In recent systems, the GPU heat load is significantly higher than the CPU heat load (often more than twice as high), and in addition the GPU heat sink air flow is significantly poorer than the CPU's. So you really shouldn't be concerned about an "upside down" CPU heat sink; it will work just fine.
That's right, but they can rely on an higher temperature drop between the heat source and the heat sink, for GPUs are usually able to run hotter than CPUs nowadays. I.e. if the sink shall dissipate the same heat load and can rely onto a higher limit in terms of heat source temperature, a lower overall conductance is needed.
As said above, heat pipes with mesh wick are able to transfer heat with ease when horizontal; they are designed to work against gravity
, but it is quite obvious that the maximum heat transfer efficiency is guaranteed when you use them as thermosyphons. As far as an heat pipe is working as an heat pipe (so as far as it is not put into a condition which goes beyond its heat transfer capability) you would not encounter any drop in efficiency and you can consider it as a nearly isothermal device, nevertheless the thermosyphon effect enhances its heat transfer capability (think about it as a boosting effect).
The main concern is to design your system relying on performance achieved into the worst condition case, thus it is not a priori true that your expectations would be met if you think the CPU temperature will be the same in upside down configuration. Moreover, the heat source shall not go beyond the heat pipe maximum heat transfer capability, which depends on its orientation. Summarizing, you should expect a drop into the efficiency and a drop into the maximum heat load manageable.
Several years ago I tested some twicked heat pipes, based on a design driven by computing industry. At that time all the computing heat pipes were of mesh type. I did some research and I see that sintered powder wicks have been introduced into the computing market since that time. The problem with those high density wicks is that the increasing in heat transfer capability is achieved by decreasing the maximum length at which the efficiency is acceptable. That is because the capillary pump effect is progressively counteracted by drag forces affecting the fluid soaking the wick. The more the wick is dense, the higher is the maximum heat load, the higher are the drag forces, the lower is the maximum length. The production cost is also pretty higher with respect to the mesh. I guess that these points make the powder wick heat pipes acceptable on high end graphic cards, which demand high dissipation capability and are not cheap at all!