The Truth About The Pencil Trick

[reaper79]

I decided to start a new thread on this and not continue the old one. When a thread is over 20 pages and has been active for 2 weeks I tend to stop reading it, and I'm guessing others do too.

Anyway, I was digging around inside my GP32, doing some recon for a possible battery mod, and a wire came disconnected from the R52 resistor. I don't have a pic up, but I resistor modded my unit a while ago, but I didn't do the conventional mod where you replace the resistor(s). Instead, I connected a 1.5megaohm resistor (1500k) to the circuit board and wired it in parallel with the 470k R52 resitor, this brings the resistance down to 360k, and my vcore up to 1.96. Before I resoldered the wire, I thought for shit and giggles lets investigate the pencil trick. It just didn't seem right that such a small voltage increase could lead to such huge overclocks.

For those that don't know the trick, it envolves applying graphite from a pencil to the R52 resistor, and in some cases the R48 resistor to up the vcore voltage to the CPU and thus overclock your gp32 better. :rolleyes:

You can read about it here

So, as the story goes I got an old mechanical pencil and applied graphite to the R52 resistor. I had the unit hooked up via an external battery pack, and I had a digital multimeter set to read the vcore. Original reading was 1.79 volts with my addon resistor disconnected, this is the stock voltage for my unit. First application of the pencil gave me 1.79v, second application 1.79v, third application and the gp32 would turn on but not boot (just like what has happened to those that have tried the pencil trick), voltage read out... 3.1 volts to the vcore, no joke. Removed some graphite, and the unit would now boot, reading... 2.68 volts. That's right, your unit will still boot even when the voltage is well beyond the 1.65-1.95 recommended range. I played around with the pencil trick a bit more (for about an hour) and got voltages anywhere between the stock 1.79 on up to 3.18 at the top, and many voltages in between. My unit booted with everything under 2.68 volts, and would not boot with anything over 3 volts. 2 morals to this story:

1) If your unit is not booting after the pencil trick, it is because your vcore is around 3 volts. However, after having my vcore over 3 volts at least three times, this shows that the voltage will not outright destroy your cpu.

2) Just because you are running stable with no crashes does not mean your CPU vcore is within safe limits. I left mine at 2.2volts for a bit and played through some games in Drmd with some overlocking settings, and it still worked. The key to remember is that vcore is proportional to the overclock, a 10% vcore increase (1.8 to 1.98) should on a 166mhz gp32 allow you "roughly" a 10-15% overclock (182-190%) If your jumping from 156 to 200mhz, think proportion, then think how high your vcore may be. If your getting a 25% overlock, your probably at a 20% vcore overvolt. (This is all a rough estimate, other factors such as silicon quality and such make a difference to the success of the overlockability).

Overclocking a cpu will ALWAYS lessen its life. It just may not be enough for you to notice. If your CPU has the potential to last 10 years, overclocking it may lessen it to 5 years. You probably wouldn't notice that would you? If your running at 220mhz and 2.6 vcore right now, maybe you'll get 6 months to a year... maybe. Is that worth it?

Since then, the graphite has been remove and sprayed down with contact cleaner. The resistor wires have been resoldered, and I'm back safely at 1.961 running at 172mhz. Cheers!

 

[Quiest]

Thanks for the good info!

 

[Zap]

Where di you measure vcore ? Is it really between T7 and T8 ?

 

[KickinWing]

No it's not worth it since almost every program for the GP32 runs fine at 156mhz.

 

[reaper79]

Where di you measure vcore ? Is it really between T7 and T8 ?

​

Yepperz, one is vcore, the other, T7, is gnd.

No it's not worth it since almost every program for the GP32 runs fine at 156mhz.

I'd argue that everything is best at 166mhz. This also is the most common limit to those programs that can use overclocked settings. Mine was only stable up to 154mhz before resistor modding it, and I noticed a healthy difference between that and 166mhz. Especially when running Drmd, GPengine, or OSnes9x. I think that the extra headroom helps mainly with processing the sound rather than graphics.

 

[KickinWing]

maybe, but my non lit unit only went to 156, i have a flu now that goes 166mhz and i notice no difference and i continue to run the programs at 156mhz.

 

[Drak]

maybe, but my non lit unit only went to 156, i have a flu now that goes 166mhz and i notice no difference and i continue to run the programs at 156mhz.

​

well for the osnes9x its important to have that oomf that the overclock gives. I run it at 172 mhz, and from 160 its a difference of 4 fps in most games that I play.

 

[GP40]

1) However long it takes, it will lower cpu life, whether its noticable or not
2) Its inaccurate, takes ages and then the coating wears off after a while
3) ICBA

 

[Goity]

maybe, but my non lit unit only went to 156, i have a flu now that goes 166mhz and i notice no difference and i continue to run the programs at 156mhz.

​

well for the osnes9x its important to have that oomf that the overclock gives. I run it at 172 mhz, and from 160 its a difference of 4 fps in most games that I play.

​

You have a 172mhz osnes9x? Is there anywhere for me to download this?

Anyways back on topic.
I never really did like this mod. My Gp32 runs at 180mhz and I'm happy with it.
The resistor mod is much more accurate and there's less room for frying the cpu like the egg it is.

 

[Robster]

Nice report, reaper79, thanks! I've always thought the pencil trick would be very hit-or-miss, and your research is pretty conclusive.

One thing I have to disagree with: why do you (and so many other people) say that overclocking will shorten the CPU's life? I say that's bollocks!

Now, there's two things to consider here: first, the electrical effect of applying more voltage than is recommended. My understanding of this is that there will be no effect at all, until the voltage is raised to the point where the insulation layers on the silicon break down and you get some sort of avalanche failure. As you've shown with your experiment, a brief period of operation at 3.1V didn't cause this to happen, so I think it's clear that 2.2V won't do anything bad.

The other thing is the thermal cycling. This is a big problem for P4's and Athlons which are overclocked and even with the best external cooling solutions, the CPU's core temperature can easily go over 100 degC. This causes thermal expansion etc, and eventually will cause stress fractures in the little wires that connect the silicon to the lead frame. Also the copper interconnects on the actual silicon are fragile, if you have a look at an electron microscope picture of those 130nm structures with 7 or 8 metal layers, you'll wonder that it works at all! The metal interconnects are also vulnerable to breakage from thermal stress.

However, none of this is relevant to the GP32. Even at 166MHz, with 2.2V core, the CPU doesn't heat up anything like a P4. The temperature rise is barely noticeable. And so the thermal stress is also lower and won't affect the chip's life in any measureable way.

I'm saying: increasing your core voltage a reasonable amount, and overclocking to 166 or even 200MHz is completely safe and won't shorten your GP32's life.

As always, I'm quite happy to be proved wrong! Anyone got any actual evidence?

 

[Drak]

maybe, but my non lit unit only went to 156, i have a flu now that goes 166mhz and i notice no difference and i continue to run the programs at 156mhz.

​

well for the osnes9x its important to have that oomf that the overclock gives. I run it at 172 mhz, and from 160 its a difference of 4 fps in most games that I play.

​

You have a 172mhz osnes9x? Is there anywhere for me to download this?

Anyways back on topic.
I never really did like this mod. My Gp32 runs at 180mhz and I'm happy with it.
The resistor mod is much more accurate and there's less room for frying the cpu like the egg it is.

​

ya goity its in the Overclocked stuff page in Beta Testing on one of the last pages

 

[I izda Boi inda Hood!]

maybe, but my non lit unit only went to 156, i have a flu now that goes 166mhz and i notice no difference and i continue to run the programs at 156mhz.

​

well for the osnes9x its important to have that oomf that the overclock gives. I run it at 172 mhz, and from 160 its a difference of 4 fps in most games that I play.

​

You have a 172mhz osnes9x? Is there anywhere for me to download this?

Anyways back on topic.
I never really did like this mod. My Gp32 runs at 180mhz and I'm happy with it.
The resistor mod is much more accurate and there's less room for frying the cpu like the egg it is.

​

ya goity its in the Overclocked stuff page in Beta Testing on one of the last pages

​

Bloody hell, Drak! It's 3.17, GET TO BED!!!!

Oh, you're in Canada.... :blink:

 

[reaper79]

Nice report, reaper79, thanks! I've always thought the pencil trick would be very hit-or-miss, and your research is pretty conclusive.

One thing I have to disagree with: why do you (and so many other people) say that overclocking will shorten the CPU's life? I say that's bollocks!

​

It is pretty hard fact that anytime you increase voltage/current/temperature of any kind of circuitry, it WILL shorten its life. There is a difference between shortening life and causing failure. To make an analogy: Take two identical cars, and drive one in first gear at 50mph and drive the second in 5th gear at 50mph (feel free to use kilometers if your not accustomed to the screwy U.S. system) and guess which engine will fail first overtime?

There are two known events that happen gradually over time. Electromigration and Silicon Migration. The atoms of the metal internally will shift over time. Electricity and heat will cause this. Extra electricity and heat will accelerate it. It works backwards if you lessen the voltage/temperature. Usually, if everthing is within reason, an overclock shouln't lessen the life to where it would make a huge impact. Say from 10 years to 7 years. Will you blame your overvolt when your GP32 dies in 2010? Probably not, you'll just figure that it was old and it had its use. Besides, by that time you will probably have gotten a new gp32 or perhaps some other handheld that emulates the PS2 and Xbox. To most, overclock is an acceptable risk.

Definition I found:
"Electromigration - This is when metal atoms wander into the dividing layers on a microprocessor. It is caused by the combination of electricity and heat. Processors are designed to run within certain heat and electrical specifications. If this occurs to a great degree and enough metal atoms wander off of the lines in a processor, they may permanently ruin the processor by thinning a connection so that it does not work effectively, or even making an electrical connection where one is not intended to be. Overclocking and raising voltage supplied to a processor increases the risk of electromigration. "

Atomicmpc.com Faqs
"Doesn’t overclocking kill your computer quicker?
Not really... Well it DOES shorten the processor's life, but if you do it properly it should still last as long as you need to until your next upgrade. It is VERY difficult just to cause your system to go POP! and not work anymore at all, just from overclocking.

If you are really worried, try and avoid increasing the core voltage much, and get a good Heatsink/Fan (HSF). Overheating and silicon migration are the top killers of CPU’s when overclocking. "

Tom's Hardware.com
"Your CPU could be damaged by so-called 'electromigration'. Electromigration takes place on the actual silicon chip of your CPU in areas which operate at a very high temperature, and can cause permanent damage to the chip. Also, electromigration does not immediately damage your chip. It is a slow process, which more or less shortens the life span of a CPU."

In other words, YES, it will ALWAYS shorten life. Maybe not enough to be noticeable, but that is just luck of the draw and how much you overvolt. The recommended voltage is 1.8+/-.15 volts or 1.65-1.95 per the white sheets for the processor. The designed clockspeed is 133mhz max. So long as it is within these specs, it should last a good long time. Changing the resistor or the pencil mod to up your voltage to around 1.95 volts isn't really overvolting. It's still within design specs. 2.1 would be overvolting. It should probably be safe for 2-3 years at least, but who knows for sure. But if you are using the pencil mod, and have no idea what your true voltage is, you may be running at 2.5 volts at a 200mhz overclock, that is around a 30% overvolt and way out of spec. I'm willing to bet your CPU will die within the year. It is hard to blow a CPU outright unless heat is a serious issue, and for these arm9's, it isn't. Electromigration takes time, but it does and WILL occur. It is proven chemistry and physics. Why else would chips have so many ratings and guidelines and approximate usage hours? Chips will die naturally, even if they aren't used. Electricity and Heat just accelerates natures course to return to equilibrium.

Cheers!

 

[RockOut]

I have a quicck question. I'm not saying that you are wrong, but have you ever really had a computer or a handheld die from overclocking or "overuse"? I mean not from burning it up immediately, but from just having it overclocked for years?

EDIT: I just used the pencil trick today, and in about 15 minutes and 30 pencil strokes later, my 156mhz flu GP32 can go to 192 mhz.

 

[Hooka]

Robster: As a small side note, there is the possibility that OCing your GP32 too high for too long may actually cause an overheating of the batteries (due to drawing more power from them than is recommended) instead of the CPU, therefore increasing the internal temp of the GP32 (and the CPU) to a level that damage may take place. For instance my first GP32, ran 156Mhz, I OC'd it at 156Mhz and played for a couple (around 8 hours straight) the back became hot, it crashed and the screen started to have problems. I noticed that after this I couldn't play for more than about 1/2 hour at even 40Mhz with brand new batteries after waiting a couple days before using it. I sent it back on warranty and got a new one. So it's possible that OCing did damage my original GP32 but this is very inconclusive evidence.

P.S. Have you finished testing that GP32 yet?

 

[bast525]

Please don't spread stories about OC'ing killing or even shortening the life of your GP unless you can prove it. It's stuff like this that keeps people from wanting to try new things. Not that long ago people were dead convinced that you would fry your GP if you overclocked it past 133mhz. Now we know this is not the case. Sure it might wear it out faster over time but really how much life span was in the GP's CPU to begin with? I have an NES that was bought for me as a kid almost 20 years ago that still runs perfectly (minus the need to clean the contacts of the cartridge port every so often), and every time I go to my local DMV I see 20+ year old computers still running on monochrome green 8" monitors, old orignal IBM 33mhz computers that are on 8 hours a day every day of the week.... I'm sure the GP's CPU will last longer than my interest in playing it will, overclocked or not, unless the voltage is WAY high, but as my own testing proved the GP WILL let you know if you're running too high of a voltage, check the other thread. Yeah you can overvolt it to 2.5v and it WILL boot up but after 10 mins or so will lock up and the LCD will go crazy.

There are several guys who have done the pencil mod already and are having no problems. Until someones GP fails due to that mod people should really refrain from trying to put a scare into others talking about how the pencil trick will shorten the life of your system.

 

[Axeman]

I have a quicck question. I'm not saying that you are wrong, but have you ever really had a computer or a handheld die from overclocking or "overuse"? I mean not from burning it up immediately, but from just having it overclocked for years?

EDIT: I just used the pencil trick today, and in about 15 minutes and 30 pencil strokes later, my 156mhz flu GP32 can go to 192 mhz.

​

Yes,

I have. I OC'd my old Celeron 400 to 500Mhz and then ran a batch of video conversions on it. It died the next day.

So basically, I was running it at its highest capacity overclocked, which most of the apps on the GP32 do, and it fizzled. So yes, it does happen.

Basically, the laws of physics state that any chemical reaction doubles in speed for each 10 degrees C rise in temperature. Therefore, even the standard degredatikon that a chip undergoes would be sped up by overclocking.

So there.

HOWEVER: it's alost cvertanly of little impact with the GP32, as there is, I believe, not one recorded case of death through overclocking. All we'd find is that, like mentioned above somewhere, it may die a little earlier (like 8 instead of 10 years or something). So dont worry about it!

 

[bast525]

99% of the time when a computer CPU fries from overclocking it's due to insufficient cooling. I have no idea how this would effect the GP but I think hardly at all since it has NO provisions for cooling and we're not talking about turning up the wick that much anyways.

But here's my example. I have an Athlon XP 1600+, which I believe was originally rated to run at either 1.2 or 1.3ghz. I have had it overclocked to 1.7ghz for over two years now (equivalent of an XP 2000+ according to my bios, that's what it sees it as) and I've had no problems. I've even run the Sisoft Sandra 'burn in' test which is a benchmark test that puts 100% load on the CPU to see if the comp will lock up or overheat. I ran that test for 24 hours straight. Temps climed about 10 degrees Celsius higher than they usually are, but the computer did not lock, and it's worked fine in the two years since. I'm using just the stock factory cooling fan that came with the retail CPU package.

Anyways, this topic seems to be getting confused a bit, half of us are talking about the GP's CPU frying from excess voltage, and half of us are talking about the GP frying from overclocking. They are two seperate things. You could fry your CPU from too much voltage even at 133mhz, whereas many people that have GP's that can OC well past 166mhz with no mods at all, seem to be having no problems after years.

 

[jcabrer]

I guess since I'm responsible for all this, I should comment on whether or not I think overclocking will shorten your GP32s life span.

This discussion brings to mind Arthur Lydiard and Jim Fixx, both have been known as "The Man Who Invented Jogging". At the time of his death when he was only 52, it was estimated that Jim Fixx spent a total of seven years jogging. By the way, he died of clogged arteries. In a bit of irony, studies later proclaimed that jogging can add seven (7) years to your life.

Arthur Lydiard, whose research into running for your health preceeded Fixx by 20 years, lived to the ripe old age of 87.

So what lesson can be gained from all this? Basically, that my GP32 will outlast all others, and that if you wait 20 years to overclock your GP32, it will die of clogged arteries

 

[Zap]

Where di you measure vcore ? Is it really between T7 and T8 ?

​

Yepperz, one is vcore, the other, T7, is gnd.

Well then, after doing the pencil trick my GP overclocks at 224 MHz with 1.4V at vcore.. that's most unusual..