Blog 26th Sept 2026 – Linux power fault on Gigabyte GA-Z77-D3H So I've been working pretty much non stop on a linux RAID system for the last 2 months or so and have to say, I absolutely love Linux. Once I got past the struggle of figuring out how to do stuff in the terminal (a problem that knowledge resolves) I found a fluidity and command that I'd realised I'd not had for years with Windows... something I'd really missed. Originally my Linux system had been based on an elderly Asus motherboard with 4GB ram and a 2 core Intel CPU. While it's true that Linux can do more with less, this arrangement did struggle when multiple events were being handled. Heavy RAID access for example hugely delayed logins and althought the system coped, the speed hit was a problem. After a trawl on line - I figured a decent and reasonably inexpensive improvement could use something like an i5 or i7 Intel CPU in an LGA 1155 socket - coupled with a motherboard capable of supporting an SSD and at least 4 SATA drives (ideally with at least two SATA ports running at 6GB/S). The first 2nd hand candidate I went for came in at 35 UKP - a GigaByte GA-Z77-D3H with 16GB of DDR3 RAM and an i5 CPU inc fan and heatsink but no guarantee other than a screen grab of it running Linux Mint. When it arrived something interesting happened - the fans spun up for around a second and then immediately died. I had to switch the PSU off, before switching back on and repeating the process only to find the fans did the same thing. Getting a tad worried, I unplugged the PSU, earthed myself and removed the CPU fan and CPU (and memory). I checked the CPU socket pins for damage but found all were ok. I freshened up the thermal paste on the heatsink base and checked for any memory damage but found none. After putting the lot back together - it fired up perfectly. hhhhhm?!? Running with the added security of a UPS, the system worked for a week solid 24/7. During that week I configured the RAID, cloned the existing Linux boot drive to a new-old-stock 480GB SSD using the superbly excellent linux software FoxClone, upgraded to an i7 CPU and to 32GB of RAM. I configured TimeShift to run daily snapshot backups as well as created my own BASH scripts under Cron timer control to run timed RSync backups of all the key data to a remote external drive. Linux was... simply gorgeous to use. But then one day I looked in and found the machine powered off. There followed nearly three weeks of me having to debug a nasty intermittent power problem with the motherboard. In short ATX power supplies have an input called PS_ON (the green wire on the 24 pin main power connector). This input is controlled by an open drain output on the GigaByte GA-Z77-D3H motherboard generated by an IT8728F chip. Normally the PS_ON line floats high to close to 5v (via the stand by rail). Whenever the motherboard wants the PSU to switch on the main +5v, +3.3v, +12v and -12v rails the IT8728F chip will drive the PS_ON signal to ground. The IT8728F makes the decision based on a number of inputs including the on/off button on the front panel but also the power state required (so that it can cope with special sleep or hibernate states). The fault on my particular motherboard was that very occassionally the IT8728F chip wasn't holding the PS_ON output to ground when it should - and even if it did when the machine was first started (so that the machine could properly boot), occassionally it would allow the signal to float high at random times - which as you can imagine immediately caused the machine to switch off. It didn't take long to spot the likely problem. The surface mount IT8728F (called a Super I/O controller) is a 128 pin QFP arranged as a square with a pin pitch of .5mm. Viewing the four sides under very high magnification confirmed that while three were immaculately soldered, one full side was poorly soldered with a large number of dry joints where the pad copper was still visible. Many of the heel fillets on that side were missing and likely as not one or more signals were failing to make it to the chip and so causing the intermittent problem. Damn... the one tool I don't have is a hot air iron for dealing with surface mount repairs. I've used drag soldering in the past but not with a pitch as small as .5mm. After weighing up the pros and cons, I instead looked for a 2nd hand motherboard and found a cheap Gigabyte GA-Z77-DS3H with a known and as it turns out minor audio fault for 15 UKP. This worked perfectly from day one and the audio fault turned out to be confined to the socket at the back of the motherboard (but where the Audio break out header connector on the PCB worked fine). The main difference between the two Gigabyte motherboards is that the DS3H has a far simpler BIOS, providing very few overclocking options. It provides safe vanila CPU and memory settings which suits me just fine, given its job as a file server. Through all of this... Linux Mint was the one known. A stable reliable friend that didn't make anything worse... and somehow (very graciously) coped with the system unexpectedly powering itself off multiple times over three weeks. Comment | Back to Quick Links