[mdavis]

Hi all. I overclocked my iBook 500/66 to 600/100 a long time ago and I've had my G4 for a while but now i want to overclock it, too. I have 2 CPU thermometers so I can monitor the temperature, plus the machine automatically does to sleep when it gets too hot anyways. (I know this because I have done a lot of fan/airflow modding and sometimes configurations don't cut it, the machine gets too hot, and it goes to sleep to cool off). Anyways, I used the PLL config tables from here:

http://www.xlr8yourmac.com/systems/g...PU_Module.html

to increase my processors' clock speed. The problem is, I can't seem to get the core voltage correct. (I THINK that's it...) I set up the PLLs for 1.25Ghz for both processors and set the core voltage to 1.95V (default is 1.85) but on booting, the machine gets almost to the login portion of booting (near the end) but then kernel panics. Every other voltage possibility (both with 1.16 and 1.25Ghz PLLs crashes right away at startup. Usually I get that weird "You computer needs to be restarted - hold the power button" message but sometime it's just a kernel panic and sometime it just locks up. This always happens about 3-5 seconds into the "Grey Apple" startup screen right at the beginning of startup. As I said above, ONE TIME I was able to get the machine to boot almost all the way through the "Aqua bar" boot screen, almost into the login portion but it still crashed. And again (re-capping for clarity) I've tried a number of core voltages at both 1.16 and 1.25Ghz.

I should note that the 1k Ohm resisters I used to configure the PLL registries (4 additional resisters are needed to bring the machine to 1.16 or 1.25) I nabbed from a Mac LCIII (won't miss them) and they are slightly physically bigger than the original resisters on the G4's PLL registry. When tested, the original resisters (from the G4 PLL) come up at about 990-995 Ohms and the new resisters (from the old LCIII) show up at about 1005-1015 Ohms) Seems within acceptance... Now I'm no expert, but it seems to me that as long as the PLL resisters are CLOSE the to correct resistance, it should work (I don't believe the CPU speed is determined analogly, but digitally - meaning a close enough resister will set a given PLL to 0 - they can only be set to either 1 or 0, not in the middle) But then again, I don't know... Both the original G4 resisters and the new ones are labeled "102." I should also note that the resisters on the G4's TOP PLL registry look different than the resisters on the BOTTOM PLL registry.. The top one's are black and the bottom ones are light blue...

Well if anybody has any information that might help I sure would appreciate it. Thanks!!!!!!!!

PS. After restoring the PLL and VID to default, the machine works just fine (I'm using it right now)

[G4ME]

wow thats hard core, i never had the balls to break open my CPUs and do some soldering to get and extra 50 MHZ, as some people have done. but there will always be the people who enjoy messing with crap like that.

i hope you can get it working, talk about saving some cash.

Also i would try and use the correct resistors, my bet is that they are very finiky with the voltage.

[Spliffdaddy]

I think you found the limit of the CPU core.

Even if you did manage to get it to boot the OS, it's gonna come down hard the first time its under a load.

Increasing the core voltage is probably not going to stabilize the CPU. As a rule of thumb, if it won't load the OS while overclocked at default voltage - then adding more voltage isn't likely to help.

Lowering the temperature of the CPUs can work miracles, however, and this is what you should concentrate on doing. As an example, my 2.53GHz Pentium4 will run stable at 2.85GHz provided the CPU temp never exceeds 114F. I have to drop back to 2.66GHz on warm days. A few measly degrees is all it takes to destabilize the machine and result in a 200MHz slowdown. (yes, I own the world's worst 2.53 P4 in terms of overclocking).


The fact that you aren't overclocking the memory bus tells me that the RAM, chipset, and PCI-based hardware aren't holding you back - it's the CPU(s) crapping out.

[Cipher13]

Originally posted by Spliffdaddy:
I think you found the limit of the CPU core.

Even if you did manage to get it to boot the OS, it's gonna come down hard the first time its under a load.

Increasing the core voltage is probably not going to stabilize the CPU. As a rule of thumb, if it won't load the OS while overclocked at default voltage - then adding more voltage isn't likely to help.

Lowering the temperature of the CPUs can work miracles, however, and this is what you should concentrate on doing. As an example, my 2.53GHz Pentium4 will run stable at 2.85GHz provided the CPU temp never exceeds 114F. I have to drop back to 2.66GHz on warm days. A few measly degrees is all it takes to destabilize the machine and result in a 200MHz slowdown. (yes, I own the world's worst 2.53 P4 in terms of overclocking).


The fact that you aren't overclocking the memory bus tells me that the RAM, chipset, and PCI-based hardware aren't holding you back - it's the CPU(s) crapping out.

Agreed... concentrate on cooling, see where that gets you.

Did you have any luck at 1.16GHz? Either way, a few resistors won't cost more than fifty cents... go buy some new ones, try that out, and if you have no luck, you've wasted less than a dollar, and you can concentrate even more on cooling. The LCIII resistors aren't *too* far out, but you don't know what kind of tolerance for error there is...

[Spliffdaddy]

Before you spend money on cooling the machine...

open the side of the case and sit a big box fan on the floor aimed at the CPU heatsinks.

Or

take the computer outside on a chilly day.

The idea is to see if a few degrees of temperature drop will help stabilize the processors.

then, if you can successfully load the OS at 1.25GHz - AND - the same level of temp reduction can be duplicated indoors without too much fuss (larger/extra cooling fans) - you have succeeded. Run a CPU-intensive task such as the SETI client for a few hours ( 2 clients on the dually). This will keep the CPUs hot long enough to check their stability. Generally, if you can run an overclocked machine for several hours (preferably overnight or longer) at full load without problems - it could be considered 'stable' and it should live a relatively troublefree life. Never any guarantees, though.

The heatsinks on the PowerMacs are large and very efficient but appear to be choked by lack of airflow. Adding a decent fan to the heatsink fins and perhaps another fan to pull in fresh air from outside the case should lower CPU temps at least 10F degrees....maybe more.

Be careful with the core voltage! This is the stuff of last resort. Heat is your enemy and raising core voltage adds more heat than it does stability. If you were REAL CLOSE to having a stable overclocked machine, I'd say it's safe to increase the core voltage up to 10% over default. Keep in mind that raising the core voltage 10% will likely pull the CPU temperature up 15%.

Think of a CPU as a lightbulb. The more voltage you feed it, the brighter it gets. If you keep raising the voltage, at some point it's gonna make a 'pop' sound and stop working.

Good luck, and dont mess with the core voltage until it's your last resort.


edited to add:

The current G4 cores are lucky to hit 1.25GHz. I'd bet the yield Motorola is/was getting on the 1.25's is pathetic. There is probably little chance of finding G4s rated less than 1.25GHz that will meet or exceed 1.25GHz. I'd also bet that the G4s at OVER 1.25GHz are based on a refined core that's not shared with the 1.25GHz and lower G4s. Remember, Motorola is rating the core speed based on its intended use in a personal computer...and the expected temperature of the CPU in those conditions. Let's assume for sake of argument that the 1.25GHz G4 is expected to experience a sustained maximum operating temperature of 140F degrees when equipped with an approved cooling device. Maybe Motorola could offer this same G4 processor marked as 1.5GHz IF they could be assured that the CPU temp would never exceed 120F degrees. They have to take into consideration the ambient room temperatures of everything from perhaps 50F degrees up to un-airconditioned attic rooms at 80F degrees. They have no clue where it will be used and at what temperature the surrounding air will be. Maintaining a CPU temp of 120F and using 80F air to do it might be difficult. But YOU (the overclocking end-user) could make sure the temp never exceeded 120F degrees, right?

Big overclocks are more likely to occur on AMD or Intel processors because they both market a broad range of CPU speeds that all share the same (or similar) core. The same batch of CPUs could be labeled and sold as 1GHz up to 1.8GHz for example. In this case, the smart overclocker would choose the slowest available retail speed knowing that the core is identical in the much faster model. This is how the 533MHz (coppermine) Celerons could overclock to 1GHz - nearly a 200% overclock. To make matters even more interesting, there is no way to change the multiplier on Intel processors (since the P2 days). All overclocking is done by increasing the frontside bus speed. The FSB is usually tied-in with the AGP, PCI, and ISA busses - and can result in a lot of out-of-spec speeds and wierd problems with hardware.

[Eug]

I'm still surprised that Motorola/Apple can get 1.42 GHz out of these 0.18 u chips in numbers good enough to base an entire PowerMac line around them.

[gizzard]

The 1.25 and 1.42 GHz G4 chips are manufactured using the 0.13 micron process.

[Eug]

Originally posted by gizzard:
The 1.25 and 1.42 GHz G4 chips are manufactured using the 0.13 micron process.

??? Sez who?

[mdavis]

I thought I had email notification on, but I guess not... I'd say this is a pretty good turnout. (BTW, I'm a MacNN old-timer but I forgot my old l/p) Well lets see... Well thanks for the help all, but I have to disagree with the temperature theory. I have thermometers directly on top of each CPU core with thermal paste so I get a pretty dead-on accurate reading with almost no delay, and the machine is crashing before it even hit's 20C (and at 1Ghz it idles at ~40C) I suppose it's possible portions of the CPU are heating up too quickly but I doubt it... And about the resisters, I doubt they're the problem. ~1K Ohm will set each bite to 0. So it's either 1 or 0, and I doubt the machine would even boot if one CPU's PLL was different then the other...

Thanks for all your help everyone but keep it come'n! (I'll also post my fix if I get it working!)

-Morgan

[Catfish_Man]

Originally posted by gizzard:
The 1.25 and 1.42 GHz G4 chips are manufactured using the 0.13 micron process.

No they're not. The .13 micron G4 has a 512k L2 cache, which the announced PowerMacs don't have (source: Motorola Press Release announcing the 7457).

[mdavis]

OK I got 1.16Ghz working. The stock 1Ghz CPU temperature was about 40�C (104F) and now the machine is running at 54-54.4�C (129.2-129.92F) constantly (no matter what I'm doing) I don't really know how I made it work... I'm at the stock 1.85V core voltage; I believe I tried that before but I'm not toally sure... I probably forgot to, that's all there was too it. haha. Well, I added some fresh thermal paste to the top of the cores and I also added one of the 2 extra PSU fans I had (from replacing the original PSU fans with quieter ones/lowering voltage) to the inside of the front air intake vents to help airflow through the machine (and to the CPU heatsink) as I have 3 HDs and they kinda block airflow... 54� is pushing it so 1.25Ghz is out of the question (for now... we'll see about liquid cooling) If you have questions I'm glad to answer them. [email protected]
Btw, I'm thinking of creating a Mac overclocking site.. Could be a good resource.
Laters - Morgan

[JB72]

Good deal Morgan!

I'd like to take my dual gig to 1.16 in the future, but not until I do a bunch of other mods/upgrades. I'm not as ballsy as you.

[mdavis]

Hey guys it's me again... I eventually put my CPUs back to stock 1Ghz speed because I was getting sick of fan noice. I just got a Verax cooling kit. It's amazingly quiet. However I heard it actually keeps the cpus cooler...So I opened my machine up to change it back to 1.16 but couldn't get it quite working consistently. I was putting everything back to 1Ghz when I lost a top side PLL resister marked "018." Does anybody know what resistance this is and where I can get one? I also started a new thread with some additional information here: http://forums.macnn.com/showthread.p...+018+resisters

Thanks!!!

[jamil5454]

If you were really extreme, you could buy a minifridge and mod it to put your tower in that. That might solve a few cooling issues...

...or you could follow "Tom's Hardware Guide"'s lead and cool the cores directly with liquid nitrogen. They got a P4 to run at 5ghz that way.

By the way, AMD has been running .18 cores at over 2ghz for a while now, but thats a CISC processor so it's probably not a good comparison.

[Catfish_Man]

Originally posted by jamil5454:
If you were really extreme, you could buy a minifridge and mod it to put your tower in that. That might solve a few cooling issues...

...or you could follow "Tom's Hardware Guide"'s lead and cool the cores directly with liquid nitrogen. They got a P4 to run at 5ghz that way.

By the way, AMD has been running .18 cores at over 2ghz for a while now, but thats a CISC processor so it's probably not a good comparison.

Actually they're all RISC internally these days (and some of the RISC chips are looking suspiciously CISC like). The primary reason AMD clocks higher than the G4 is because the pipeline is about twice as long. They've hit 2.6GHz on .13 micron SOI, so pretty similar to the G5.