Umm, way too much paste, and to the gentleman stating the gap is bigger than a standard PC, yes you are correct on some, but others make just as good of contact. You should apply the appropriate amount to each area.
Easy enough to check with carbon / transfer paper. You can make yourself "test shims" and shim each chip until your transfer touches. I never had to stack many shims at all in any of the S3, S3+, S4, or S5 systems I have.
Too much paste acts as a heat insulator rather than heat transfer and you end up losing any potential gains, or making it worse. To know if your repaste job helped or hurt your system you need to take temps using a temp gun, or other measuring device before and after completing the work making sure everything is the same in both cases. On the S5 for example you aren't getting temperatures from the chip on the Antminer display, and those temperatures have more to do with ambient air being moved past the chips and heatsink than the chip itself.
Of course you need to check again after a couple of days when it has really "set up" / "burned in"
Remove the sides, make a graph representing the miner broken down X amount of times in small squares, check the temp in each square and report it. Do your repaste job and repeat the same test. Then give it the normally recommended 48 hours to sit up and take your readings again.
In my opinion and experience if there is so much paste on the chip that when you put the heatsink against it the paste is pushed out and covers a ring around the outside of the chip you have went way too far and should cleanup, then start over.
You certainly are not helping the situation by insulating the heat in, and most threads like this aren't measuring properly.
I hope that helps someone with something but even with Asics when discussing thermal paste, less is more. Don't make a blanket of paste to cover the chip, make a bridge for the heat to travel across.
One tiny dot on most chips are fine, and some you may need a small amount more. Try the transfer paper. It does not add much time to an already tiring procedure and why not be accurate when performing a task such as this?
As the other gentleman stated, about the size of half a grain of rice. You can add more if it is really needed, but remember, more can harm the very thing you are attempting to correct and is the most common cause for a lack of better temps. With every other variable equal try one hash board the way you did it and the other the way many of us are saying, and show everyone what you find?
That sounds like a good project to add to the list.
In fact I've seen tests done where using too much paste results in reduced effectivity.