their phones or clock radios, but they do overclock their PCs. Pocket PCs are more than just appliances, they're little PCs. Overclock away! (but only if you're willing to take the risk of frying your device)

[revolution.cx]

I was contacted by the author yesterday and I got back to him today. He never responded and instead included the phrase "Representatives from Wibble-wobble.com and Revolutionary Software Front did not immediately respond to e-mailed inquiries."

His article is long on suggestion, short on facts.

Notice how Intel bins a chip: test it at the fastest speed and slow it down until it doesn't have problems, toss it in the respective bin and sell it. So much for damaging a chip by running it too fast.

Here's my short response:
-----------


Saw your article. I'm pretty disappointed. I guess scare sells though.

Why would Intel or Dell say anything other than don't change the clockspeed, just buy the expensive chip instead?

Unlike desktop processors the Xscale is completely speed configurable in software. This is something Intel enabled by design - there is no "over clocking" going on.

And just like Intel bins their chips via testing, the consumer (and developers) can do the same. So far there have been 0 reports of problems related to changing the clock speed with our product and that's with a combined runtime far greater than any testing Intel has ever done.

[Cracknell]

That yahoo article might as well say Overclocking can cause cancer and reduction of sperm count from all the CPU radiation.

duh'

[revolution.cx]

And the article reminds me why I avoid mainstream computer press: they never link to anything off their site.

[Bradskey]

People should do what they want, because naturally I don't care what happens to their handhelds. But seriously, the benefit of OC'ing a PPC is negligible. Unless you're bent on playing a port of an old Quake game or something similar the added cycles probably don't yield a lot of benefit, especially with the 100 MHz bus starving the CPU. Then your battery life goes all to pot, or at least becomes less predictable. Pocket PCs are not like desktops, where you choose your PCB and CPU and other components and mix and match and tweak and modify and do all kinds of cute OC's and hacks to squeeze all the performance you can out of them. Handhelds are pretty much designed to function one way. Anything out of spec is higher risk than on a desktop, with little potential benefit, IMO. Your mileage may vary.

[Janak Parekh]

Well, the other twist here is that most people aren't actually overclocking their chips, but rather using XScale to scale up the frequency within rated limits. If I've heard correctly, all the 200MHz units are actually using 300MHz hardware underclocked. Actually overclocking via Pocket Hack Master is another thing.

Revolution.cx, about the author contacting you at the last minute: a lot of authors do try and contact people at the last minute. It might be them, or it might be sudden deadlines sprung up on them. I agree it's frustrating.

--janak

[hulksmash]

It is possible, however, that in the case of the iPaq 1910 that running it at an overclocked speed of 400Mhz may cause unforeseeable damage to your system since there is very little room left for ventilation in that unit, as in "practically none." I believe that there may not be noticeable difference between 200Mhz and 400Mhz since PPC 2002 was standardized specifically around the strongarm processor, which runs at 206Mhz. I haven't notice much of a difference between my Jornada 568 and iPaq 3955, except for perhaps a slight increase in response of "stored" web browsing services via Internet Explorer on Avantgo, etc. If someone has a different experience or opinion, I am interested in hearing it with a fair amount of substantiation in real-world observance. Thanks and great post! :n00b: ( I had to post the noob faces, never saw em before, no other signifigance other than that :lol: )

[revolution.cx]

I wish the author had told me he had a deadline, instead it was a very casual inquiry.

On a 200mhz 1910 the speed up is noticeable, otherwise I wouldn't bother.

Intel makes a big deal about the ability to change the clock speed of the XScale on the fly - notice the name of the chip????? The whole point is to change the speed depending on what the processor is doing. They spent many millions to make the chip do that and to market it as such.

ViewSonic devices have speed adjusting software built in.

People also don't realize that Clear Speed slows down the processor for increased battery life, can set a speed based on what program is running and will even lower your backlight for you as the battery gets weaker.

[seanturner]

People also don't realize that Clear Speed slows down the processor for increased battery life, can set a speed based on what program is running and will even lower your backlight for you as the battery gets weaker.

Actually, Pocket HackMaster will also allow you to change the processor speed based on the processor utilization. It's quite a nifty feature in that it makes the XScale truely scale.

[revolution.cx]

The 1910 won't run at 400 mhz. 300 mhz is the maximum stable. 333 runs fine but some units have problems with storage cards.

So I set my 1910 at 333, strapped it in an insulated foam container,turned the backlight on maximum, volume on full and ran an arcade game we are developing that uses every bit of CPU it can get. Left it that way through several full battery cycles. No problems.

It is possible, however, that in the case of the iPaq 1910 that running it at an overclocked speed of 400Mhz may cause unforeseeable damage to your system since there is very little room left for ventilation in that unit, as in "practically none."

[revolution.cx]

Whoops, forgot to mention that Clear Speed does that too (autoscale based on need). Even adds a speed indicator and manual override in the nav bar.

People also don't realize that Clear Speed slows down the processor for increased battery life, can set a speed based on what program is running and will even lower your backlight for you as the battery gets weaker.

Actually, Pocket HackMaster will also allow you to change the processor speed based on the processor utilization. It's quite a nifty feature in that it makes the XScale truely scale.

[Rirath]

People should do what they want, because naturally I don't care what happens to their handhelds. But seriously, the benefit of OC'ing a PPC is negligible. Unless you're bent on playing a port of an old Quake game or something similar the added cycles probably don't yield a lot of benefit, especially with the 100 MHz bus starving the CPU. Then your battery life goes all to pot, or at least becomes less predictable.

Completely untrue. First off, there's a big difference in speed. I've got the benchmarks on everything from VObench, movies, and PocketNester to prove it. Second, the battery life drain is not bad at all. Especially with PPCs like the Axim that have a lot of juice to spare. Tests show underclocking to 100mhz from 400mhz saves about 10-15% at best. So overclocking to 500mhz doesn't drain it like you think.

Third, can we drop the old bus is the limit quote yet? The Axims can't seem to overclock the bus just yet, but going from 400mhz to 500mhz does show a MARKED improvement in all areas even though the bus is still 100mhz. (Not as quite good as increasing the bus, but very close!) Besides that, the non Axim devices like the Ipaqs DO overclock the bus.

Finally, if overclocked correctly based on what unit you own, it's rock stable. If you use scaling, 75% the time you device won't be overclocked at all. Personally I don't like underclocking in steps so I simply use two settings: 100mhz under 25% cpu usage, and 497mhz above 25%.

[jizmo]

How I've understood, this whole bus issue is a bit like gears in the car. 400mhz XScale is like a car that goes on a gear that's too low, it'd go faster if it could change up. These overclocking utilities don't change your PDA to a bigger gear, they just press the pedal even further down thus increasing speed.

I run my e310 Toshiba overclocked to 236mhz most of the time, it gives me that extra boost that makes emulators and games run smoothly. While someone may claim that faster speeds aren't that important in everyday use, I myself can notice with ease if I've got the overclocking turned on or not. It just makes everything run that much snappier.

Oh, and nearly one year of use, and no problems here. Doesn't drain the batteries or affect stability. My bet is that e310 could be overclocked even further without harming the device (ipaqclk that does the overclocking claims that the its maximum speed of 236mhz is perfectly safe value).

/jizmo

[Deslock]

There's also an interesting psychological aspect here - people don't overclock their phones or clock radios, but they do overclock their PCs. Pocket PCs are more than just appliances, they're little PCs.
That's true... when I OCed my HP1910 to 472 MHz (kaboom... instant hard reset), it was very nostalgic. I OCed many PCs over the years (but don't bother anymore as I no longer play games). However, if my clock radio or phone operated slowly, I would also want to overclock it. So, I think for most users, OCing handhelds and PCs is primarily from a genuine need for better performance (though in some cases, it's probably that people just like to tinker).

BTW, I OCed my Palm too. While it was very responsive for most apps, using large spreadsheets, scrolling images, and doing file operations was painfully slow. My PPC is generally sluggish, so I leave it overclocked pretty much all the time.
Increased performance comes at a price. Faster clock speeds increase power consumption, decrease battery life, and add to the heat the device gives off. This can lead to system failures and the loss of critical data, analysts and vendors warned.
The article seems factually correct to me. Running my HP1910 at 300 appears to be perfectly safe. But I have noticed a slight increase in temperature in the casing, and there *could* be future side-affects from heat. Also, there is a slight decrease in battery life (though it's almost insignifacantly small). Lastly, if you overclock at higher levels, you can run into problems (generally with data loss and not hardware damage). So, the article may be a little sensationalistic in a couple spots, but it's not wrong.

[goirish!!]

People should do what they want, because naturally I don't care what happens to their handhelds. But seriously, the benefit of OC'ing a PPC is negligible. Unless you're bent on playing a port of an old Quake game or something similar the added cycles probably don't yield a lot of benefit, especially with the 100 MHz bus starving the CPU. Then your battery life goes all to pot, or at least becomes less predictable.

Completely untrue. First off, there's a big difference in speed. I've got the benchmarks on everything from VObench, movies, and PocketNester to prove it. Second, the battery life drain is not bad at all. Especially with PPCs like the Axim that have a lot of juice to spare. Tests show underclocking to 100mhz from 400mhz saves about 10-15% at best. So overclocking to 500mhz doesn't drain it like you think.

Third, can we drop the old bus is the limit quote yet? The Axims can't seem to overclock the bus just yet, but going from 400mhz to 500mhz does show a MARKED improvement in all areas even though the bus is still 100mhz. (Not as quite good as increasing the bus, but very close!) Besides that, the non Axim devices like the Ipaqs DO overclock the bus.

Finally, if overclocked correctly based on what unit you own, it's rock stable. If you use scaling, 75% the time you device won't be overclocked at all. Personally I don't like underclocking in steps so I simply use two settings: 100mhz under 25% cpu usage, and 497mhz above 25%.

I own HackMaster and have used it from time to time, basically just playing around, but not all day every day. As a matter of fact I turned it on the other day and some how wound up with a device induced hard reset. I was running at 497 and have never had that happen before. No biggie as I save everything pretty often.

My actual question(s) is do you run it all the time, and if not about what percentage of the time are you at 497 and for how long at a time?

Just wondering what you notice on a day to day basis.

[bikeman]

Jizmo - how did you overclock your e310? I have one that I want to fool around with.

[jizmo]

Jizmo - how did you overclock your e310? I have one that I want to fool around with.

Perform a search in google for "js overclock" .. it's an old utility, originally intended for iPAQs, but works with other ARM processors as well.

I don't know how it goes with XScales, but 206mhz -> 236mhz really makes a difference in emulators, for example; many of them become actually playable. In fact, 236mhz ARM is by far the fastest PPC as 400mhz XScale is finely slower than 206mhz ARM.

/jizmo

[Bradskey]

Third, can we drop the old bus is the limit quote yet?

You can't drop it if it is still and factor, and in almost any device that we would generally recognize as a GP computer, the bus speed is a factor. I didn't say it was a limit, or that it precludes ANY performance increase. CPU intensive apps are naturally less affected than memory and I/O intensive apps. You may (or may not) eventually reevaluate what you consider to be a sizeable performance increase if devices with a 200MHz+ bus ever become available. I personally wouldn't be eager to OC the bus on such a device, since it affects more than the memory and the CPU, and not knowing the intricate details of the design you never know whether there is a cheap shortcut component in there somewhere that just can't handle it. If it works for you then I'm happy for ya'. ;-)

[Jonathon Watkins]

Completely untrue. First off, there's a big difference in speed. I've got the benchmarks on everything from VObench, movies, and PocketNester to prove it. Second, the battery life drain is not bad at all. Especially with PPCs like the Axim that have a lot of juice to spare. Tests show underclocking to 100mhz from 400mhz saves about 10-15% at best. So overclocking to 500mhz doesn't drain it like you think.

Rirath, what app do you use to overclock the Axim? Can it go past 500mhz? Is it Completely stable?

[goirish!!]

Completely untrue. First off, there's a big difference in speed. I've got the benchmarks on everything from VObench, movies, and PocketNester to prove it. Second, the battery life drain is not bad at all. Especially with PPCs like the Axim that have a lot of juice to spare. Tests show underclocking to 100mhz from 400mhz saves about 10-15% at best. So overclocking to 500mhz doesn't drain it like you think.

Rirath, what app do you use to overclock the Axim? Can it go past 500mhz? Is it Completely stable?

Although not directed toward me, NO it can not go past 497mHz. Anything past that and it automatically locks up. (PocketHackMaster)

[Jonathon Watkins]

Although not directed toward me, NO it can not go past 497mHz. Anything past that and it automatically locks up. (PocketHackMaster)

OK, so, how stable is it? Does it seem faster to you Goirish?

[goirish!!]

Although not directed toward me, NO it can not go past 497mHz. Anything past that and it automatically locks up. (PocketHackMaster)

OK, so, how stable is it? Does it seem faster to you Goirish?

Well it does seem somefaster, but don't expect miracles. As far as the stability, read some of my questions earlier in this thread.

[bmhome1]

PocketHackMaster only overclocks the processor when pushed to maximum demand. It underclocks the processor when demand falls below (either preset or adjustable) that threshold. The PPC is never run at only full overclocked (unless 100% games and movies). So using it wisely could actually INCREASE potential component life.

[Jonathan1]

Has anyone ever heard of a Pocket PC actually dying from an OC? Everything I've ever read is that you get an unstable system which in turn requires a hard reset which in turn fixes the problem.

You'd think that if a device started getting sufficiently warm in one's hand the person would have enough common sense to scale it back before something REALLY bad happened. *shrugs* You’d think.

This article sound like some of the wonderful articles I read back when OCing started becoming mainstream (Or if nothing else not relegated to only the hackers of the world) Where zdnet, pcworld, and the like all clamed that you’d kill your system if you OCed it. Now that its become more common you don’t hear nearly as much doom and gloom about OCing a puter. So it goes with the Pocket PC.

[sjcbulldog]

So, I have watched this thread and working for a semiconductor company I felt compelled to respond. Most silicon companies that have speed bins for a product want the maximum yield to the fastest speed bin. This gives the maximum flexibility for the company and in general the demand for parts in the faster speed bins increases over time. However, the marking (labeling on the top) of these parts may be delayed in their production flow until the part is ordered from the customer. This is so that if a customer orders a 300 MHz part, and the semiconductor company only has 400 Mhz parts in inventory, they just mark it as a 300 Mhz and ship it. This is simplifying a little, but basically this is how it works. This means many times you get a device with a 300 Mhz part in it, but the part really passed the 400 Mhz tests. The more mature a given device is, the more the company has worked on yield issues and the more this is likely to be true. Also, not that the semiconductor company is going to test the device under the worst case temperature and voltage conditions that are specificied for the part, not necessarily the conditions the part will operate under in your device.

However, a company may also really have yield to the lower speed bin and you may have a lower speed bin part. This would not really affect the part during test, as the pins of the part are connected to a tester and not a system. If the part fails and causes the pins to go to a state that is unexpected from the device under normal operation, but tester is tolerant to this condition. However, if the part fails in your device, the circuits in the device connected to the pins of the part could cause damage to portions of the device.

Now, having said this, in order to see damage the following must be true:

o The device must really be of the slower speed bin
o The device must really fail in your device
o The failure must put the pins in an unexpected state
o The unexpected state must cause real damage to your system

I think the percentage of case where all of these are true is very small. However, the semiconductor companies and the PDA manufacturers have to state that this damage can occur to cover themselves, no matter how unlikely it is that this will be a problem.

This is generally not an economic issue, as the people how build the devices will not buy 300Mhz parts and sell them in a 400Mhz device because of the liability they will incur. In turn for a 400Mhz they will pass the costs of the faster part to the consumer.

Hope this sheds some liight on the subject.

Thanks

[Fishie]

Intresting read, thanks.

[bikeman]

Jizmo - thanks for the information. I installed it and it is running with no problems. I guess now my e-310 is safe until Sprint has a Pocket PC phone edition.

[PetiteFlower]

Of course, even if it is extremely unlikely for this damage to occur, it's still possible. And if, on that small possiblity, damage does occurr, you're screwed since you voided your warantee. Call me a wuss but it's not a risk I'm willing to take, especially considering that there's nothing I do on my machine that I feel I need extra speed for. Pocket Nester runs just fine even at the power save mode of 200mh, and that's probably the most taxing thing I run on it!

[revolution.cx]

You can read the follow up article here:

http://www.idg.net/ic_1191645_9713_1-5069.html

Which provides a more balanced take on the subject (of course I'm biased, that's me they interviewed).

I also got this tidbit from a person who works at a company that makes the machines that manufacture chips:

"If the processor will overclock without a change in voltage, and timings are not an issue, then there should be no problem. Heat dissapation may be worse as mentioned.

Sometimes the manufacturers don't rate the chips at the higher speed ratings because they fail to perform at the -5% voltage 85C point...and sometimes all the chips will work at the higher speed and they simply grade some at the lower speeds to fill the demand. I would think no matter what the architecture if Intel makes these chips with the 858 or 859 process 400MHz is duck soup...after all, they ship 1Ghz P-IIIs and Celerons and Ultra Low Voltage 750Mhz P-IIIs with this process."

Notes:
- 858 and 859 are Intel manufacturing lines.
- The processor does "overclock" without a voltage change.
- Intel states explicitly that all XScales do not need a heat sink because heat is not an issue.

[Jonathon Watkins]

Certainly makes sence. I'e been overclocking desktop processors happily since 1995 and have never damaged a CPU this way. 300Mhz to 400Mhz sounds just fine. :P

[Jonathon Watkins]

And I'm now running my Axim at 500Mhz thanks to XCPUScalar 400 by ImmierSoft. I do the same as Rirath - below 25% it's at 100Mhz and above that it runs at 500 Mhz. It does feel snappier. Very good.!

See here for details: http://www.pocketpcthoughts.com/forums/viewtopic.php?t=7720&highlight=overclock+free

[Jasper K Abraxxis]

It IS possible to run atleast some models of the Ipaq 1910 at 400 Mhz, as I have one that runs very happily, and I must say, perfectly stable, at this speed.

It has run this speed (well actually 398 Mhz, but lets not split hairs) for about a month now using the COMPLETELY freeware "Speed Stepper".

Anyone who says that running an Ipaq 1910 at double its default clock speed would not mnake a noticable difference is either A\ - devoid of all sensory input! B\ - a gibbering moron, or C\ - working for Intel. :)

It makes a HUGE difference! Mpegs play smoothly with few or no dropped frames, games like Doom and Quake run FAR smoother, and even the simple task of negotiating menus etc is much more responsive.

The unit itself does not get noticably hot, and battery life is reduced..... by about 10%! Who gives a rats anus? Its much faster and pleasant to use.

Now, just because some Ipaq 1910s can run at 400 Mhz doesnt mean all will. I dare say that different units, from possibly different manufacture dates/batches use different grade XScale Cpus. But give it a try, Speed Stepper is a free download, easy to use and the worse that can happen is you will have to do a hard reset (so back up your Pocket PC first). Oh BTW - I have been overclocking Palm devices and Pocket Pcs for many a year now - I have never even heard of hardware damage from overclocking such devices - its crap, the worst that happens is the system will lock up (hard reset) or it will become unstable to use (knock the speed back a little) - so have fun. :)