Perhaps looking at improving cooling would dissipate the heat and increase performance.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Overclocking the G5 CPU
- Thread starter simie
- Start date
- Sort by reaction score
With the Power Mac G5, this isn’t always practicable.
To wit: with my A1047, I have an unfixable issue with the main board’s memory controller heatsink, part of the U3/Backside complex, in which it runs abnormally hot.
Discussions on MR forums and elsewhere have indicated this is related to poor/failing solder points on the main board around this complex. I only learnt this after I finished a complete disassembly of my G5 — all of it, including the power supply — and clean-out, followed by all-new Noctua thermal paste on said heatsink (along with the daughtercard heatsinks for each processor). None of this made a material difference in the high temps of the complex, not even when at a complete idle.
So long as I still have this G5, and so long as I can’t find a known, working main board without this fault, my options are to either run at reduced power mode, or to run as little as possible as full power mode. An additional sacrifice I’ve had to exercise is to run with only half of the RAM modules in place, as running with all eight pushes that complex into far too high temps and kernel panics (which triggers spontaneous reboots, given the way I’ve set it up).
Anyway, keep this in mind as you theorize around the G5 architecture.
I think that it is a shame that you have to run your Mac this way because if this fault, I would expect the solder points to be of a higher quality than what they actually are. There was an issue Nvidia and thier poor soldering some years ago and the courts made them repair or replace faulty components.
I would like to think that mine at the moment is free from this fault.
Automatic mode is available to Linux and Mac OS X on the G5 where it will automatically switch between Reduced Mode and full power when the load on the system demands.
You can also set reduced mode/performance mode in Open Firmware, but that setting is only good for using reduced mode while using Open Firmware, a good idea if you need to spend a lot of time booted in OF.
By default Desktop Mac's boot in full power mode, not an issue when you boot straight into the OS, because the OS power manager will take over right away anyway, but in OF this can cause the fans in ramp up to "Jet takeoff mode". Loud and annoying when you are trying to use Open Firmware for longer than a few minutes.
Sadly you can't overclock the G5 using OF, only switch between reduced and full power mode using bus slewing.
Remember Apple mostly used the bus speed to set the CPU. Most G5's just use a 2x multiplier of the bus. Except the 600Mhz bus G5 1.8Ghz and the iMac G5 with the 600Mhz bus.
So figuring what sets the bus is the key to overclocking the G5, and to figure the multiplier you need to compare the 900Mhz( 1.8Ghz 2x ) bus G5 CPU to the 600Mhz( 1.8Ghz x3 ) bus G5 CPU.
Last edited:
I dunno but I've been hacking to get my 4GB sticks working (and btw Win11 terminal sucks but no macports for m1 ventura to install telnet so no choice).
And on OF:
dev /smu
words
EDIT: Nah it seems you are right - there is only some commands to extract it I think.
Last edited:
It is what it is, and I hope your G5 continues to run smoothly for many more years to come!
That said, the G5 was needlessly complex from the outset, and this is ultimately what befell PowerPC use on the Mac side.
I mean, that G5s needed a discrete PPC 405 CPU just to power on the system, to insisting only IBM be involved in chip design and not Motorola/Freescale under the still-extant AIM partnership, underscores just how many things could (and did) go wrong with Power Mac G5s.
For as much as I love having a G5, when one pans back and sees how Apple raced ahead to sell a rushed, 64-bit product, only to backtrack down to 32-bit in the hasty move to Intel, suggests the G5 test case was an example of “just because it’s technically feasible to premiere 64-bit desktop computing doesn’t mean it’s technically prudent just yet to do so.” It’s frustrating, because that rush ultimately tripped up Apple when trying to publicly promise ever-faster processor speeds for the G5, when doing so was ever-more inefficient (i.e., heat and power usage).
Even Low End Mac’s overview of the G5 line reveals how, even when the system was still within AppleCare, one of the most frequent components to fail on every version was the main board:
It’s frustrating from a historical perspective, because had Apple just stuck with pushing co-operation in the strained, but still extant AIM alliance instead of picking favourites (Apple — the “A” in AIM — were that linchpin), the run-up to reliable, 64-bit computing on PowerPC could have happened more organically, more reliably, and with the benefit of cooler-running, multi-core, 64-bit processors (viz., PA Semi’s PWRficient) across Apple’s entire line — all without throwing in with Intel (and at the cost of dawdling in 32-bit land a bit longer as penalty for that architectural switch).
But where PowerPC, under a terminally-divided AIM (no thanks to Apple, under Jobs) failed in the G5 was to deliver the front-end performance without all the power and heat consumption. Jobs was an impatient man — oftentimes a net benefit for product innovation, but at other times with the shortsightedness of shooting his own foot.
Now I’m going to open this wooden soap box I’ve just stepped down from, fish out a bar of soap, and go clean up in a shower.
I mean there was this guide for dumping the rom http://macos9lives.com/smforum/index.php?topic=4295.0
I would reverse it - if I know the format - I've experience only with user space binaries.
And beside there should be some way to attach debugger
AHT link and software
github.com
GitHub - upekkha/AppleHardwareTest: Apple Hardware Test Download Links
Apple Hardware Test Download Links. Contribute to upekkha/AppleHardwareTest development by creating an account on GitHub.
github.com
Last edited:
I posted earlier an article from BBoettjer, talks about 12C EEPROM
www.st.com
This website will help identify the EEProm.
Here is a guide to reading and writing to the device.
deepbluembedded.com
So, if this information is correct, it would be possible to copy off the data of the EEPROM and modify it and reprogram the device and test the processor speed. YOU would have to analyse the data that was read from the device carefully, before attempting to modify the data that has been copied.
Standard I2C EEPROM - STMicroelectronics
Discover I2C EEPROM products covering a large scope of applications, such as wearables, mobile devices, smart city or industrial automation.
www.st.com
This website will help identify the EEProm.
Here is a guide to reading and writing to the device.
Reading And Writing Serial I2C EEPROM With PIC – DeepBlue
I2C EEPROM Code Example Project With PIC Microcontroller. Read And Write I2C Serial EEPROM. Interfacing Serial EEPROM With PIC.
deepbluembedded.com
So, if this information is correct, it would be possible to copy off the data of the EEPROM and modify it and reprogram the device and test the processor speed. YOU would have to analyse the data that was read from the device carefully, before attempting to modify the data that has been copied.
Last edited:
lc04A is this the IC
SN74LVC04APWR
6-ch, 1.65-V to 3.6-V inverters
This could control voltage on the daughter board
I have searched and searched and come up with nothing
I have checked numerous databases and there is no chip with this name. The only way to find out what the hell this is to connect up to Analyzer to find out if this is an eprom as the name is similar to some Eeproms.
Last edited:
After searching I think that this could be our Eeprom AT24C256 instead of 24AC256. Its design matches that of the Eeprom range from ATMEL.com
ATMEL240 24AC256N ST27 tells us it was manufactured by ATMEL and other Eeproms have the 24**256 in them. This may be what we have been looking for. Remember Apple could have had made a custom part.
ATMEL240 24AC256N ST27 tells us it was manufactured by ATMEL and other Eeproms have the 24**256 in them. This may be what we have been looking for. Remember Apple could have had made a custom part.
Last edited:
OK I just forgot how to write to dev property (in OF) - I try something like:
But it crashes
Nvm you need:
"(hex-bytes)" encode-bytes " reg" property
OK I did it:
I'll probably post a tutorial on a separate thread when I have the time but basically I get the idea from here:
And used original samsung 4gb sticks (with the first 2 being original 1GB ones to boot)
encode-bytes " "(hex-bytes)" " reg" propertyBut it crashes
Nvm you need:
"(hex-bytes)" encode-bytes " reg" property
OK I did it:
I'll probably post a tutorial on a separate thread when I have the time but basically I get the idea from here:
And used original samsung 4gb sticks (with the first 2 being original 1GB ones to boot)
Last edited:
I say write it on here my friend as we are trying to achieve the same goal.
I'll ok - well as it turns out you need to tinker a little more - currently after the 1gb is utilized - linux crashes.
I think I need to modify the ram-map and/or (possibly if that doesn't work as well) - the dimm-info
Update for you guys following this thread, I am going to look on Ebay to acquire the equipment to test my theory out.
I am also going to look for a working CPU form a G5 as to retrieve the data from the Eeprom.
What we need to do is to analyse the chip and read what the pins on the IC actually do then we can access the data held within it.
I will keep you guys posted on my success.
If all goes to plan and I am correct when I have the data form the Eeprom, you can assist by examining it and attempting to modify it before sending back to the device and testing.
I am also going to look for a working CPU form a G5 as to retrieve the data from the Eeprom.
What we need to do is to analyse the chip and read what the pins on the IC actually do then we can access the data held within it.
I will keep you guys posted on my success.
If all goes to plan and I am correct when I have the data form the Eeprom, you can assist by examining it and attempting to modify it before sending back to the device and testing.
Last edited:
Actually there might be a `set-frequency` in the `words` list (for a quick try at least).
Except it may be bogus:
defer set-frequency
: ferror
throwferror throw ;
It will take some time, but it will be worth the effort, I am thinking of getting a 1.8Ghz 970PPC daughter board as these are easy to obtain, probably even get the whole PowerMac rather than just buy the CPU card.
I just want to add - if anyone is wanting to kick somebody trying to dump the dimm-info (for example) which is conveniently
... 00000400 bytes total
Well also another very important thi... this sentence is 500 character totals .
Anyway the solution is:
0 > " dimm-info" get-active-property
3 > .s
3 > ffa322b8 400 dump
Where the address and size of the dump are displayed on the second line.
And tbh nevermind how angry I'm this read here:
Is a useful read
This is my dump (it looks like there is some recognition of the modules):
... 00000400 bytes total
Well also another very important thi... this sentence is 500 character totals .
Anyway the solution is:
0 > " dimm-info" get-active-property
3 > .s
3 > ffa322b8 400 dump
Where the address and size of the dump are displayed on the second line.
And tbh nevermind how angry I'm this read here:
Technical Note TN1061: Fundamentals of Open Firmware, Part I: The User Interface
TN1061: Describes the Open Firmware User Interface, Forth and the Open Firmware language used in Apple products.
developer.apple.com
Is a useful read
This is my dump (it looks like there is some recognition of the modules):
ffa322b8: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa322c8: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa322d8: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa322e8: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa322f8: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32308: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32318: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa32328: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa32338: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa32348: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa32358: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa32368: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa32378: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32388: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32398: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa323a8: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa323b8: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa323c8: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa323d8: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa323e8: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa323f8: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32408: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32418: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa32428: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa32438: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa32448: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa32458: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa32468: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa32478: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32488: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32498: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa324a8: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa324b8: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa324c8: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa324d8: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa324e8: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa324f8: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32508: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32518: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa32528: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa32538: 80 08 08 0e 0b 01 40 00 05 25 50 06 82 08 08 00 |......@..%P.....|
ffa32548: 0c 08 70 01 02 05 07 30 45 3d 50 32 1e 32 28 02 |..p....0E=P2.2(.|
ffa32558: 17 25 05 12 3c 1e 1e 00 06 37 c3 80 14 1e 0f 00 |.%..<....7......|
ffa32568: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 13 df |................|
ffa32578: ce 00 00 00 00 00 00 00 00 4d 33 39 35 54 35 31 |.........M395T51|
ffa32588: 36 30 46 42 34 2d 43 45 37 00 00 00 00 13 26 00 |60FB4-CE7.....&.|
ffa32598: 00 00 00 6d 56 4e b1 21 00 00 00 00 00 00 00 00 |...mVN.!........|
ffa325a8: 00 00 00 00 84 50 01 20 25 25 20 32 32 1e 3c 37 |.....P. %% 22.<7|
ffa325b8: 80 08 08 0e 0a 61 40 00 05 3d 50 00 82 08 00 00 |.....a@..=P.....|
ffa325c8: 0c 04 38 01 02 00 07 3d 50 50 60 3c 1e 3c 28 80 |..8....=PP`<.<(.|
ffa325d8: 25 37 10 22 3c 1e 1e 00 00 37 69 80 1e 28 00 00 |%7."<....7i..(..|
ffa325e8: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 12 b3 |................|
ffa325f8: ce 00 00 00 00 00 00 00 01 4d 33 20 37 38 54 32 |.........M3 78T2|
ffa32608: 39 35 33 43 5a 33 2d 43 44 35 20 33 43 05 43 51 |953CZ3-CD5 3C.CQ|
ffa32618: 05 cd e1 00 59 41 44 34 43 30 4c 00 00 00 00 00 |....YAD4C0L.....|
ffa32628: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
ffa32638: 80 08 08 0e 0a 61 40 00 05 3d 50 00 82 08 00 00 |.....a@..=P.....|
ffa32648: 0c 04 38 01 02 00 07 3d 50 50 60 3c 1e 3c 28 80 |..8....=PP`<.<(.|
ffa32658: 25 37 10 22 3c 1e 1e 00 00 37 69 80 1e 28 00 00 |%7."<....7i..(..|
ffa32668: 00 03 00 00 00 00 00 00 00 00 00 00 00 00 12 b3 |................|
ffa32678: ce 00 00 00 00 00 00 00 01 4d 33 20 37 38 54 32 |.........M3 78T2|
ffa32688: 39 35 33 43 5a 33 2d 43 44 35 20 33 43 05 43 46 |953CZ3-CD5 3C.CF|
ffa32698: 06 45 d2 00 59 45 41 36 44 33 4c 00 00 00 00 00 |.E..YEA6D3L.....|
ffa326a8: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| ok
I'm sure Sonnet/Newertech/Powerlogx looked at all this stuff back in the day. Likely a lack of higher clock rate G5's spoiled the market for upgrade cards, but also the DMCA could have played a factor.
If Apple obfuscated the code, then it violates the DMCA to reverse engineer it.
It doesn't matter to us, but it would if we were trying to sell it as a product.
We don't have the documentation we need from IBM to know the PLL settings of the G5, and even if we did what could be really hope to achieve without a supply of higher clock rate G5's?
Maybe we can get a 2.7Ghz G5 to run @3.0Ghz if we can cool it, but is it really worth it for a 10-20% overclock?
Sure, some things you do because they are there, but the cost benefit here is too skewed to really make this worthwhile. Given the time and money it will take to reverse engineer all this, and given the cost of much faster more capable MacPro1,1 and 2,1 that can be picked up for next to nothing, it's not really worth it.
If you really want a fast PPC just buy a Quad G5 or a Dual 2.7Ghz G5.
If you want the fastest PPC you can reasonable get, look into running Mac OS X on Power9 hardware native or even Power7.
It can be done, and it's likely going to be a lot easier and more productive than trying to overclock a G5.
The 2.7GHz G5s were 970fx chips. Your datasheet is for the 970MP, the multicore revision used in the Late 2005 models.
At the same time, Motorola was very much losing interest in the business at about the same rate as IBM. Apple bullied IBM for the G5 because Moto was so lax on improving the G4 - keep in mind the 7400/7410 was basically the 750 with AltiVec bolted on, and it wasn't until the G4e that we saw actual improvements.
The G5's biggest issue is that Apple took workstation/server class hardware and kneecapped it to try and keep the price at a prosumer level. This meant shortchanging the poor things on cache, designing the U3 northbridge in house (which was part of why it was unreliable), etc.
Well, those and it being the era of everyone focusing on clockspeeds and long pipelines to the detriment of actual scalable performance. The 970 isn't great but it's not like Netburst in the Pentium 4 was a real winner either.
I've seen a few people say that the IBM-designed U4 northbridge in the 970MP models cost more than the CPUs themselves. Reading the specs on it, it's laughably overbuilt for what Apple used it for, so I can see that.