specs

PD has advantages when you use the cables at 2A for a long time. But that is typically for charging a phone or a laptop in a short time.
 
With a PD-adapter the power consumption is pretty bad for the Rock5.
With a constant Voltage power supply it is only bad in combination with NVMe and good in combination with eMMC.
When powered of, without removing the power cable, the Rock5B still uses about 0,5W (slightly worse than a RPI4, which is already bad).
 
When using eMMC the Rock5B uses about 1.8-2W in idle. When you are using a general 30-65W PD adapter make it 2.4-3W in idle (the DC-DC conversion from  20V to 5V is very inefficient, if you have a small and cheap PD adapter this adds to the  inefficiency).
 
When using an NVMe you could get the power consumption as low as 4W in idle, with a 12V constant voltage adapter.
CPU 0-3: 1200 ( 600 Mhz - 1800 MHz / conservative) CPU 4-5: 816 ( 408 Mhz - 2400 MHz / conservative) CPU 6-7: 600 ( 408 Mhz - 2400 MHz / conservative) GPU : 300 ( 300 Mhz - 1000 MHz / simple_ondemand) DMC : 1560 ( 528 Mhz - 1560 MHz / dmc_ondemand) Here the pcie_aspm (NVMe connection) is set to powersupersave.
 
With the pcie_aspm set to performance and all the cpu and gpu settings to max you get 6W in idle.
CPU 0-3: 1800 (1800 Mhz - 1800 MHz / performance) CPU 4-5: 2400 (1000 Mhz - 2400 MHz / performance) CPU 6-7: 2400 (1000 Mhz - 2400 MHz / performance) GPU : 1000 ( 300 Mhz - 1000 MHz / performance) DMC : 2112 ( 528 Mhz - 2112 MHz / performance) The Kingston A2000 might not be the most energie-efficient NVMe, but other brands generally also focus on performance instead of efficiency.
 
The above summary is made by parsing the settings from:
  /sys/devices/system/cpu/cpufreq/policyX/
  /sys/devices/platform/fb000000.gpu/devfreq/fb000000.gpu/
  /sys/devices/platform/dmc/devfreq/dmc/
  /sys/module/pcie_aspm/parameters/
(You can also use the sys-tree to quickly change the settings with a script).
 
Ondemand and conservative are generally similar for power consumption in idle mode.
 
Earlier experiences that low power consumption and NVMe are a bad combination is true for the Rock5 as well. PowerSuperSave can not be considered a stable option for all NVMe drives. But still with lowpower solutions NVMe might be preferable to eMMC since it is slightly better protected against data loss.
 
In performance mode, during a compilation, the consumption easily reaches 13W (with NVMe, without large USB users, without much GPU use).
(I use a cheap power supply with monitoring to get the general idea of power consumption, not a calibrated solution).
 
Finally you could use a constant power supply of 5V to lower the power consumption in idle. I would not recommend that , I think optimizing for 1.5A at peak performance is better, resulting in 9V or 12V for most cases. (Instead of PD, QC is probably cheaper and more power efficient).

I think the power supply might deserve it's own topic.
 
Like balbes mentioned PD is meant for devices with internal battery. There it saves the battery during fast charging by preventing overcharging.
The Rock 5B is further 'borked' since the negotiation starts after the system starts. The advantage is it can be open source (more or less), but the big disadvantage it is so slow most PD chargers stop negotiating during system start in the meantime.
 
If you realise that the power supply in the Rock 5B (IP2315) is just a step-down converter with some logic and a PD car-adapter adds an step-up adapter to the equation you might realise it makes no sence to operate the power input at 20V. While the resistance in the cable might be minimized the power loss in the step-down and the step-up converters are maximized. Most power supplies can achieve 95% efficiency, but cheap step-up/step-down converters can only achieve that when V_in is almost equal to V_out. Normal PD-adapter should work more efficient than 12V car-adapters, but even then if they very efficiently produce 20V which the R5B very inefficiently transformes to 3.3-5V does that make sense?
 
A 20W PD car-adapter adds roughly 0.5W to the power consumption. With a 65W/20V PD car-adapter the adapter might add more, but the step-down converter in the R5B will definitely work more inefficient than at 9V.

I think the power supply might deserve it's own topic.
 
Like balbes mentioned PD is meant for devices with internal battery. There it saves the battery during fast charging by preventing overcharging.
The Rock 5B is further 'borked' since the negotiation starts after the system starts. The advantage is it can be open source (more or less), but the big disadvantage it is so slow most PD chargers stop negotiating during system start in the meantime.
 
If you realise that the power supply in the Rock 5B (IP2315) is just a step-down converter with some logic and a PD car-adapter adds an step-up adapter to the equation you might realise it makes no sence to operate the power input at 20V. While the resistance in the cable might be minimized the power loss in the step-down and the step-up converters are maximized. Most power supplies can achieve 95% efficiency, but cheap step-up/step-down converters can only achieve that when V_in is almost equal to V_out. Normal PD-adapter should work more efficient than 12V car-adapters, but even then if they very efficiently produce 20V which the R5B very inefficiently transformes to 3.3-5V does that make sense?
 
A 20W PD car-adapter adds roughly 0.5W to the power consumption. With a 65W/20V PD car-adapter the adapter might add more, but the step-down converter in the R5B will definitely work more inefficient than at 9V.

I think the power supply might deserve it's own topic.
 
Like balbes mentioned PD is meant for devices with internal battery. There it saves the battery during fast charging by preventing overcharging.
The Rock 5B is further 'borked' since the negotiation starts after the system starts. The advantage is it can be open source (more or less), but the big disadvantage it is so slow most PD chargers stop negotiating during system start in the meantime.
 
If you realise that the power supply in the Rock 5B (IP2315) is just a step-down converter with some logic and a PD car-adapter adds an step-up adapter to the equation you might realise it makes no sence to operate the power input at 20V. While the resistance in the cable might be minimized the power loss in the step-down and the step-up converters are maximized. Most power supplies can achieve 95% efficiency, but cheap step-up/step-down converters can only achieve that when V_in is almost equal to V_out. Normal PD-adapter should work more efficient than 12V car-adapters, but even then if they very efficiently produce 20V which the R5B very inefficiently transformes to 3.3-5V does that make sense?
 
A 20W PD car-adapter adds roughly 0.5W to the power consumption. With a 65W/20V PD car-adapter the adapter might add more, but the step-down converter in the R5B will definitely work more inefficient than at 9V.