Zenbook S 16 with Ryzen HX AI 370 not reaching s2idle.

I've been trying to debug high power usage on suspend (~30%/day, https://github.com/BNieuwenhuizen/zenbook-s16?tab=readme-ov-file#standby-battery-duration), and tried to debug this through amd_s2idle.py. It indicates that the laptop actually doesn't reach s2idle:

$ sudo python amd_s2idle.py 
Location of log file (default s2idle_report-2024-08-09.txt)? 
Debugging script for s2idle on AMD systems
💻 ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA (ASUS Zenbook S 16) running BIOS 5.35 (UM5606WA.308) released 07/17/2024 and EC 3.22
🐧 Arch Linux
🐧 Kernel 6.11.0-rc1-1-git-00269-gd9ef02e56f0f
🔋 Battery BAT0 (ASUSTeK ASUS Battery) is operating at 100.86% of design
Checking prerequisites for s2idle
✅ Logs are provided via systemd
✅ AMD Ryzen AI 9 HX 370 w/ Radeon 890M (family 1a model 24)
✅ SMT enabled
✅ LPS0 _DSM enabled
✅ ACPI FADT supports Low-power S0 idle
✅ HSMP driver `amd_hsmp` not detected (blocked: False)
✅ PMC driver `amd_pmc` loaded (Program 9 Firmware 93.9.0)
✅ USB4 driver `thunderbolt` bound to 0000:c6:00.5
✅ USB4 driver `thunderbolt` bound to 0000:c6:00.6
✅ System is configured for s2idle
✅ NVME Micron Technology Inc 2400 NVMe SSD (DRAM-less) is configured for s2idle in BIOS
✅ GPIO driver `pinctrl_amd` available
✅ IOMMU properly configured
How long should suspend cycles last in seconds (default 10)? 600
How long to wait in between suspend cycles in seconds (default 4)? 
How many suspend cycles to run (default 1)? 
Started at 2024-08-09 22:18:12.742055 (cycle finish expected @ 2024-08-09 22:28:16.742069)
Results from last s2idle cycle
○ Suspend count: 1
○ Hardware sleep cycle count: 1
○ Wakeup triggered from IRQ 9: ACPI SCI
○ Woke up from IRQ 9: ACPI SCI
○ gpe10 increased from 180 to 186
✅ Userspace suspended for 0:10:18.052243
❌ Did not reach hardware sleep state
🔋 Battery BAT0 lost 585000 µWh (0.74%) [Average rate 0.34W]

At this point I'm not sure what can prevent the HW from reaching s2idle and since apparently the SMU seems to be involved looking at the source of the amd_s2idle.py script, I thought it was worth reporting here.

The full log is attached. s2idle_report-2024-08-09.txt

The issue seems to happen every suspend cycle I've tried (will try multiple boot cycles in a bit).