Bench
Every measurement published here was taken on this rig. Where a note gives a method rather than a number, it is because the number is a property of the sample, not of the procedure.
| Board A | X670E, dual Promontory 21, AGESA 1.2.0.3c |
| Board B | Z790, LGA1700, ME 16.x |
| Memory | 2x16 GB and 4x16 GB DDR5 UDIMM, Hynix A-die and M-die |
| Cooling | 360 mm AIO, open bench, ambient logged per run |
| Power | 1000 W ATX 3.0, single rail |
| Programmer | CH341A (3.3 V modified) and Raspberry Pi via spidev |
| Meters | bench DMM at the socket probe points, 100 MHz scope for transients |
| Thermal | K-type thermocouples on the inductors, IR for survey only |
| Host OS | Linux 6.12, lm-sensors, turbostat, pciutils |
Logged conditions
Ambient temperature is recorded at the start and end of every run and reported with the result. An open bench runs materially cooler than an assembled case, which matters most for refresh timing and for VRM thermals — both notes say so where it applies.
What is not here
No scores, no comparison charts, no product recommendations. The rig exists to make measurements reproducible, not to rank parts.