Introduction to Vlatch

The Apex and Extreme Z590 motherboards from ASUS brings an interesting new feature named “Vlatch”. It brings on-board circuitry for CPU Core Voltage min/max peak detection at much higher speeds than what is normally possible with . The values are recorded by the embedded controller (EC) on the motherboard and is visible in the latest HWInfo version. HWInfo is also able to reset the stored min/max values. This can help you find the ideal load-line setting for your overclock by minimizing over/undershoot which normally is not measurable by the user.

Test setup

  • Intel Core i9-11900K @ 4.8 GHz 1.3V set
  • ASUS ROG Maximus XIII Apex
  • G.Skill 3200C14D-16GTZR (2x8GB)
  • Corsair H115i
  • Tektronix DPO 3054 with differential probe to CPU die-sense
  • HWInfo64 v6.43-4370
Oscilloscope connected to the motherboard

Method

The system was set to use Manual CPU Core Voltage at 1.300 V to avoid power saving features to drop the voltage in idle. Before each test the Vlatch values were reset in HWInfo and the scope measurement history was cleared. A single run of Cinebench R20 was the performed after which the reuslts were recorded. The measurement resolution of the Vlatch feature is 6.25 mV and the vertical resloution of the scope is 10mV

Test results

Load-line Calibration Level 4

Vmin/Vmax/Vpp recorded by HWInfo with LLC=4
Oscilloscope capture at LLC=4
VminVmaxVpp/Delta
Vlatch/HWInfo1.063 V1.288 V0.225 V
Oscilloscope1.06 V1.32 V0.220 V
Summary of results with LLC=4

Load-line Calibration Level 8

Vmin/Vmax/Vpp recorded by HWInfo with LLC=8
Oscilloscope capture at LLC=8
VminVmaxVpp/Delta
Vlatch/HWInfo1.113 V1.288 V0.175 V
Oscilloscope1.12 V1.32 V0.200 V
Summary of results with LLC=8

Result analysis

The Vmin values recorded by the Vlatch feature seems to match quite well with the oscilloscope measurements, which is arguably the most interesting parameter to look at when it comes to processor stability. Further investigation is needed to see if this correlation matches with end user testing. The Vmax values differ by 34 mV in absolute terms. The recorded values for Vmax are also quite a bit lower than what was expected. The Vpp/Delta values are matching at LLC=4 but has differs by 25 mV at LLC=8. All in all the values match decently with the scope measurements and I look forward to seeing test results done by other users to see if it can help when figuring out their overclocks.