It has strong multithreading support including tuning to avoid two threads writing to the same cache line (which causes cache line bouncing and greatly bottlenecks parallel throughput).
Other caches do not broadcast notices when they discard cache lines, and this cache could not use such notifications without maintaining a count of the number of shared copies.
Recall that the cache lines are tagged with the real address of the original persistent memory location; cached results are later transparently returned there.