So I had some time to go through old eBay listings... because getting the information I needed another way was proving difficult. I needed board shots. On the left (pink background) is the same board I have in my base model 2020 iMac that came with an i5-10500 and a Radeon Pro 5300 4GB GPU. On the right is a board shot from the next tier up, but not the highest tier. This variant came with an i5-10600 and a Radeon Pro 5500XT 8GB GPU by default, and could be configured with an i7-10700K or an i9-10910. I wasn't able to find a board shot from the highest tier iMac, but it came with an i7-10700K and a Radeon Pro 5500XT 8GB GPU and could be configured with an i9-10910 and up to a Radeon Pro 5700XT 16GB GPU. As you can see, there are some power stages missing from the base model. I am no expert, but I count at least 3 VRMs missing. That means that the mid-tier and higher boards were probably overbuilt for any configurations with an i5, but could safely handle either the i7 or i9 chips if the end customer decided to upgrade at time of purchase. Furthermore, the third photo shows the underside of the heatsink that was designed to cool both the CPU and the GPU, regardless of the configuration. As you can see, the CPU, GPU, and GPU VRAM chips make contact with the heatsink, while the VRMs do not.
With this information in mind, I can make a few fair assumptions:
1. IF the VRMs on the mid-tier and higher boards were over-built for the i5, then the VRMs on the base-tier board is under-built for anything higher.
2. IF the power consumption of the i7-10700 and i7-10700K are near identical under a full load, then they can be treated as near identical for the purposes of considering their suitability for an upgrade.
Thus, under a full load scenario there is a chance that the VRMs on the base-tier board will have to pull more power than they were designed to; and, because they are not actively cooled, will run hotter than intended by Apple. This introduces a level of risk that can only be mitigated by avoiding full-load scenarios, in which case the upgrade is pointless.
As such, I think this project is dead in the water... because I am not willing to risk burning up these VRMs.


