PDA

View Full Version : Who said you can't?



kim jong il
14th August 2004, 04:24 PM
Recently I posted requesting info about my processor type (G4 400 - (PPC 7400)). After some sleuthing I guessed that my G4 400 (daughter card's CPU) was made by IBM not Motorola. Some of you may be aware that IBM had a superior manufacturing process and could supply fast chips when Motorola were stuck at less than 500 MHz (cache addressing reasons). Version 2.8 of the processor addressed these issues in late 1999...... My machine is Nov 2000 (PPC copper process v2.8)

I'm guessing here that as my G4 CPU core did not say it was manufactured by motorola so it was probably someone else........ IBM don't admit that it was theirs but they did supply chips to Apple using their superior manufacturing process and this must have been one of them.

Clockchipped it to 450 MHZ. Perfect.....

500 MHz? Bongs once then again half heartedly. Won't boot. No surprises here. SO, I boost CPU core voltage from 2.1 V to 2.5 V and restart. PERFECT. CPU core changes temperature faster than before but is perfectly stable at 100% duty cycle. (not more than 8 C warmer than case/HD temperature - (around 40 C)) A full 25% faster than standard.

Now wait for my 550 MHz test results. I still have half a volt to play with and the L2 cache claims to be good for 320 MHz (or 660 MHz processor speed at 2:1 ratio). I just have to add cooling before I try the next step.

I just had to tell someone.........

kim

kim jong il
15th August 2004, 01:43 PM
I would like to add that this modification is not for everyone. (I won't include the warranty warning as none of these machines should be covered by warranty anymore - if they were it would be VOID)

If you have a pre december 1999 G4 and the processor is made by Motorola, this modification will NOT work (L2 cache addressing problems). Likewise, on a PCI G4 it will NOT work (probably) also. Yikes? Seriously think about forgetting this. Sawtooth? check out the CPU core ID first. If you have a Motorola it probably will not work unless it was manufactured after December 1999 and then only if it is processor version 2.8 or higher. NON Motorola? Consider it.

To have any chance of success, you WILL need an (AGP Gigabit ethernet motherboard) G4 PPC 7400 version 2.8 (or higher) and preferably the processor will NOT be made by Motorola. IBM had the superior manufacturing process but were forced by contractual obligations not to sell faster chips than Motorola could manufacture (i.e. 450 MHz). It would appear that they adhered to this ruling by selling underclocked chips that were, in reality, capable of speeds up to 650 MHz (100 MHz BUS) or 667 MHz (133 MHz BUS). NOTE: The IBM process was yielding these in December 1999.

DO NOT ATTEMPT THIS UNLESS:

a. you are prepared to fry your processor
b. you have a good idea what I am talking about
c. you believe you know enough not to kill your machine or are prepared for this potential eventuality.

If you need to ask for further details you probably should not attempt this either (but you are welcome to email or PM me about this modification).

kim

Disko
15th August 2004, 01:47 PM
Sounds great. :)

How about a step by step "how to" guide with pics? I'll put it up in the article's section if you'd like. :)

kim jong il
15th August 2004, 05:13 PM
Hey Disko, I'll work on it (although the original post was boasting, NOT a try this). My concern is that someone with a Motorola CPU will attempt this and cook their computer (the daughter boards are the same)(CPU ID?: Motorola say 'Motorola'; IBM say nothing, but you will need to physically see the CPU core to determine this). Motorola CPU's have been reported to fail within 12 hours of a 50 MHz speed up, so be warned. This speedup was a full 100 MHz or 25% (REAL speedup/performance boost, not like BUS overclocking which is another whole story with these MB's and CPU's)

This modification was only undertaken after considerable research suggested that I had an IBM CPU and that at the time of manufacture IBM were achieving and exceeding the speeds described in the post above. Fiddling with the CPU core (voltage or BUS multiplier) is not for the faint hearted and an extra problem with the daughterboards (with the IBM CPU daughterboard) is that none of the board components are labeled (e.g R8, R73, whatever) so the person attempting this must actually KNOW what they are attempting (I found no resources on the web that categorically referred to my machine- I used cross referenced documents pertaining to CPU core voltage, BUS multipliers and deductive reasoning).

Wrong CPU or inadequate knowlege will/could probably end in TEARS. I could produce a simple step by step article on how to kill a PCI graphics G4 ('Yikes' OR 'Sawtooth', but probably not 'Mystic')(in the wrong hands that is), but I would prefer not to. A simple step by step is easy but some people will attempt it on machines that are not, and never were or will be, capable of the speeds described.
NOTE: due to the exchange rate in 1999, 2000, many people may have bought the low end multi processor capable G4 'gigabit ethernet' OR 'mystic' motherboard machines which seem, in Australia, to have the IBM CPU's. The above pertains only to these machines. The upshot being that if you have a single or multi processor G4 7400 (v2.8 or higher)(with above motherboard manufactured during late '99 to end of 2000, you can almost be guaranteed to get an extra 50 MHz without voltage adjustments. 100 MHz with (IBM only)

I'll work on the article, but it will have to contain multiple warnings, e.g. Do not attempt/do this until you are sure of.......... etc.

The upshot is, in a nutshell, that you have to be fairly sure you have the IBM processor.

cheers, kim

Byrd
16th August 2004, 06:56 PM
Kim, what site did you go off to perform the voltage modification?

And have you any pics of the mod you did to the chip itself, along with describing how you went about it?

Thanx :)

JB

kim jong il
16th August 2004, 08:28 PM
Hey bYrd, I'll put something together starting tomorow (Tuesday 17/4/04) (EDIT: try Tuesday 17/8/04) It will have to be progressive. I'll start with some basic theory (complete with relevant links) follow through with the results of my deductions about my particular machine based on this. My capacity to take pictures of any decent quality are a little limited, but I'll do my best. Finish up with methods and materials and results. You may be amazed that only one link had to be removed and three added for the complete procedure.

cheers, kim

(I'm back at work this week after being on holiday for 5 weeks so I have a little less time on my hands than recently)

Ozi
16th August 2004, 11:32 PM
Hi kim! Sounds like an awesome speed bump! 25% is huge, and 40 is nothing for a CPU. The G4 in my pBook gets up to 63, and then the fan keeps it there.

Also, the family 4MP camera could be lent to you to do the pics. If you want, I am more than happy to come over with it, and also have a perve on your overclocked G4. :) I am free from 430 tomorrow afternoon, if you want to do it then. Give me a call if you want me to come around, home or mobile number. :D

~Ozi

purana
17th August 2004, 08:37 PM
Originally posted by kim jong il@Aug 16 2004, 08:28 PM
(Tuesday 17/4/04) It will have to be progressive.
You mean 17/08/04 I am sure :)

jameso
17th August 2004, 08:53 PM
If you go too..

http://www.xlr8yourmac.com/G4ZONE/sawtooth...hCPUdesign.html (http://www.xlr8yourmac.com/G4ZONE/sawtooth/SawtoothCPUdesign.html)

and look halfway down the page there is a guide on overclocking you g4 in firmware. I've done it and had no problems but i also didn't really notice a speed increase. Kim can probably tell us more because he mentioned it above. do you know anything about this Kim?

kim jong il
18th August 2004, 06:20 AM
Hey jameso. Hopefully you are genuinely interested otherwise this is fairly pointless, however, here is an explanation of things as I understand it. I could be completely wrong but i don't think so. Furthermore I do not want to get in a bitchy debate with anyone on the factors that infleunce speed in this particular machine (or any for that matter) in this thread. There are loads of resources that describe the factors that affect any machines performance in excruciating technical detail. As a result I will not do this. However, be warned, what follows is a simplification.

CPU overclocking is quite different to Bus overclocking. With CPU overclocking the processor is physically performing more functions, whereas with Bus overclocking you are speeding up the rate at which various parts of your machine communicate with each other (Bus referring quite simply to the actual physical electrical data path). In this case the open firmware hack that speeds up the system Bus just increases the speed at which the CPU talks to the motherboard (more or less). As the system Bus is not any kind of significant bottleneck in the machines referred to it does not result in any meaningful performance boost. Data gets shifted between buffers faster but still only gets processed at X speed. Sort of pointless really but a semi worthwhile academic exercise in some ways. (Bus overclocking may improve benchmark scores marginally but this improvement in performance is a mathematical illusion)

Some Xbench (1.1.3) scores to give a basic idea of the difference CPU clock speed makes are shown below (Want to find Xbench? Google it. Donationware) Benchmark: G4 800x2 = 100

A standard AGP Gigabit ethernet G4 motherboard: (bigger numbers are better)

G4 400x1: 46-54
G4 450x2: 67-78
G4 500x1: 73 (me)
G4 500x2: 71-82

The Xbench calculated CPU average performance increase was 24% (G4 400 to 500 MHz speedup). Some CPU functions were only 20% faster, while others were a whopping 32% faster. I have given a range of values as some people will insist on running benchmark tests with background tasks running. It means that results need to be interpreted and absolute comparisons should be examined closely. On overall system performance, my single G4 500 gives me about the same performance as a dual 450. Where it loses out in multithreading limitations it makes up for in speed.

If there is anything you would like clarified, just holler.

cheers, kim

kim jong il
18th August 2004, 07:14 AM
Originally posted by iodine+Aug 17 2004, 08:37 PM--></div><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td>QUOTE (iodine &#064; Aug 17 2004, 08:37 PM)</td></tr><tr><td id='QUOTE'> <!--QuoteBegin-kim jong il@Aug 16 2004, 08:28 PM
(Tuesday 17/4/04) It will have to be progressive.
You mean 17/08/04 I am sure :) [/b][/quote]
Nice catch. I&#39;m going to post progressively into the past............ (actually how clever is that?)

In reality? Like I said/implied; first night back at work after 6 weeks off (5 AL + 1 LWOP) and I was buggered. This is my excuse and I&#39;m sticking to it.

emac_man
19th August 2004, 09:47 AM
Now, if you can clockchip a 2.5Ghz dual G5 to 3.2Ghz (or at least my eMac 1Ghz out to 1.25 ;) ) I&#39;ll worship you as a god ;)

kim jong il
19th August 2004, 11:18 AM
Originally posted by emac_man@Aug 19 2004, 09:47 AM
Now, if you can clockchip a 2.5Ghz dual G5 to 3.2Ghz (or at least my eMac 1Ghz out to 1.25 ;) ) I&#39;ll worship you as a god ;)
You would only need to remove three resistors to achieve 1.25 GHz (this is the best modification type as it is quick and clean). I could even do this for you for a small fee, everlasting goodwill or your firstborn.........

This of course would invalidate your warranty, but people have run 800 MHz emacs out to 1.33 GHz without CPU core voltage modification, indicating that not only do they use a low power G4 7445, but that it is dramatically underclocked to ensure low heat output and stability in all operating conditions. At this speed (1.33 GHz) peak power dissipation is 26 watts, which is quite alot but it should not even really need supplimentry cooling although some might add this for peace of mind (e.g. a tiny pentium II fan would be ideal). 1.25 and 1.33 GHz speeds are very possible at extremely small risk to the machine. (even if you have a high temperature lock up, which is improbable, the CPU will be undamaged and will just need to be run at a lower speed or supplimentry cooling added.)

cheers, kim

EDIT: The article Atomic refers to BELOW is a beauty (one of the best I have read). Micro soldering is not for everyone however. You would not believe how easy it is to damage the ultra fine circuitry on the motherboard until you actually do it

Atomic
19th August 2004, 11:23 AM
Originally posted by emac_man@Aug 19 2004, 09:47 AM
Now, if you can clockchip a 2.5Ghz dual G5 to 3.2Ghz (or at least my eMac 1Ghz out to 1.25 ;) ) I&#39;ll worship you as a god ;)
You can proclaim yourself as God and overclock your eMac yourself. A guide to doing this is available here&#33; (http://www.lbodnar.dsl.pipex.com/eMac/eMac-upgrade.html)

zen
24th August 2004, 12:24 AM
Overcloking an Apple CPU&#33;? :o

never though this is possible&#33;&#33;&#33; :lol: GOOD SHITZ&#33; AND KEEP US POSTED WITH YOUR PROGRESS B)

kim jong il
24th August 2004, 05:57 AM
Originally posted by zen@Aug 24 2004, 12:24 AM
Overcloking an Apple CPU&#33;? :o

never though this is possible&#33;&#33;&#33; :lol:


Certain macs are more amenable to modification than others. Overclocking your powerbook, for example, would be an extremely bad idea. Possible and do-able but still a really bad idea. Early G3&#39;s and G4&#39;s are good candidates (early G3&#39;s and Motorola G4&#39;s should generally be restricted to a maximum 50 Mhz increase)

I will keep people posted. I have promised to write an article about the above mentioned modification but have been a little pre-occupied the last week or so but should have time this weekend (hopefully)

cheers, kim

zen
24th August 2004, 08:03 AM
I&#39;ve only just joined the mighty Apple world not long ago, this thread has sertainly changed even more about how I view the Apple machines now.

This is just an incredible experience (the mac experience that is)