Question: Does PC consume RAM sticks following the installation of an AMD GPU?
Question: Does PC consume RAM sticks following the installation of an AMD GPU?
G'day everyone, I have an issue that I need urgent help with before I throw more money at the problem. I hope I'm posting in the right place.
Up until recently, for many years my PC had been rock-solid, completely stable. It is my main workhorse PC that I do almost everything with; Gaming, 3D CAD, music library, network storage, web - the only thing I don't really do is video editing. The spec was as follows:
Motherboard: Gigabyte Z490 Gaming X AX
CPU: Intel 10700K
Cooler: CoolerMaster Hyper 212 Evo cooler, NO overclock, factory bios settings
RAM: Kingston HyperX 2x16gb
GPU: Gigabyte Windforce Nvidia GTX 1070 (I forget exact part name)
Storage:
Samsung 840 EVO 512gb SSD as OS / game storage drive
2x WD Green 2TB bulk storage
1x WD Red 2TB bulk storage
PSU: EVGA SuperNOVA 750w
OS: Windows 10 64 bit - enrolled for extended security updates through to October 2026
Monitor: AOC G2770PQU 144hz 1920x1080
I try to run most games at 144hz, I cap some games at 120fps and rarely at 60fps.
Now as I said, up until about a month ago the system was rock solid. Never crashed. Performed absolutely flawlessly. So what changed? Well, I bought Cyberpunk 2077 and realised quickly that the GTX 1070 wasn't quite up to the task, at least not above 60fps. So I decided before the AI boom inevitably raises prices I would purchase a new GPU. Having used Nvidia all my life and wondering if the grass was greener, I decided to swap sides and try out an AMD GPU. The GPU I purchased is a
PULSE Radeon 9060 XT OC 16GB.
Immediately, I ran into issues. Unlike the Gigabyte 1070, the 9060 did not seat well into the motherboard. The cooler made contact with the SATA ports and would not allow the PCIE slot lever to click without physically forcing the lever itself. It did eventually, click, however, and I remember reading once upon a time that if it does click - all the PCIE pins are touching and all is well. At the end of the day, it's just plastic on plastic we're talking about here so there shouldn't really be any problems, right? I didn't have the patience for a return procedure, so I just sent it. This may have been a mistake.
After about a month of solid, reliable use (we're talking like a hundred hours of Cyberpunk by this point), and me thinking everything's fine, Cyberpunk 2077 crashes. Not a crash to desktop, one of those crashes where the entire computer just freezes and the sound glitches out. Needed a force shutdown to resolve.
I reboot, and Windows crashes again after a while. Not even gaming. This progressively gets worse and worse, with crashes happening sooner and sooner, until I start attempting diagnosis. All of the BSOD codes were relating to memory, so I ran MemTest86 (UEFI mode). This did not go well. MemTest86 found so many errors with the primary memory stick in slot b (the one closes to the CPU) that it was still reporting them 9 hours after I started the test.
So my first thought is that perhaps driving the new GPU was so intensive that it was the death knell for my somewhat old memory sticks. After all, I knew the CPU and motherboard were reliable, so it couldn't be anything else? How would a GPU kill RAM sticks? I even found the dead RAM stick's 'heatsink' clogged with dust (probably due to being close to the CPU cooler fan), a smoking gun maybe? the other one was clean..
So me, thinking old and tired RAM was the culprit, bought a brand new pair of G Skill Ripjaws 2x16gb sticks. I figured even if this wasn't the problem, I should probably buy some as an investment (sort of joking). I wanted more storage space anyway, so I bought a 1TB Kingston NVMe drive at the same time as I bought new ram. I figured hell if I'm going to spend money I'm going to at least profit from this. Never had an NVMe drive before, wanted to see how it performed.
Well guess what. After installing the new ram sticks, Cyberpunk keeps crashing.
With Windows 10's inevitable demise in late 2026, and with no intention of ever using Windows 11, I decided dual-booting with Linux was a prescient idea. So I did just that, splitting the NVMe drive into two partitions with Windows 10 on one and Linux Mint on the other.
I honestly tried using Linux as a daily, but there are simply not enough native programs yet. Not Linux's fault - I think it's a fantastic OS - but I digress. I tried playing Cyberpunk on Linux through Steam. You know what? I was shocked with how well it performed. But lo and behold, it almost immediately threw up Kernel errors. Fine, this might just be a Linux issue - after all Cyberpunk is known to be a buggy mess, and Linux gaming isn't exactly mainstream yet. But it still crashed on my squeaky clean Windows 10 installation...every time...without fail, after about 10 minutes of gameplay.
At this point I'm stumped. I ran a Prime95 test and it BSOD after one pass. Ok, so it crashes only under intense load. I can live with that, just no more Cybperpunk or other demanding games.
So lately (the last week) I've been playing Warframe, which doesn't use anything like the resources a AAA game like Cyberpunk uses, and it has been utterly reliable. There was one - ONE - crash, when I first started playing it, but otherwise it has been running fine for about a week...until today.
I wasn't even gaming at the time. I was browsing the web on windows 10 and all of a sudden my mouse just stops responding. The screen goes black for about 2 seconds, then comes back, and the AMD software goes 'hey you just had a driver crash! wanna report it?"
I go to write the report, and half way through the PC crashes completely and reboots itself. It now will not boot, and beeps continuously during POST. There was no BSOD.
This is now the current situation:
- One of the brand new G Skill memory sticks is either toast or feigning death. The one that is in slot b, the same slot that the dead Kingston ram stick was in. I know this, because the PC currently will not boot with it plugged in - I get constant error beeps during POST, and the motherboard's diagnostic LED's are showing DRAM error when this stick is plugged in. I unplug it, move the other stick to slot b, and it boots fine.
- it does not matter what GPU is plugged in, if any. I tried both, and none. The situation remains dependent on the now-dead memory stick being plugged in.
So now the tl;dr:
* system running fine for years
* install new GPU that doesn't seat quite right in the MOBO, PC runs fine though?
* a bit of time passes, memory stick in slot b dies
* replaced with brand new memory, new NVMe OS drive, fresh OS install
* a bit of time passes, memory stick in slot b dies.
So what are the potential culprits here?
* AMD 9060 GPU is bad
if this is the case, do I simply chuck the 1070 back in, replace the dead G Skill memory stick with the remaining good Kingston one, run them for max compatability and send it? If I did this, how do I know there isn't permanent damage elsewhere as well?
* MOBO is bad
if this is the case, I have the initial problem of finding another LGA1200 MOBO new, which will be difficult now it's outdated and may have to risk buying second-hand
if the MOBO is bad, the question is did I kill it cramming the GPU in? How do I know this won't happen again with any other MOBO I purchase?
* CPU is bad
The Prime95 BSOD leads me to believe possibly the CPU is (has become?) unstable? How could I confirm this?
From this point I don't know how to proceed. I don't want to just keep swapping components without understanding what the problem is, because I don't have a lot of money and I simply cannot afford to purchase anything without being dead certain at this point. Should I RMA the 9060 just in case? After all, the PC only started crashing after I installed it...
The rest of the system diagnostics seem to be fine - temperatures and voltage levels all seems completely normal, Windows never reported any hardware issues.. The only errors that are reported anywhere are to do with memory and then only after it crashes...but I've had two different sets now? So it must be a memory controller issue somehow?
The PC is now shut off permanently until I can resolve this issue. I am willing to turn it on with a good ram stick installed to try and obtain any diagnostic information should it be requested.
Thanks in advance for any help.
Welcome to the forums, newcomer!
What BIOS version are you currently running?
Power Supply: EVGA SuperNOVA 750w
EVGA offers several SuperNova series PSUs; which model do you own and how recent is it?
Compared to the Gigabyte 1070, the 9060 installation didn’t fit well on the motherboard. The cooler touched the SATA ports and prevented the PCIE slot lever from clicking properly—physically pushing the lever caused it to fail.
Examining the cooler and PCB layout, I don’t see any reason the cooler or shroud would block the card from fully seating in the PCIe slot. It’s possible a standoff is misplaced beneath the board, making the PCB flex downward.
Your GPU: Gigabyte Windforce Nvidia GTX 1070 (part name uncertain)
+
Did you run DDU in Safe Mode to uninstall all GPU drivers (Intel, Nvidia, AMD), then restart and manually install the newest driver from AMD’s support site using an elevated command?
Thank you for your prompt response! The BIOS version is presently F24a. It seems the system started as F23 before the RAM upgrade. After installing the new RAM, I updated the BIOS. To clarify, it was operating on F23 until the first memory error occurred.
I also noticed hesitation when trying to install odd numbers. When compared to the 1070 model, there doesn’t seem to be any noticeable difference physically. Upon inspecting the motherboard, everything appears intact— the metal bracket is properly aligned with the chassis, and there’s no visible movement. This suggests the issue might not be related to flexing. However, this seems like the most straightforward fix, though I did break the motherboard in an attempt to bypass the problem.
In safe mode, I used DDU as expected. I ensured it was activated because some Nvidia tools might still run under normal Windows and wouldn’t be removed properly. I followed a guide to remove all Nvidia-related files correctly.
Regarding AMD GPU drivers, I did not use an elevated command. I recall downloading the driver installer from the chip vendor site and also fetching the AMD software, intending to have everything ready at once—similar to what I do for Nvidia GPUs. Windows has improved its native GPU support, allowing me to adjust the monitor speed to 144Hz without additional installation, so I considered the official driver update a standard step.
Since you're concerned about a possible mobo/CPU/Ram problem, remove the dGPUs and use integrated graphics. I hope you hadn't altered the OS drive during troubleshooting. Seek stability by using a reliable RAM slot and a solid NVME connection. Start from there.
It's an EVGA SuperNova 720w, a model preceding the current 750w version. I've been checking voltages and 3.3v, 5v, and 12v consistently within acceptable ranges each time.
I didn't realize I was troubleshooting at that moment, but now it's clear—I should have addressed one issue at a time.
As mentioned in the original post, the system boots and operates correctly with a single RAM stick, though I later relocated it to a different slot.
Apologies for mentioning 'slot b' instead of the correct 'slot A2' from the manual; I should have specified that dual-channel setup requires slot A2 paired with B2. I've moved the working RAM stick to slot A1 now.
It's an EVGA SuperNova 720w, which comes before the present 750w model. I think they never offered a 720W unit before. Please share a picture of the stickered side with the PSU so we can review it. Upload your image to a platform like Imgur and send us the link in this discussion.