osm2pgsql tablespaces on SSDs.

Hello,

I'm not sure if this is a simple question with a straightforward answer, but I'll ask anyway.

I know for spinning HDDs splitting the tablespaces across multiple drives has been shown to increase performance when loading the planet file into postgres using osm2pgsql, but what about SSDs?

All things being equal, which of the following options is faster (or more efficient) when farming out the 4 osm2pgsql tablespaces (i.e. main_data, main_index, slim_data, and slim_index):

1. a single large (1TB) SSD with 1 partition.
2. a single large (1TB) SSD with 4 partitions (one for each tablespace).
3. 2 medium sized (512GB) SSDs (one for data and one for indices).
4. 4 smaller (<=512GB) SSDs (one for each tablespace).
5. Maybe some other configuration?

I have a 1TB SSD I can test option 1 and 2 with, but if options 3 and/or 4 are not any faster, I'd rather not spend the money to find out.

Has anyone benchmarked this?

Thanks,