I think that all the instructions in a RISC take the same time what
makes me believe RISC is associated with synchronous.
It's not true that all instructions in a RISC CPU take the same time. Even in the classic RISC pipeline, the pipeline will stall in some circumstances. Pipeline interlocks and control hazards are the ones that Hennessy and Patterson go into some detail about, but they also mentioned data cache misses.
Even if you ignore those complications, it's unrealistic to design a general-purpose CPU where every instruction takes the same time these days. Programmers want floating point, and SIMD operations, and so on.
Can you tell me if RISC is associated with asynchronous and CISC is
associated with synchronous?
On, that's easy: They're not.
Truly asynchronous CPUs are rare, for a few reasons. One of them is that you can't (currently) prototype asynchronous logic on a FPGA, and it's much harder to express in Verilog or VHDL than synchronous logic.
Having said that, it is easier to implement a RISC pipeline in asynchronous logic than a CISC pipeline (e.g. Sproull's Counterflow Pipeline). However, event his is a bit misleading because the boundary between RISC and CISC is quite fuzzy these days. Most modern CISC CPUs use RISC cores internally, and dynamically translate the CISC instruction stream into RISC code. The main reason for this is that the same features which make RISC attractive (e.g. fixed-size instructions which can be buffered) are also the ones which make superscalar execution work.
So to the extent that asynchronous logic works, modern CISC CPUs can benefit almost as much as RISC CPUs can.