G.Skill Trident Z5 NEO RGB 32GB (2 x 16GB) DDR5 6000 CL30-38-38-96 1.35V F5-6000J3038F16G AMD Expo Issue with 7950X3D
G.Skill Trident Z5 NEO RGB 32GB (2 x 16GB) DDR5 6000 CL30-38-38-96 1.35V F5-6000J3038F16G AMD Expo Issue with 7950X3D
Dragonosis
12-28-2023, 09:10 AM #1Build: 7950X3D, 4090, 360mm AIO with 3 case fans, Sabrent Rocket 4 Plus, ASUS ROG x670e crosshair hero motherboard. I have Expo One enabled and the PC is working perfectly. However, I've experienced a few restarts recently (only a brief blue screen once). Running Karhu and Prime95 both failed after about 45 minutes. Prime95 reported a fatal error about rounding issues during FFT processing, suggesting a hardware problem. The stress.txt file mentioned thermal changes and CPU voltage adjustments didn’t help. I’m considering whether to tweak the thermal paste or leave it as is, and I’m curious if this issue is common among users with this RAM. Should I ignore it since my games ran fine, or should I wait for a BIOS update or new chipset drivers? Also, I’m wondering if downclocking would be safe. Thanks in advance, and I’ll keep you updated. P.S. I’m also running it alongside another system. Is this something urgent to fix before the next update?
Machi_Gamz
12-30-2023, 11:28 AM #3RAM instability doesn’t harm the hardware directly. The main risk is losing your Windows installation, which would mean reinstalling it—a process that’s generally manageable. Be aware that ASUS has reported problems with recent BIOS updates for memory support. You might succeed with 6000, but it could also fail. The problem appears to relate more to the 6000 ratio; you may find better results at 6200, though adjusting some voltage settings could improve stability. For memory speed, higher frequencies are preferable, while lower FCLK overclocks are better. This varies by which CPU part is unstable. At 6000MT/s, the default FCLK should be 2000MHz, which should run smoothly even at 1.3V. Fixing the FCLK at 2000MHz, setting SOC to 1.3V (the highest AMD recommendation), and testing could help. If that doesn’t work, lowering SOC to 1.2V might assist. If none of these steps succeed, try 1.25V and tweak other voltages. Remember, memory voltages like 6000CL30 are stable for Hynix chips, which tolerate over 1.5V, but staying at or below 1.4V is wise to avoid cooling issues. Other options include VDDIO_MEM, VDDP, VDDG, and VDD Misc. Voltages—usually they don’t matter much except for FCLK overclocking. Adjustments of 0.005V or smaller are usually safe, aiming for 0.01V steps if needed. If you’re concerned about voltage limits, 6000CL30 is a reliable memory type, and higher voltages are generally safe for memory modules.
Thank you for sharing all this detail. I was concerned about affecting the SOC since I heard X3D chips could fail, and during setup with the AIO cold plate I struggled to lower it properly—had to apply pressure to move the top latch. While doing so, the pre-applied thermal paste touched the CPU. I eventually had to lift it back and reposition it several times, which made me think a paste reapplication might be needed. Currently, temperatures range from 29°C to 40°C during idle, with gaming at around 60°C and underload at 40°C while browsing. I ran Karhu and Windows memory diagnostics and encountered errors, suggesting the issue might be related to RAM rather than the CPU. I’m planning to follow your advice and see if it stabilizes. My benchmark scores are near average or slightly above, and I haven’t experienced crashes except during YouTube playback. I’m hopeful you can help resolve this and will update you as I make adjustments. Thanks a lot—I’ll try right away!
BlitzSquadHD
01-07-2024, 04:05 PM #6Switching to DIMM A1 and B1 as well. Appreciate your assistance once more. Reassembling now, finishing later.
TwistedHeartt
01-12-2024, 11:00 PM #7It might not function well on A1 and B1 due to the structure of memory systems. You can attempt it, but you’ll likely need to revert to A2/B2 afterward.
I made sure my RAM was properly seated. I haven’t re-seated it yet; I just set VDDCR_SOC Voltage to 1.3 and am about to run Prime95. I also noticed a warning in the event viewer earlier today: the embedded controller returned data when it wasn’t needed. It seems the BIOS might be trying to access the EC without syncing with the OS. This information will be disregarded. No further steps are required, but you should contact your computer manufacturer about updating the BIOS.
)
george_griveas
02-13-2024, 11:27 PM #10I simply left Expo on and increased the VDDCR_SOC voltage to 1.3. After two tests, one failed at 45 minutes, but both Prime95 Small FFT AVX and Karhu passed after an hour and ten minutes. I ran Karhu for an hour and ten without issues, using the same settings for over two hours with no problems. So Expo is now running smoothly. Thanks a lot! I also have another concern. In HWiNFO I noticed a brief spike at 128C on CPU (Tctl/Tdie). Uncertain if it was an error since it quickly returned to normal. Could be a HWiNFO issue. During the Prime95 test, temperatures stayed below 79°C, with CCD1 and CCD2 under 74.8°C and 67.1°C respectively—averaging around 63°C. What are your thoughts? Also, should I try running a blend with AVX enabled for about 6 to 8 hours tonight to be safe, or stick with Large? Your decision is welcome )