Solver - macro loop - want it faster

So it's taking Solver a couple of seconds per row -- that doesn't seem unreasonable. The code isn't pretty, but it's adequate, and not the cause of the performance issue.

You might look to see if you have a lot of volatile formulas that could be replaced by non-volatile equivalents, so they don't need needlessly recalculate as Solver iterates.

I will agree with shg, the code itself seems functional and completing within a reasonable time frame. If you need to speed this up, you might need to go a completely different direction.

You say that you are not very experienced with VBA. Are you likewise inexperienced with spreadsheet programming in general? Other programming languages? How about the math/algebra associated with this problem?

I note that your "objective function" (column T) is essentially a "least squares" type objective function -- suggesting to me that this problem might reduce to a simple "linear regression" that you can solve using the LINEST() function instead of Solver. The challenge here is exploring the function (J:N) to see if that function can be linearized to something in the form y=mx+b (it looks to me like it should -- you may need to spend some time with the algebra of the function). This will probably generate the most improvement because, on my machine, a 1000 LINEST() functions doing 1000 different regressions can calculate near instantaneously.

One complication is the presence of the constraints for z and mu. In both of the examples you gave in the other thread, z wants to exceed the upper limit of 0.06, so Solver fixes z at 0.06. Do you have enough experience with the 70000 regressions you need to do to know how often z and mu will naturally fit within their constraints? If we can say that z will always be 0.06, we can simplify the problem to a single parameter, and perhaps speed it up that way. If not, you will at least need to figure out how to work those constraints into the regression. Off the top of my head, that means a) performing the regression, b) testing the constraints, c) if either parameter is out of bounds, perform the regression again constraining the offending parameter to its max/min value. Even performing the regression twice like that will probably be much faster than Solver and VBA.

I haven't done the actual work (obviously), but that would be a quick overview of how I would try to approach this. If this works, you can bypass Solver and VBA, and program the whole problem using spreadsheet functions, and, I expect, compute these much faster than 500 in 20 minutes.