[OFF-TOPIC] Gurobi vs Mosek via Yalmip

[G]

Hi,

does anybody know about performance comparison between Gurobi and Mosek used through Yalmip ?

I am dealing with Mixed Integer LP formulated thanks to Yalmip Big-M support and, on my particular instance of the problem, Mosek is incredibly slower than Gurobi.

[Johan Löfberg]

MILPs are intractable, so it is common (and expected) that some problems are solved efficiently using some solvers but slow on other, and vice versa.

[G]

Ok, thanks for the answer.

By the way I am talking about

Gurobi - 4 CPUs laptop, 4 Gb RAM -> approx. 1h30min
Mosek - 16 CPUs server, 64 Gb RAM -> more than 1 day !!

Exactly same Matlab code, of course

[Erling Andersen]

The current status of MOSEK is that for linear MIPs then Gurobi is faster than MOSEK in most cases. MOSEK shines in other areas. If you do not care about anything else than speed then use Gurobi/CPLEX/Xpress for linear MIPs. 

If your problem is a conic MIP we should do quite good though but Johans comment always apply in the MIP area.

If you want to use MOSEK because of price or another feature, then contact MOSEK support and we might be able to help to get better speed for your problem