Since i'm interested if this would work or not, I configured a power sipper x300 build using a 5700g on the Deskmini as follows to simulate a "quiet" Jupiter x300:
-Parked 5 cores, 3 cores enabled (5600g = Park 3 cores, 3 enabled)
-Disable SMT (Hyperthreading)
-Undervolt to 1.05v @ 3.8ghz manual clock
-GPU clock unchanged @ 2ghz
This is using the Noctua l9a (NFa9x14) so keep in mind the cooling capacity is going to be better than the Jupiter.
Result #1 (CPU workload)
21.9w Package power in Cinebench with performance similar to 1165g7 Tiger Lake laptop. Room to enable another core or two, or SMT.
About 60% fan speed of the NFa9x14 and temps are under 50c, it would be hotter on the Jupiter but it's hard to say by exactly how much but it looks promising.
Result #2 (Gaming result - Witcher 3 720p High):
3 Cores @ 3.8ghz 1.05v
36w GPU + 9W CPU (3 cores)=
45w on sensors
At 45 watts, this wouldn't be as quiet, but it still should be possible on the Jupiter without 100% fan speed. Because I have noticed the GPU has better heat dissipation on the Zen3 die, it doesn't get hotspots as bad as the the CPU cores. So even though the package power draw is almost double, the hotspot temps won't be as bad as a CPU benchmark.
Now the above compared to the baseline (my x300 config) to test for lost performance.
8 Cores, 4.4ghz @ 1.3v manual locked
44W GPU + 17w CPU =
61w (Note: GPU wattage may be higher because SOC amperage increased)
Same fps! There is no perceptible difference in gaming results by dropping to 3 undervolted and underclocked cores, at least in this game since the CPU utilization isn't very high (60%).
Cyberpunk was at 90% util so maybe some games would benefit from a bit higher clock speed, or by enabling 4 cores.
Because of my memory configuration I could not test an undervolted SOC, it was at relatively high 1.15v in both tests.
Actual gaming results would be lower than 45 watts when using less SOC voltage. Up to 5 watts less on iGpu if using 900mv SOC.