Full disk usage, no data transfer. Occurs unpredictably during gaming or video encoding, or other workstation tasks.
Full disk usage, no data transfer. Occurs unpredictably during gaming or video encoding, or other workstation tasks.
Hello everyone, I want to clarify that this problem matches exactly what was described in the thread. The main variation lies in my hardware configuration. My system has no SATA devices, and all SATA controllers are turned off in the BIOS. Windows 10 doesn’t recognize them as devices. I’m using a Windows 10 Professional System spec: AMD Threadripper 3970X with ASUS Zenith II Extreme MB (not the Alpha model), with the latest BIOS update. It has 128GB DDR4 3200 CL14 RAM (8x 16GB modules) - G.SKILL Trident-Z RGB, D.O.C.P. enabled in BIOS. The AMD Radeon RX5700XT OC 8GB is installed on ASUS ROG STRIX, and the ASUS ROG Thor 1200W PSU was recently replaced under warranty; it shows 2.9v instead of 3.3v in BIOS. Storage setup includes eight NVME SSDs, five Samsung 980 Pro (four 1TB, one 500GB), and three Corsair MP510 1.92TB drives in the ASUS Hyper M.2 PCIe card. This card is connected to Slot 3, which supports PCIe RAID mode or Bifurcation in BIOS. Currently, it’s set to Force PCIE Gen3, with three drives Gen3 and one Gen4; I intend to switch soon. For now, I limit the Gen4 speed on that drive to match the other three. Two SSDs (one 500GB and one 1TB) are in the mainboard front slots, and two more 1TB SSDs are in DIMM slot 2. I use RAID0 configurations for some drives, but most are single disks—Windows and games are set up as individual drives. Three 1.92TB Gen3 drives are in one RAID0, and three 1TB SS 980 Pro units are also in RAID0. This was mostly done just for fun. I rarely store much data on the RAID0 volumes. Lastly, my system runs a custom liquid cooling loop with EK WB components. I use CableMod cables for both the motherboard and GPU. Previously, I overclocked the CPU, but I’ve since returned to stock BIOS settings. My Windows installation was created from a USB flash drive made via Media Creator, converted to ISO, and burned to a new drive. The original file had no changes. I’ve tried various fixes in Windows, run Memtest86 on RAM, performed CHKDSK and SMART checks, and scanned both block and CHK errors. No issues were found. I regularly update drivers for chipset, processor, GPU, and keep everything current. This installation has been taking longer than usual—about 18 months since the fresh setup. The hardware remains unchanged; I added the three Gen3 SSDs in early 2021 and upgraded my 5x SAS 980 Pro drives from Samsung 970 EvoPlus in early 2021. Many articles mention problems with SATA/AHCI controllers, but I didn’t enable SATA ports in BIOS, so those solutions didn’t help much. I’m curious if others have faced the same issue with only NVME SSDs and no SATA devices. From what I’ve seen, the problem tends to affect my C drive or D drive specifically. The C drive holds a single 500GB Samsung 980 Pro, while the D drive contains one 1TB drive without RAID0 configuration—its placement isn’t clear from the BIOS when setting up RAID arrays. I noticed during upgrades that the BIOS chose drives differently: sometimes one on the mainboard, one in DIMM.2, and another on the front for the array. To confirm, I’d need to remove the drives and inspect their serial numbers. Right now, I’m contemplating backing up all data as a precaution, wiping the current storage layout, and reinstalling Windows from scratch while reconfiguring everything. It’s becoming increasingly frustrating when I play games or run overnight tasks and end up with a nearly empty drive—while the performance tab shows 0% usage, disk activity is at 0%. Please let me know if anyone else has encountered this situation.
I suggest starting with the memory profile CL14 at 3200MHz on Ryzen, though it may be too aggressive—CL16 is more typical for this platform. CL14 at that frequency works well in a dual-channel two DIMM setup, but avoid quad channels unless necessary. Turning off DOCP would be a good first step. Alternatively, you can mimic a CL16 3200MHz configuration with timings of 16-18-18-36 at 3200MHz.
I adjusted the settings manually at 3200Mhz. On the Zenith II Extreme board (uncertain if it's an ASUS model) they included a DRAM RAS# PRE time option between the third and fourth timing choices. It wouldn’t let me set it to 36, so I opted for auto. They suggested 22 as part of the DRAM training. These kits typically follow the G.SKILL spec (14-14-14-34), where the DOCP setting also had RAS# PRE at 14 and worked well for a long time before this problem began. At first I thought the problem was with the 3.3v rail, but after replacing the PSU it still showed lower readings but not redline. I suspect the issue might be related to multiple SSDs sharing that rail—maybe removing some would help. END EDIT I’ll give it another go. I used Corsair CL16 RAM in quad channel before, but faced stability issues; Windows reports no problems with SSDs. I swapped the RAM about a year prior to the problem. Those kits cost significantly more (around $2100 AUD for eight modules) and all have the same serial numbers despite being sold separately—possibly luck was on my side. They claimed testing was done on AMD platforms, but not necessarily for quad channel. I’ll try again, easing the timing back to CL16. I’m not very comfortable with sub-timings, so turning off DOCP and setting it manually at 3200MHz without changing timings might reveal how it performs under standard JDEC settings. It could take some time before I share results, as progress can be slow. Sometimes the last occurrence was about six weeks ago. Occasionally, during Diablo IV, it happened multiple times—first around 40 minutes after start, then 10 minutes, and finally after the first mobs were killed. Edited October 24, 2023 by Manbull-Dave Added update, modified RAM timing as advised.
Tonight I updated. I was running Diablo IV again but it froze. I checked the task manager – it showed the D drive at 100% activity, yet not actually reading or writing. The event viewer displayed numerous errors indicating the device might not exist. In Device Manager everything looked normal, but when I clicked on a drive to open properties, the manager also crashed. The AMD RAID tool confirmed all drives were present and functioning properly. I’ll need to restart to clear Windows from its instability since the D drive isn’t responding for SMART checks. Last time I verified there was nothing unusual in the SMART data for any drive. After rebooting, I’ll recheck the SMART info and run the AMD RAID utility in admin mode just to be safe. I accidentally launched it without admin rights, which made things tricky because both Taskkill and Task Manager were stalling. I also noticed some errors in the event log linked to Ryzen Master, installed a long time ago, and decided to uninstall it. Despite that, the SMART data from Crystal Disk Info still looked normal. All drives are now detected after restarting, and seem okay. In the AMD RAID tool, two drives not in a RAID appear as legacy devices, while six in RAID0s show up as disks. The only problem is with one drive (C or D) that isn’t part of an array. I haven’t used Ryzen Master much since over a year. I’m unsure if the issue lies with my drivers/OS or something deeper. My plan is to reset my storage settings back to BIOS defaults, turning off NVMe RAID and running drives individually. If needed, I can use MS Storage Spaces to manage them separately. Alternatively, I could just assign each drive a different letter. Another thought: maybe the hardware itself has a fault, but the BIOS and HWInfo64 are showing the 3.3V rail at lower levels (sometimes dropping to 2.96V) even after changing my PSU in March. I’m not sure what else could be wrong. I considered that having eight NVMe SSDs drawing close to the 3.3V rail might stress it, but with a 1200W Platinum PSU and a UPS, that seems unlikely. Right now, the system is idle and draws under 250W thanks to the OLED display on my Thor 1200W monitor. Playing D4 uses significant CPU and GPU power, but stays well below the PSU’s capacity (around 550W). I saw online that many users face similar problems in Windows – no definitive solution seems available. Most advice suggests tweaking SATA drivers or adjusting Intel Rapid Storage settings. My TRX40 doesn’t support IRT features and I don’t use SATA devices, which are still detected in Device Manager. I’ll try the game again, this time with files from a different drive, to see if the issues continue. One more note: I turned off DOCP, set primary DRAM timings to 16-18-18-36, and speed to 3200MHz. FCLK Auto is on, other RAM timings are auto, CPU clock auto, and RAM voltages auto. So apart from that, I’m letting the motherboard decide the rest during testing.