I recently picked up a new PC with a Ryzen 9800X3D, MSI B850 Tomahawk MAX II, 32GB DDR5-6000 CL36 memory, RTX 4080 Super, Corsair RM1000e 1000W PSU, and a Noctua NH-D12L cooler. My goal is to keep stock performance while reducing temperatures, power use, and fan noise.
I was planning to enable the memory profile, try a negative Curve Optimizer value on the CPU, possibly set VSOC near 1.20V, and undervolt the GPU by adjusting its voltage/frequency curve in Afterburner. The technician who assembled the system advised against all of this, saying that memory profiles are technically overclocking, could cause instability, and that CPU undervolting might damage the motherboard.
Would you enable EXPO or XMP with this memory kit? Should I start with an all-core negative Curve Optimizer value or leave the CPU completely stock? Is setting a 94°C thermal limit worthwhile? Since there have been reports of Ryzen 9800X3D failures on some systems, should I manually limit VSOC to 1.20V even though I have an MSI board and an updated BIOS? Finally, is adjusting only the voltage curve in Afterburner the right approach for undervolting the RTX 4080 Super?
5 Answers
Run the system at stock first so you have a baseline. If temperatures or noise are actually a problem, try a modest negative Curve Optimizer value and test it thoroughly. Per-core tuning can produce better results, but an all-core starting point is simpler. An undervolt normally does not damage the motherboard, though an unstable setting can cause crashes, errors, or game failures. A 94°C limit is optional; the processor already has its own safeguards, and lowering the limit mainly trades some boost performance for lower temperatures.
For the RTX 4080 Super, Afterburner’s voltage/frequency curve is the usual method. Pick a target voltage and frequency, flatten the curve above that point, then test several demanding games while watching for crashes, driver resets, artifacts, and performance changes. Save the profile only after it is stable. You can also adjust fan curves, since the Noctua cooler and a well-tuned case may already keep the CPU comfortable.
There is no need to manually change VSOC just because of reports involving other boards. Keep the BIOS updated and check what the board is actually applying. If you do set a limit, use a sensible value and verify that memory remains stable; changing voltages without a specific reason can create unnecessary troubleshooting. Also, undervolting does not make a chip automatically safer if the setting is unstable, so test with demanding games and tools rather than assuming a setting is good because the PC boots.
I would enable the memory profile. On an AMD system that generally means EXPO, although some kits are advertised with XMP and may still work through the board's memory-profile options. Without it, the DDR5-6000 kit will usually run at a slower JEDEC default speed. It is technically memory overclocking, but it is a normal feature the kit is designed to support. If it causes crashes, disable it or try a less aggressive setting and test stability.
The safest approach is to change one thing at a time. Enable EXPO first, confirm memory stability, then try a small CPU adjustment if you still want lower temperatures. Start with fan curves before chasing tiny efficiency gains. Your hardware is already high-end, so reliability and noise levels matter more than squeezing out a few extra benchmark points.
Would you start with an all-core value, and would setting the thermal limit to 94°C provide any meaningful benefit?