- From: Adelaide, Australia
- Registered: 2014-05-12
- Posts: 877
Power Mac G5 - Three LED Flashes with Good RAM (Guide)
I originally posted this on MacRumors on the 25th of November 2013. I'm reposting this information here for future reference.
Several years ago, I owned a Power Macintosh G5, a 1.8GHz Single Processor. The machine was a first generation, codenamed "Omega" and manufactured some time between June 2003 and June 2004.
When I acquired it, the machine would power on without a chime, and the LED on the front would flash 3 times, pause, then flash 3 times again, with the sequence repeating until the machine was powered off. The machine would never display any video, and over time the fans would slowly ramp up to maximum speed.
The flashes of the LED indicate a specific error code, designed to help isolate a fault in the system:
A power-on self test in the computer’s ROM automatically runs whenever the computer is started up after being fully shut down (the test does not run if the computer is only restarted). If the test detects a problem, the status LED located above the power button on the front of the computer will flash in the following ways*:
• 1 Flash: No RAM is installed or detected.
• 2 Flashes: Incompatible RAM types are installed.
• 3 Flashes: No RAM banks passed memory testing.
• 4 Flashes: No good boot images are detected in the boot ROM (and/or there is a bad
sys config block).
• 5 Flashes: The processor is not usable.
* Note: The status LED lights up when the power button is depressed at startup. Do not count this light as one of the diagnostic flashes.
Now, the first step is to attempt reseating the memory, replacing with known good memory, and ensuring that the memory is in the correct slots. Power Mac G5s don't respond well to unmatched RAM or memory installed in the incorrect slots.
However, sometimes the issue goes beyond that. In the case of the system I experimented with, a reseat of the memory could occasionally allow the machine to start, but it would lock up during startup, requiring a hard power off. New, known good replacement memory would do the same.
One issue I found commonly discussed on the various Mac forums relates to the solder joints below the memory slots. The theory is simple; reseating the memory flexes the logic board enough to create a momentary connection of the broken solder joints, resulting in a near successful startup attempt.
This sounded promising, since it was consistent with the behaviour my machine was exhibiting.
A common test for the issue was to slide out the fan assembly and run a hair dryer back and forth over the memory slots from the front side of the board. I've highlighted the area for reference:
The theory is based on the fact that as metals expand when heated and shrink when cooled, the heat from the hair dryer would slightly expand the solder joints, temporarily rebonding the connections and allowing the machine to boot once again, or at least until the joints cooled and the connection was once again broken.
After 5 minutes under the hair dryer, my Power Mac G5 booted up, albeit briefly. The system locked up at the "Starting Mac OS X" screen, and the fans began to ramp up once again.
The longer I heated the machine, the longer it would last. With 15 minutes of heat, I made it far enough to format a new SATA drive and prepare it for installation before the G5 locked up once again.
Here's where it becomes interesting.
Removing the Logic Board from the system, I painstakingly reflowed each solder joint below the RAM slots with a soldering iron to rebond any broken connections, then refitted the board to the machine.
After powering it on, I saw the same 3 flashes, followed by the telltale sound of the fans ramping up. The G5 was still dead.
Tracing The Fault
Using the handle of a screwdriver, I probed around inside the enclosure, applying pressure at various points on the board to see if it would affect the machine's behaviour at all.
After applying moderate pressure to the area highlighted below, being careful to not damage the small surface mounted components in the process, the machine started up.
Removing the Logic Board for inspection once again, I found that the components are directly behind a large chip on the underside of the Logic Board.
This is the integrated circuit that handles the transport of data between the memory, processors and system bus, and it looks almost identical to the render of the G5 System Controller from the Power Mac G5 Introductory Keynote. Skip ahead to 78:00 if you don't wish to watch the entire thing.
Looking at the Block Diagram for the Power Mac G5 from the Service Source manual, it appears Apple also refers to this custom IC as the U3 Memory Controller / Bus Bridge, and of course, it directly interfaces with the memory slots.
I studied a document released by IBM in 2004, titled "Development of BGA Solution for the IBM PowerPC 970 Module in Apple's Power Mac G5". This document is freely available from IBM's website and is worth a read if you're interested in the minute technical details of modern circuit board assembly.
The U3 Memory Controller / Bus Bridge is affixed to the board using a Ball Grid Array, or "BGA" for short. This chip packaging technique places the electrical connections on the underside of the chip, allowing for a compact chip package that is easier to place on more densely packed circuit boards.
One of BGA's disadvantages is that the electrical connections can be subject to stress and fatigue over time, either through flexing or heat expansion and contraction.
Given enough time, those electrical connections can crack, and the process can occur much sooner if the solder used is of a poor quality, as we saw with the nVidia GeForce 8600M GT problem that plagued 2007/2008 model MacBook Pros, or with inadequate cooling, as seen with the "Red Ring of Death" issue that affected the Microsoft Xbox 360.
The Power Mac G5 does not have these factors playing in its favour. The machine runs more on the warm side, evidenced by the use of large heatsinks not only for the G5 processors themselves, but also the extremely large heat pipe arrangement on the back of the board to cool the U3 Memory Controller / Bus Bridge.
Adding to the problem is the fact that the Power Mac G5 uses a very large Logic Board that is subject to flexing with thermal expansion, and the removal and insertion of memory into the slots also flexes the board.
As these machines have aged, some of the electrical connections below the U3 Memory Controller / Bus Bridge have fractured, resulting in an intermittent boot or no-boot situation.
Since the Memory Controller can not interface with the memory slots, it is unable to detect any memory slots that can be used by the system, and so it throws the 3 flashes of the LED on the front:
• 3 Flashes: No RAM banks passed memory testing.
The Repair (Attempt)
So, if you could repair those electrical connections, the Power Mac G5 would run reliably again. Most homebrew BGA repairs involve either the use of a heat gun, to heat only the affected chip, or an oven, to evenly heat the entire board and reflow all the components, along with the affected chip.
BBISHOPPCM's World on YouTube has a video titled "Repair PowerMac G5 Motherboard Failure (NON-Watercool Only)". This is perhaps the most well known example of an oven reflow of a Power Mac G5 Logic Board.
This is where I need to stress that reflowing the chip is a difficult process, compounded by the large size of the Power Mac G5 board. While it can repair a Power Mac G5 Logic Board, it isn't guaranteed to work, and if not performed correctly, it can irreparably damage the Logic Board. Unless you have experience with PCB work, I would recommend speaking to someone that is more confident with BGA repair to get this task done.
Due to the large size of the G5 Logic Board, it is very easy to warp the board when heated. It is absolutely critical that the board is supported during the heating process. If you watched the video from BBISHOPPCM's World linked above, you would see that he used coffee mugs evenly placed to support the entire board, not just the affected corner.
You will also need to strip the board of as much as possible, as there are some wires and components that could melt or explode in the process, such as the PRAM battery. As a result, this can be an involved repair at times. It's worth visually inspecting the board and removing as much as you possibly can first.
I opted for the heat gun method, and I can't say it went well, although it doesn't mean this method isn't possible to perform correctly. If I had to try this again, I would try the full board in oven method.
Using a heat gun, heating the affected chip and the memory slots, I have managed to warp one Logic Board using excessive heat. After this occurred, the board appeared bent and would no longer sit flat. The heat pipe would no longer affix to the board, and the board would no longer fit the case correctly, let alone the reliability and functional issues that come with a warped circuit board.
If I had to use a heat gun again, I would heat only the affected chip, and would try to heat it for no more than 30 seconds to one minute.
While I would like to write a definitive guide on how to repair the G5 Logic Board, I haven't had success with the one occasion that I attempted this repair, and the failure to repair this board allowed me to perform the dissection necessary to write this article.
However, I do believe that I have identified the problem spot on the board, and if I could source another G5 to repair, I would likely be successful the second time around. Unfortunately G5s in Adelaide, Australia are a little thin on the ground, and I haven't been able to find another yet.
The aim of this write-up is to add to the already large amount of information out there about repairing the Power Mac G5 and to assist those who are seeking this information to repair their own Power Macs.
As many Power Mac G5s have since been scrapped and harvested for their enclosures, there are fewer available to enthusiasts and collectors. Prospective owners that wish to attempt to revive a dead system may find this article useful. If even one machine is repaired using this information, then it's been well worth my time in writing it.
Best of luck.
Last edited by iMic (2014-08-14 04:28:20)
Resident Professor of Alternative Methodology
Faculty of Macintosh Restorations & Modifications - "It works, let's fix it!"