System operation during defined stresses Overall performance at varying conditions
System operation during defined stresses Overall performance at varying conditions
I own a fairly recent build (see details below) that experiences a full system freeze under certain circumstances. The freeze is absolute—the display turns black, no signals are registered, and the machine won’t power up. It doesn’t trigger a BSOD. To recover, I must perform a power cycle at the PSU; simply holding the power button isn’t enough. This issue occurs consistently even with BIOS settings unchanged, whether XMP is enabled or disabled, and whether AMD’s easy overclocking is active. The problem is sporadic. Initially, it only appeared when using Handbrake for transcoding. It can also happen during Blender renders, especially with Barbershop animation. I’ve tested Prime95, HeavyLoad, and Cinebench for extended periods without encountering any instability. Memtest86 has also been run numerous times with no failures. This suggests the fault arises under real workloads, not from simulated ones. I suspect potential causes include power supply weaknesses (voltage drops), CPU or memory faults, or motherboard issues. It doesn’t seem to be related to overheating, as the CPU reaches ~80% and GPU ~70%. I’m not convinced it’s a thermal problem. I’d like suggestions on what to try next or which component to inspect first. Thank you! Build specifics: CPU – AMD Ryzen 9 7000X; Mainboard – Gigabyte B550M DS3H, BIOS F15; RAM – 2x Corsair Vengeance LPX 16GB DDR4 3200 C16 (dual-channel); Cooler – Noctua NH-D15S SSD 1TB M.2 NVMe; PSU – EVGA SuperNOVA 550 GA; GPU – Asrock Challenger Intel ARC A750 (previously XFX Radeon RX6400); Case – DIYPC Silence BK with Noctua cooling fans; fan controller absent.
No logs detected in the Windows Event Viewer around the crash time. Error, Warning, and Critical entries were not found. It does seem unusual. A quick check on the system bio might help assess stability upgrades for your newer Ryzen 7000 series.
Good observation. I didn't mention that detail. The Event Viewer appears tidy. Messages flow normally until the startup prompts, after which nothing happens until the boot sequence begins. No issues or warnings prior to the event. The BIOS version on this board is already up-to-date. Gigabyte offers both F15 and F15a versions; one might expect the 'a' version to be newer, but it isn't. Considering the release timeline, the 'a' variant is actually six months behind.
If nothing appears in the event viewer, it's probably not an OS or software issue. It seems unlikely a complete crash would stem from memory problems, but testing with just one module installed might reveal instability. Swap it out and see if the issue persists. Also, monitor the VRM temperature during rendering.
Thanks for the suggestion. I hadn't considered the single-stick memory test. I'll attempt it outside of work hours. Regarding the VRM question, what does "overly hot" refer to? In one of my earlier tests, I monitored HWINFO and noticed the average VRM MOS temperature rising sharply about 10 minutes before the crash. This aligns with the workload start, as other temperatures (except CCD1) increase together. The VRM MOS temperature stays below 79°C on average, and it never goes above 80.0°C at full resolution. The graphs are included if needed (VRM is the yellow line). Spoiler: it matters.
Certain boards tend to heat up significantly on VRM units, but the temperatures stay within acceptable ranges, suggesting overheating isn't the main cause of the crash. Consider testing the single channel first to observe the behavior. It might be worth trying before replacing other components. If the issue persists with the single channel RAM, explore switching to a different PSU or temporarily using another board as a donor for diagnostics.
It could be possible. During further tests, I attempted one memory stick at a time and encountered failures consistently. It seems there might be two faulty DIMMs or another underlying issue. The HWINFO log from the later runs indicates the VRMs reaching extremely high temperatures. I wasn’t sure why this happened in my earlier test, but it could be an anomaly or a timing error. In the VRM temps I observed using Blender to trigger the crash, they rose from the 60s into the 90s or higher. Failures typically occurred when VRM temperatures exceeded 95°C. Sometimes they were around 95°C, other times up to 104°C, but never below that threshold. As an experiment, removing the top fan/radiator cover and installing a subpar 120mm fan on top of the case as an exhaust led to significantly longer failure times—over 20 minutes instead of less than 10. While I’m not certain VRM issues are the root cause, I plan to order replacement parts and test if it helps. I’ll also add another 140mm fan to the CPU cooler. The NH-D15s include one, so adding a second should improve airflow through the front intakes and potentially stabilize the VRMs. I’ll keep investing in Noctua and update this post when they’re available for installation, unless I discover better alternatives before then. Thanks for the tip about the VRMs; I’ll follow it closely before replacing expensive components like power supplies and system boards.
The updated components were installed and tested again. I increased the NH-D15s cooler by another 140mm (now it has both) and installed a 120mm exhaust. This setup provides significant airflow. I also placed an NVMe heat sink on the back of the motherboard near the exposed VRM chips, while the rest remain on the front and under a heatsink. Although it doesn’t look appealing, it’s improving heat contact and lowering VRM temperatures quickly (though still warm to the touch). Despite these changes, the issue persists. I’m not upset about the extra parts since they improved the build quality overall, but they didn’t solve the problem. The system still crashes under stress, but it doesn’t shut down or panic. Power LEDs and fans stay on during the incident. Both +5V and +12V rails remain active, and the power button doesn’t work. I can’t hold it down to stop the system, so I have to switch power settings on the PSU to turn it off and restart. The problem occurs with both Linux and Windows. I’m focusing on the system board as the likely culprit. We’ve eliminated VRM overheating, BIOS updates (three revisions tried), and memory issues. The absence of a proper shutdown suggests the PSU might be the main cause. I’m planning to order an MSI B550M Pro-VDH board for comparison and testing. If that doesn’t work, we’ll need to check the CPU or PSU. It’s the same situation as before. I’ll update later with results, unless someone has better advice!
Following the motherboard replacement, everything seems to function properly. I completed a 4+ hour Blender render, transcoded a video queue of 173 items, and ran another 8+ hour Blender render with some Portal 2 elements. The exact issue remains unclear, but it was clearly hardware-related, likely involving the motherboard. It probably stemmed from a VRM problem, which I resolved by swapping the board. I’ll reach out to Gigabyte to check if they offer a replacement program. If not, I’ll install a lower-power CPU—possibly a Zen 2 model—and pass it to someone for light use. Thanks for your assistance, Solado. Your prompt response and support during troubleshooting were valuable.