My younger son recently built a new gaming PC for himself. He wanted to play some newer games that his older PC could not do at the framerates that he wanted. Another reason was that occasionally, the old PC would not boot up – and he would have to try numerous times. During the Covid lockdown, computer parts were ordered from various suppliers and eventually arrived. He was able to build his new PC. A brand new custom built PC powered by the latest 12th generation Intel Core i7-12700K. The build was completed eventually and he took it home and has been using it quite happily except for other problems.
A couple of weeks ago, he was having trouble with his Superloop NBN service – his router wasn’t functioning, as he could connect via wireless, but could not access the internet. We had a spare Asus RT-AC68U router, so I went over to help swap over his old TP-Link AC1750 to this Asus router. It went well, and I also plugged in a network switch in the study so that both computers would have an ethernet conenction. I asked about the old computer and was told that it stopped working some weeks earlier. I had a quick look at it while I was there and it just would not boot – power would come on, but nothing on either display was visible.
I took the old PC back home with me together with the TP-Link router. The TP-Link router turned out to not pick up an IP address from the NBN modem, as I had plugged it into my Telstra Smart Modem and could see that it was not connecting. Then I set a static IP on the TP-Link and it worked. When I set the TP-Link back to dhcp, it picked up an IP address and was working – something must have gone wrong with the configuration as it wasn’t working before and now it did. Anyway, this article is not about that router but about the old PC.
This particular PC was comprised of a Corsair case, with a Gigabyte GA-Z97X-Gaming 3 motherboard. It had an Intel Core i5-4690K processor and 8GB of ram. It also had a Radeon RX-570 graphics card in it. The processor cooling was a Corsair H100i liquid cooling kit, which I had previously written about in 2014 – https://j0hn88.wordpress.com/2014/12/16/repair-it-of-course-corsair-h100i-liquid-cooling-standoff/
During that build, a standoff had been broken which I repaired. Now, back to the PC. I plugged in my diagnostics card – a Debug King card into the PCI slot. It was awkward to view the display as it was facing towards the power supply, but I had enough of a view to see what was going on. When I powered up the PC, I could see that the debug card was show that voltages were available, but didn’t seem to change status. This was almost as if the main bios wasn’t running. I tried various things, like pressing keys on the keyboard, but nothing appeared to work.
There was no PC speaker plugged into the motherboard, so I scrummaged around in my parts boxes and found a speaker with the right plug to fit the motherboard. Powering on again – effective silence. A beep would have be good, to show that it would try to boot, but nothing. A search of the internet showed that a lot of people had been having this issue since 2016 or so, but no real resolution to this problem.
I removed the Radeon RX-570 graphics card and connected the monitor to the motherboard DVI port. Still no display. I decided to verify that the Bios was doing something, by removing the memory. By powering up the computer without memory, I should get some error beeps. Sure enough, I get a constant beeping – so that was good, as it means that the CPU is most likely fine.
I plugged the memory back in, but in the other slots – I had a bit of trouble inserting the memory as it just didn’t click in as it should but was definitely locked in place. Powering on again, I saw that the behaviour had changed. The debug card was showing memory check then eventually display check, and then i felt that it should work. About half a minute later, there was a short beep from the speaker, and the bios screen was displayed on the monitor – ok, some success.
I didn’t have the hard disks connected, so powered off, connected the disks and powered on again. Success, I could see Windows had started up and then a login screen – great. I didn’t have the password, but was able to shutdown. Now let’s try the Radeon RX-570 graphics card. It was installed, then powered up – and … the original problem happened – there would be a period click from the speaker as the computer reset itself. Could the graphics card be faulty – I didn’t think so, but all I could do is to power off, and remove it. On power up, there was no change in its behaviour – constant reset was happening.
This is very strange – it worked a moment ago, before I put in the graphics card. Let me try removing the memory again so did this and then install the memory, power on, and the computer would boot up. Some success at least – removing the memory somehow forces it to work again, when the memory is installed again. I had some spare memory and tried that as well – works with my memory, but if I put the graphics card back in, then it caused the computer to fail and stay like that, until I do the memory trick.
So what is happening and why does the memory trick seem to work. Removing the memory must change the cmos settings, such that installing memory causes an update to the cmos which then works. This seems particularly like a bios problem. Once it was working again, I checked and found that the bios was version F5 which was the latest at the time of the build. I saw on Gigabyte’s website that there was an updated F8d bios in 2016 – with improved memory compatibility. I also replaced the cmos battery – which was a CR2032 lithium cell. That didn’t seem to change anything, but at least it was a new battery.
On Tuesday evening, my son popped over for a visit and I showed him what I had found – once the PC was working, he logged on and we could see that it was functioning, but putting a graphics card in, would definitely kill it. I downloaded the F8d bios, then once I did the memory trick and it was working again, I plugged in the USB and restarted into the Q-Flash mode by pressing the End key during the BIOS startup.
The BIOS firmware update to F8d proceeded smoothly and the system was still working afterwards. I shutdown and installed the graphics card, and powered on. Well, what do you know? The graphics card was working this time. I connected a network cable, and let the computer do its Windows updates, and restarted it a few times – still working, so then shut it down. I told my son that I wasn’t sure if it was fixed, so he can leave it a few more days.
The liquid cooling system was full of dust – that is the heat exchanger – so it was a matter of taking the fans off and vacuuming the heat exchanger carefully. I could see that there was still dust in it but this would have to do for the time being. I disconnected the power and will come back to this PC in a couple of days. I wouldn’t be confident that is is fixed, until the computer consistently boots up, no matter what I do to it, like remove graphics card, install graphics card, swap memory etc.
On the Anzac Day long weekend, I had a chance to do further testing. Put the original memory back in, still ok. Remove the Radeon RX-570 – still ok. Put it back in, still booting fine. I left the power off, switched off at the power supply. Various combinations of the above and still booting ok. So it looks like the computer is behaving properly, and I am now confident that the original problem has been resolved.
What would have been the root cause? Possibly, the Bios might have been corrupted – but maybe unlikely. The Bios Cmos settings could be another cause. By removing the memory and installing the memory causes the Bios Cmos settings to be updated. Updating the Bios firmware will also update the Cmos settings. I also feel that this motherboard did not have sufficient support under the memory slots.
Usually the ATX boards are a bit deeper, but this one is not as deep, so the memory slots overhang past the last screw fitting. My other boards will be deeper, allowing an extra screw to go into the highlighted support port. Originally, the memory was in the grey memory slots, but now I have left them in the black slots. Memory inserted in the last grey slot might not have been making consistent contact. I usually try to support the board when inserting memory.
Anyway, it looks fine now – hasn’t missed a beat with booting up – so will let my son take it back next time he comes over.
R.IT 