The datasheet for 970fx cpu's does not mention the 2.7GHz cpu or give any hints about it.
Last edited:
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
I think I've some sort of success unless - like the linux is doing some smart things under the hood - I crashed the 4GB barrier of 1 GB DDR2 sticks (and it hasn't crashed - I'm still compiling Webkit-gtk on gentoo (which before on 8GB ran out of memory at some point with -j4(using all 4 cores to compile))).
I have the following combination - first 4 slots (on both sides - DDR2 1GB Samsung) then on the other 4 slots on the rest sides DDR2 4 GB Samsung.
The command is:
*It's only 18 GB - but it's fine for me at the moment - I just wanted to buy 4GB sticks wow.
EDIT: Yeah it's still running last was at more than 5 gb used (which means the 4GB stick are definitely working).
EDIT: Now I only hope my Geforce 6800 won't overheat (as it normally does under linux since I have not installed drivers for NVIDIA - I use Radeon second card normally for it) but if it does - I'll have a chance to cement the mod by adding a nvramrc function and then calling it as boot-command. And then I can also safely remove it ( I need it only for OF).
I have the following combination - first 4 slots (on both sides - DDR2 1GB Samsung) then on the other 4 slots on the rest sides DDR2 4 GB Samsung.
The command is:
dev /memory
" "(0000000000000000400000000000000040000000400000000000000100000000400000000000000140000000400000000000000180000000F00000000000000280000000F00000000000000380000000F00000000000000480000000F0000000)" encode-bytes " reg" property*It's only 18 GB - but it's fine for me at the moment - I just wanted to buy 4GB sticks wow.
EDIT: Yeah it's still running last was at more than 5 gb used (which means the 4GB stick are definitely working).
EDIT: Now I only hope my Geforce 6800 won't overheat (as it normally does under linux since I have not installed drivers for NVIDIA - I use Radeon second card normally for it) but if it does - I'll have a chance to cement the mod by adding a nvramrc function and then calling it as boot-command. And then I can also safely remove it ( I need it only for OF).
Last edited:
970FX design. For frequency scaling it is assumed that the reference clock, SYSCLK , and the related synchronizing clock, Psync run at a constant frequency. The PLL uses a fixed divider in the feedback path, but a variable, seamlessly switched divider in the forward path. The fixed feedback path allows the PLL to constantly run at a fixed frequency, avoiding the need to relock when switching frequencies. The processor clock (mclk) and bus clock (Bclk) frequencies can be changed seamlessly, while maintaining the ratio between these two clocks at a fixed value. Figure 9-4 shows this design. Note that the processor interface (PI) supports a double data rate bus. Therefore, the data rate clock (Dclk) is twice the Bclk frequency, and is constrained by the processor design to be no more than half the mclk frequency.
This indicates that it is possible to underclock to reduce power and keep the system and the cpu cooler.
Well you don't say:
localhost ~ # cat /proc/cpuinfo
processor : 0
cpu : PPC970MP, altivec supported
clock : 1250.000000MHz
revision : 1.1 (pvr 0044 0101)
processor : 1
cpu : PPC970MP, altivec supported
clock : 1250.000000MHz
revision : 1.1 (pvr 0044 0101)
processor : 2
cpu : PPC970MP, altivec supported
clock : 1250.000000MHz
revision : 1.1 (pvr 0044 0101)
processor : 3
cpu : PPC970MP, altivec supported
clock : 1250.000000MHz
revision : 1.1 (pvr 0044 0101)
timebase : 33333333
platform : PowerMac
model : PowerMac11,2
machine : PowerMac11,2
motherboard : PowerMac11,2 MacRISC4 Power Macintosh
detected as : 337 (PowerMac G5 Dual Core)
pmac flags : 00000000
L2 cache : 1024K unified
pmac-generation : NewWorld9.8.1 Initiating a Frequency Change Software initiates a frequency change by writing to the PCR. The value written to the PCR frequency field determines the target frequency being switched to. The values in the parameter fields must correspond to this new frequency. Similarly, if the voltage field is used, the voltage requested must correspond to the frequency requested. The North Bridge is responsible for changing the voltage before the frequency change when raising voltage, and after the frequency change when lowering the voltage.
Is this in OSX or is this in the Open Firmware?
Is this in OSX or is this in the Open Firmware?
This is on Gentoo.
But on OSX there is Reduced power settings which I believe does the same thing.
Great minds think alike you have just found evidence of the PVR and the cpu frequency etc.
Surely your PPC is a Quad core yet the data displays it has a dual core mac. detected as : 337 (PowerMac G5 Dual Core)
1250.000000MHz times by two gives 250.000000Mhz
I wonder if Apple bodged OSX for the Quad PPC in order to get the MP to detect all cores. In reality it is not functioning as it should. Apple would not care as they were already building Xeon Macs so why would they care.
As long as they produced something that pleased the market at the time and it bought the dollars in, they could not care less.
Last edited:
many years ago there was the photoshop test and I ran the test with a ram disk and parts of OSX disabled, when I get chance I will try to replicate the same result this time I will film the Quad in action. I am curious if I can replicate the same results.
My reason for bringing this up is that I thrashed the Xeons score and everyone else's at 13 seconds if I remember rightly. This surely proves that OSX itself is slowing down the computer with all its fancy garbage running.
Why do you need an operating system that runs everything at once. I want an OS that I can strip out all the crap that I do not need. There is very little point in going for things that are memory and process hungry and really are just gimmicks!
Perhaps we need a bit of software that can enable and disable all the unimportant parts of the OS but can reenable them at will and very easily.
My reason for bringing this up is that I thrashed the Xeons score and everyone else's at 13 seconds if I remember rightly. This surely proves that OSX itself is slowing down the computer with all its fancy garbage running.
Why do you need an operating system that runs everything at once. I want an OS that I can strip out all the crap that I do not need. There is very little point in going for things that are memory and process hungry and really are just gimmicks!
Perhaps we need a bit of software that can enable and disable all the unimportant parts of the OS but can reenable them at will and very easily.
I'm not sure if you are ironic or not
I have great evidence of those data being correct.
Most of all my fans don't spin like a rocket launcher while compiling with all 4 cores.
Least of all - before those same values where 250.000000Mhz
Here is the command I used:
EDIT: In any case I take the compliment - I'm sure a great mind.
EDIT: But if you insist of pointlessly ruining your system - be my guest.
Last edited:
I’m guessing you’re referring more to the alliance itself, coupled with Motorola’s divestiture in ’04 (of what became Freescale) and less to the continuing of RISC-based products themselves — as IBM’s POWER architecture and the embedded chips Freescale continued to develop were both derived from this root.
A balance between feasibility (yes, it can be done, even if there are steep costs and compromises) and practicality (doing all as designed, reliably and cost-affordably) was beyond reach during that time.
Apple’s kneecapping, as you put it, was that balance struck. The IBM-derived POWER4 product as a PPC 970, in a prosumer setting, was not reliably ready for prime time.
Again, this was fairly early in the 64-bit architecture realm, and Intel couldn’t produce a reliable 64-bit product of its own until 2006 (as sales of Itanium, the Intel-HP product, lagged well behind POWER ISA and what would become x86-64, and didn’t have a place in Apple’s consumer/prosumer/entry pro server products based on i386 CISC).
As you noted, the Pentium 4 Type F/Netburst, from ’04, wasn’t entirely ready for prime time, either. It wasn’t until 2006 when Core-based Woodcrest/Merom chips went into production that Intel could deliver reliable, x86-64 systems across the board.
Oh, I definitely can’t forget that window of time. I recall the Jobs keynote in which he vowed that the G5 would “soon” break 3.0GHz — in effect, publicly front-loading a promise not backed by much more than limited, in-house testing and not, say, verifying in advance that a manufacturing/fabrication process with these promised clock speeds with current generations could be produced reliably.
(Contrast this with the front-loading of keynote promises for all in-house stuff, such as the transition to OS X from OS 9, which was completed faster than originally announced.)
Given the lead time required to get all that re-design into production and manufacturing, the cost of that redesign leaves one to wonder whether this was Apple’s “forget it” moment. If so, then this probably emerged around when pre-production planning of 970MP/U4 systems got underway in late ’04.
Similarly, I wonder much the same with the roadmap from 7447 to 7448 (and nothing planned or designed beyond the 7448), as the final iteration of G4 products from Apple, much like the 970MP-based G5s, incorporated a complete redesign of the logic board to accommodate the 7448 and clock speeds beyond 1.67GHz.
Whatever the case, Apple remained that linchpin in the withering alliance, and under Jobs, Apple kind of squandered that alliance from as early as 1998 — namely, when Job ended licensing of third-party PowerPC products, instantly reducing the near- and mid-term consumer market for PowerPC CPUs.
To wit — and this is kind of important — Freescale and IBM in 2006 renewed their Power-based alliance, without Apple, as Power.org, now OpenPOWER Foundation. That alliance continues with IBM and Freescale’s successor, NXP, to this day, with both IBM’s POWERx server products and NXP’s PowerPC e5500/e6500 64-bit QorIQ CPUs (which, it should be noted, still incorporate AltiVec technology).
I have no intention of ruining my system.
I want to change a few things in my Quad to improve it, I think slimming down OSX and getting rid or even replacing the useless features would improve performance.
Last edited:
IBM was more interested in the CELL CPU used in gaming consoles than it was PPC for Apple's use.
Apple's sales of laptop Macs was far more share of revenues than desktops and the G5 chipset used far to much power. Prototype G5x laptops had to be so hobbled power and clock speed wise that they were slower than G4 based laptops that Apple was selling at the time.
For its' day the G5's were great machines for the desktop market, but they fell behind too quickly.
I'd be happier had Apple used full-blown Power6/7/8/9/10 than the switch to Intel, but maybe the costs were just too high compared with Xeon CPU's and the Core/2 was so very much faster and scaled in clock speed and power usage so much better than the G4 or anything AIM had in the pipeline.
The more things change, the more they stay the same. AppleSoC if far better than anything x86_64 has in the pipeline for mobile use, but for the desktop( workstation/pro ) market it may yet prove to be vaporware.
I'm just not sure a SoC can scale up in power and speed to be any faster or better than Xeon/Threadripper/Epyc, but I'm hoping I'm wrong.
I can't get my head over this:
1. When system is on and active (compiling something) - it can run days.
2. Sometimes when the system is not active anymore (finished compiling) - it hangs.
3. Sometimes it won't power on - with the diagnostic (LED #7) on
3. Is fixed often by just waiting (and/or wiggling components around):
I'm thinking can this be an issue with the motherboard cooling?
And also if I need to take the motherboard off - how do I not break the power switch.
1. When system is on and active (compiling something) - it can run days.
2. Sometimes when the system is not active anymore (finished compiling) - it hangs.
3. Sometimes it won't power on - with the diagnostic (LED #7) on
3. Is fixed often by just waiting (and/or wiggling components around):
I'm thinking can this be an issue with the motherboard cooling?
And also if I need to take the motherboard off - how do I not break the power switch.
Yah, I don’t know how to do that. I broke mine when I disassembled everything a couple of years ago. I power-cycle it using my Apple Cinema Display.
I may give a second go at full disassembly at some point. If I do, one of my goals is to repair the damage I caused to the front power board connection.
I would save everything and reset the PRAM, try again and then even reseat the ram. It may not be the mobo, but also given the system a good blow out in case of dust as this loves the heat.
Last edited:
I read somewhere that the graphics card can cause the system to hang at start up, Reseat the card as well.
I’m not a big fan of “I read somewhere” as a citation source, particularly when a problem described is probably unrelated to a poorly seated GPU card (as a poorly-seated GPU card might prevent display, but wouldn’t prevent booting, as G5s can run “headless”). Could you find that citations and post it here? Thanks.
The problem, as described, is probably related to the G5 hardware itself.
My own A1047 presents erratic behaviours which have eluded remedy — like soft-hanging when idle (but not sleeping, and also won’t drop to sleep if requested by pressing the power button); running high-CPU compiling tasks without issue, but auto-reboots (i.e., recovering automatically from kernel panic) when running the G5 as a Shoutcast server. The only thing I can guess causes these behaviours the overheating of the memory heatsink area becoming too critical in temperature to function without kernel panicking.
Again, this appears to be a failure of U3 design — namely, cooling — on Apple’s behalf. It’s a shame that cooling for the memory heatsink on the U3/backside is so poorly designed. While a thorough cleaning-out of the internals, including the hard-to-access backside, may redress some of these issues, it is not an assurance that these troubles won’t persist afterwards on these older systems.
PowerMac G5 startup failures, red LED beh… - Apple Community
Reset the SMU, using the button on the logic board below the lower memory slots (apparently this button is on both the late 2004 and late 2005 PowerMac G5 models, which is not what the documentation says), then zapped PRAM as it was coming up, and lo and behold, the machine started!PowerMac G5 startup failures, red LED beh… - Apple Community
discussions.apple.com
further down the thread, from 2008:
Again, these logic boards have design flaws. Those flaws, especially on survivor units (i.e., ones still running in 2022), find their way to the surface in peculiar, sometimes unrepeatable ways across different units.
When one owns one of these survivor units and is running into intermittent issues like the ones either @sasho648 or I described, the best one can do is to take remedial steps to reduce the frequency of these issues from popping up, as they will continue to do so. Now, unless you’re a board repair expert in, specifically, G5 main boards, and you definitively know what to fix (and can offer to do so for a fee), then facile, “it needs an SMU reset” for these survivor units is not really very helpful.
We have to come up with all sorts of ideas to resolve problems, my G6 Quad use to hang and refuse to boot. I use to get just the fans running at top speed and the only cure at the time was to keep resetting the computer until it booted. I was lucky in that I swapped out the ram and it has not done it since.
Super. That worked out for your unit and yours alone. The quad A1117 is also a different creature than the A1047.
I'm not sure but the issue is likely the fact that after installing the heatsinks from the air cooled other G5 I have (since the water cooler was not doing its job) - since there was some misalignments (and my own lack of knowledge but I'm learning) - I didn't screw the shunt bolts (for the power).
I thought they were just for keeping the cpus in place but after I was removing the rear fan to disassemble the motherboard - quickly figured out what they really are (by the huge wires going to the power supply).
At least that's what I think it's (I hope so).
So before they were just making contact from the occasional gravity (since I have my mac laying on the floor on its side).
I thought they were just for keeping the cpus in place but after I was removing the rear fan to disassemble the motherboard - quickly figured out what they really are (by the huge wires going to the power supply).
At least that's what I think it's (I hope so).
So before they were just making contact from the occasional gravity (since I have my mac laying on the floor on its side).
I must say that my dual 2.5Ghz which was custom built for me buy Apple has never had any issues at all. I push the button and on it comes. It never hangs or crashes, I did upgrade the ram more or less straight away and added a second HDD.
I just wonder if the quality of components in the quad is substandard, for example 35year Amiga 1200's and Atari Falcon's never have any issues with the electronics. Yes, there are folk out there that want to recap them for people but that is to make money. I had a Atari Falcon 030, that I had from new and in all the years I used it, no components were ever replaced, eventually I sold it.
Ref Apple G5 PowerPC Service Manual
No boot tone (LED on, fans spin, no boot tone, video) or No video (LED on, fans spin, boot tone, no video) 1. Reseat video card 2. Replace video card 3. Replace front panel board cable 4. Replace speaker 5. Replace front panel board 6. Replace logic board.
This is where the original person must have got their information from.
Here is something that is interesting. Here are the part numbers for the different PPC 970's and look below that different logic boards are configured for different speeds.
If it is possible to swap CPU without swapping the logic board then the logic board ***.2895 or ***2950. If it was this straight forward, we would all put 2.7GhZ PPC processors in our machines and underclock them.
Last edited:
After reading the service manual for the G5, there is an official way to test the logic board and the psu and to see if the voltage is passing through to the 970/s
You need a multimeter or voltmeter to test to perform the test.
The logic board has terminal connection that provide voltage to the cpu. If you are experiencing problems then this needs to be performed. This could indicate several things.
Faulty PSU or a faulty Logic Board.
The terminals are not accessible on the Quad so I would recommend that you remove the processor module and connect clips to the terminals and label each wire and then reassemble.
I would copy the following information straight away as it may be removed by the powers that be.
This is posted for educational purposes only for enthusiast that wants to preserve their G5.
There is other ways to test as well. I would look voltages as the system is underload and measure the voltages if your system hangs and then post the results. We can then establish what happens during the system running and hanging etc and during hang at startup.
You could have a voltmeter permanently connected to monitor the voltages. You could use something like this eBay item number:284964434191.
I am thinking of building something and mounting them somehow in Mac style box and have 4 voltage meters running with each one being labelled. I have no idea on actual design, this could help prevent damage having these voltages monitored. It could make a fancy addon for the G5. If you are into that sort of thing.
You need a multimeter or voltmeter to test to perform the test.
The logic board has terminal connection that provide voltage to the cpu. If you are experiencing problems then this needs to be performed. This could indicate several things.
Faulty PSU or a faulty Logic Board.
The terminals are not accessible on the Quad so I would recommend that you remove the processor module and connect clips to the terminals and label each wire and then reassemble.
I would copy the following information straight away as it may be removed by the powers that be.
This is posted for educational purposes only for enthusiast that wants to preserve their G5.
There is other ways to test as well. I would look voltages as the system is underload and measure the voltages if your system hangs and then post the results. We can then establish what happens during the system running and hanging etc and during hang at startup.
You could have a voltmeter permanently connected to monitor the voltages. You could use something like this eBay item number:284964434191.
I am thinking of building something and mounting them somehow in Mac style box and have 4 voltage meters running with each one being labelled. I have no idea on actual design, this could help prevent damage having these voltages monitored. It could make a fancy addon for the G5. If you are into that sort of thing.
Last edited: