That said, your CPU is distinctly unusual and so what works/doesn't work for you might not apply to those of us with rather more pedestrian processors.
Not only that, I made a mistake. It is a dual processor system, the L3 cache is per processor, not total, so it is 55 x 2 = 110 MB, i.e. 55 threads by the 2MB rule which is larger than the total number of threads available so 44.