This uses the 1.3M neurons dataset from 10x Genomics.
If you have at least 30 GB of memory (recently, tests indicated that memory usage could go up to 120 GB, we are investigating), run cluster.py to produce the following result producing in 6 hours on a small server using at most 2-3 2.4 GHz cores. If you subsample to 130K cells, this takes 16 min. The tSNE computation took about 4 hours using 8 cores. The clustering result is available here.
Visualizing and Clustering 1.3M neurons