Sometimes your choice of style is limited by the availability of the power switch type in your library.Originally posted in cdnusers.org by Tobing
However if you have options between the two, below are some considerations.
One analogy I would use is the comparison of wire bond vs. flip chip design.
1. For an already hardened block, ring configuration is much easier to implement.
You can just build a ring around the block instead of opening the block again for column.
2. Floorplans are much easier with ring configuration. Especially when you have large
blocks, memories or macros.
3. Cell placement is easier with ring configuration. This may lead to easier timing closure.
4. Less routing overhead on ring configuration. The vias attaching the power switch to
the always on grid does reduce routing tracks.
5. Column structure has advantage on IR Drop. Typically you can build a more robust
always on power structure on the upper layer with the column structure. Then drop the power
switch where it is needed. With ring you can suffer significant IR drop after the power switch
having a solid power after the voltage drop does not help much.
6. Also having power source in the middle will reduce IR drop by half compared to the edge.
Imagine if you only have one power source for a block, putting one in the middle guarantees that any
points in the block are no more than half the width away from the power source.
7. EM requirement for the power rail after the power switch could be less for the column style,
since the current can spread from the center to both sides Vs one side for the ring configuration.
8. Arguably the column configuration will use less area.
9. Column structure can do more options for power sequencing to reduce glitch on neighboring cells.
Hope that helps,