Hi,
I’m testing RDNA undervolting behavior on a compact, airflow-optimized setup. I have solid experience with RTX 5000 series UV, but RX 9xxx is new to me.
The UV Profile:
- Voltage Offset: -80 mV
- Core Clock Offset: +100 MHz
- Power Limit: 0% (Stock)
- VRAM/Fans: Stock
Specs:
- Ryzen 5600GT (→ 5500X3D soon)
- Gigabyte Radeon RX 9600 XT GAMING OC 16GB
- Gigabyte A520i AC (PCIe 3.0)
- 2x16GB G.Skill TridentZ Neo CL16 3600 MHz
- PSU: Thermalright TR-TPFX 850 Platinum
Driver:
- Adrenalin Edition 26.3.1
Results (3DMark):
- Time Spy: +2.7%
- Port Royal: +2.6%
- Steel Nomad: +2.9%
- Time Spy Extreme: +2.5%
- Speed Way: +0.7%
Observations:
- Effective voltage drop under load: ~20–30 mV
- Slight VRAM temp reduction (few °C)
- Noticeably lower fan RPM (especially in heavier tests)
- Power draw reduced by ~5–10W
Data logging via HWinfo (CSV), then charted in Gemini.
Question:
- On RDNA, is it worth increasing Power Limit for better sustained boost, or is this already close to the efficiency sweet spot?
- Does further undervolting typically scale better than adding +PL on these GPUs?
Thanks!