The Source for Mac Performance News and Reviews
Don't forget to check out all the other site features!
Apple G4 Sawtooth CPU Module Speed Settings
By Michiro Isobe
(Updated 6/22/200 for 133MHz Bus Info/Tests)
(Updated 1/27/2001 for Firmware/Timebase info)
(Updated 4/20/2001 for OS X issues with overclocked bus speeds)
(Updated 12/26/2001 for comments on OS X 10.1.2 solving time-base problem)
Disclaimer/Warning: This page is for reference information only. Performing modifications or other work inside your Mac will void your Apple warranty, may cause damage to your computer or result in personal injury. Any modifications to your CPU module may result in permanent damage requiring expensive parts replacement, data loss or personal injury. The resistors on these modules are very small surface mount devices, requiring very delicate soldering skills beyond the capability of most end users. The author and site publishers DO NOT recommend attempting any of the mods listed here.
Dual G4/500 CPU Module Speed Settings: Thanks to "Zack" for the photos of a Dual G4/500 module at this page which shows the Bus/CPU ratio resistor areas. (added 12/14/2000)
6/22/2000 Update: 133MHz Bus Speed Notes: Michiro Isobe wrote that he'd found a way to run the G4 at 133MHz bus speeds.
The following is from his entry in the Rate Your CPU Upgrade Database:
"Power Mac G4/450 rev.A
Stock 450/225/100MHz bus
Running 466/233/133MHz bus
By using two 128MB PC133 CL2 DIMMs (http://www.crucial.com/store/partspecs.asp?imodule=CT16M64S4D7E), 133MHz bus can be running. The timebase problem in my previous report [this article-Mike] was solved by firmware tweaking.
The MacBench 5 results show only CopyBits operations are improved (by ~40%) with the bus overclocking.
Michiro also sent Stream, Memory Watch and GaugePro test results.
(Note: My G4/500 (100Mhz bus) reports 234 MB/sec (max) w/GaugePro 1.02/OS 9.04)
Memory Watch Results:
" Dear Mike,
Happy Holidays!. Today, I found that Mac OS X 10.1.2 has fixed the time-base problem of bus-overclocked sawtooth. Although Apple System Profiler has shown fixed 100MHz bus clock, the clock works normally and you can fix it by Open Firmware tweaking.
(Original info added 1/27/2001) Michiro Isobe wrote:
1. G4 (Gigabit Ethernet) and G4 Cube have no time-base problem.
2. The method of "firmware tweaking" for old sawtooth models.
Old G4 models (before G4 Gigabit and Cube) have some time-base problem, which the clock run faster (or slower) when the bus clock changed faster (or slower) than default of 100MHz. This is caused by timing routines of Mac OS, such as TickCount, Microseconds and UpTime, calculate time using actual bus clock and some ##fixed## values supposing 100MHz (99.6MHz exactly) bus. So, we must change these constants when change the bus clock.
I changed these constants using "NVRAMRC" OpenFirmware script. The NVRAMRC is stored in nvram and is performed at boot sequence.
How to edit and store the NVRAMRC for 133MHz bus
(1) Boot by holding down "Cmd" + "Opt" + "O" + "F" keys to enter the OpenFirmware terminal.
(2) Enter the NVRAMRC editor. (<return> shows you should type "return" key)
nvedit<return>(3) Type following script exactly (including space).
" /" select-dev<return>(4) Type "Cntl" + "C" to exit NVRAMRC editor.
7efdc44 " clock-frequency" get-my-property 2drop !<return>
" /cpus/PowerPC,G4@0" select-dev<return>
1fca0554 " clock-frequency" get-my-property 2drop !<return>
7efdc44 " bus-frequency" get-my-property 2drop !<return>
1fbf711 " timebase-frequency" get-my-property 2drop !
(5) Save the script
nvstore<return>(6) To enable the NVRAMRC, type this command.
setenv use-nvramrc? true<return>
(7) Reboot with new timebase constant.
If you fail to edit the NVRAMRC or need to return to the 100MHz bus, you can clear the changes by resetting the nvram, booting by holding down "Cmd" + "Opt" + "P" + "R".
This TIL article explains how to reset the PRAM. (Another TIL has more details on NVRAM resets via open firmware, see commands down the page there.-Mike.)
In many cases, 133MHz bus does not work. If it seems to work, it is stable with very limited conditions.
(NOTE: Michiro later noted that at 133MHz, only 2 dimm slots were usable (even with 133mhz RAM.) Also see below for comments on problems in OS X with overclocked bus speeds. [Solved by OS X 10.1.2 update he later said.] Remember that 133MHz bus speeds require PC133 RAM *and* a mod to the CPU module to lower the bus/cpu ratio - otherwise the CPU will be so overclocked it will not boot. (I.E. - a G4/450 module is set to 4.5x ratio for the original 100MHz bus - at 133MHz the CPU will be trying to run 600MHz, which will not even boot. And for 450 and 500Mhz G4s - there is no setting that for those CPU speeds exactly, and the closest settings run 466MHz and 533MHz, which may not be reliable since it's overclocking the CPU beyond its rated speeds.-Mike)"[updated] Mac OS 9.1 resolves the timebase problem of all sawtooth models without Open Firmware tweaking. The only problem is Apple System Profiler still recognizes the bus clock to be 100MHz and displays incorrect machine speed.
(Mac OS X) Mac OS X has the same time-base problem with the original sawtooth system. Unfortunately, OS X detects Open Firmware tweaking to "timebase-frequency" and refuses to boot. Currently no solution is found.
[UPDATE: Michiro Isobe wrote that OS X 10.1.2 solved the time-base problem present with earlier OS X versions.]
Michiro Isobe "
Bus Ratio, Core Supply, L2 Signal Level and Bus Clock Details
of the Apple Sawtooth G4 System:
(1) Bus Ratio Controls:
- PLL_CFG R7
- PLL_CFG R9
- PLL_CFG R11
- PLL_CFG R13
The PLL_CFG code is shown at Table 14 in the following PDF document.
(Update - a reader noted the link is now
(2) CPU Core Supply Voltage (Vdd):
- VID4 R37
- VID3 R38
- VID2 R39
- VID1 R44
- VID0 R78
The VID code is the same with Yosemite based system. Please consult the voltage identification table of this PDF document http://www.cherrysemiconductor.com/product/PDF/CS-5165PDF.pdf
(UPDATE: CherrySemi was acquired by ON Semiconductor, so the URL above is now invalid.
Default setting was 2.1V
(3) L2 Signal Level:
L2 signal level was specified to be 2.5V by the PPC7400 design docs. But it can be adjusted a little by changing a register at the back side of the CPU daughter card.
(See Fig1.jpg and Fig4.jpg for photos identifying these)
L2OVdd = 1.260 x (R51 + R52)/R51
Default: R51 = 124 ohm, R52 = 124 ohm.
(4) Bus Clock Configuration:
But, I found that the bus overclocing also make the time base run fast. So this modification is nonsense. If you try the bus overclocking, you must find the source of time base and supply a clock independently.
- FS0 R435
- FS1 R434
- FS2 R433
- SSON R432
There is frequency table in this PDF document. sg500.pdf [link revised 1/20/2001]
- G4/7450 CPU Module Overview and Speed Settings (again by Michiro Isobe)
- Dual G4/500 CPU module Bus/Ratio Control Resistor Locations
- Dual G4 CPU Module Swap on Single G4 system
- Dual G4/500 vs Single G4/500 (Quicktime/iMovie performance tests)
- Other G4 Systems related articles (drive upgrades, fan noise reduction, performance tests...)
- A Peek at the Sawtooth CPU Module (Apple G4/450 AGP system)
- Comparing the G4 to the G3 (specifications/features)
- G4 Upgrades in B&W G3 vs Sawtooth G4/450 AGP performance (included in my XLR8 G4/400Z Review and Newer Tech G4/400 ZIF Review)
- Apple Developer Notes on the G4 systems
- For more info on the PowerPC 7400 (G4) CPU, see Motorola's PowerPC 7400 (G4 CPU) page (Includes a JPEG image of the die) and their PDF file on the G4.
- For a comparison of the G4 vs G3 CPU chip designs, see this page.
- G4 CPU Upgrade Reviews (includes comparisons to G4/AGP Systems).
= Back to the XLR8YOURMAC.COM =