I tested the effects of enabling Samsung’s “Rapid Mode” (driver level topology aware RAM caching AFAIK) on my 1 TB 870 EVO SATA drive and saw a significant improvement. As always, YMMV.
Times below are for single plots on an Intel Core I7 6700, 4 threads, 6780 MiB buffer, Synology NAS final destination
How many tests have you run? I’m curious if the benefit is real. Rapid Mode uses your systems ram as a cache for your SSD. I suspect that your sustained read/write speeds will not benefit from this, given the amount of data and the fact it still needs written to the slow SSD.
This is a single test, but on the same system with all other parameters equal. More would be better, but e.g. looking at phase 4 with an almost 2x improvement I’d say results are at least fairly interesting and beyond mere noise. I’ll try to run a similar test with 2 concurrent plots as well.
Well, I have an brand new 1TB 870 EVO sitting right here… Maybe I’ll give it a try. I’ll have to use a lower powered system (one of my GPU mining rigs). I don’t want to stop any of my heavy duty plotters.
I find this a lot in Google’s cache, but can’t pin down the source.
RAPID consumes at most 25% of the installed DRAM, up to a maximum of 1GB, but will scale down resource usage and eventually revert to a pass-through mode if the CPU core(s) or DRAM is occupied with other system tasks. (This technology is really designed to make use of excess system resources)