pSLC Cache / Write Intensive Usage

When copying games from your Steam Library, or other very large files (>10 GB), you may notice that write speeds on your SSD start at full speed, before dropping considerably. The underlying reason is that modern drives have small, fast caches, that soak up write bursts to improve performance. In the fairly uncommon scenario of writing data that's too big to fit into these caches, the drive will have to write data directly to flash, and it will probably juggle some out of its write cache at the same time, which can result in a significant loss of write speed. Newer TLC drives use part of their capacity in SLC mode for increased performance. This test can reveal the size of that pseudo-SLC cache.

The testing methodology used on this page simulates such a scenario. We write a sequential stream of 1 MB blocks to the drive in a single thread, like a typical file-copy operation would do, and measure write speeds twice a second. Please note that some other publications test this at higher queue depth, which will yield bigger numbers, that are less relevant to real-world scenarios. The drive is fully erased before testing, to ensure any caches are emptied. This test generates a heavy write load with lots of data, something that most people don't do regularly—typical consumer workloads involve shorter bursts of write activity.


Write speed starts out at 4.5 GB/s, which is pretty good for single-threaded write speeds on a Gen 4 SSD. These speeds are sustained until 508 GB have been written, which means the drive will fill all its capacity in SLC mode first. Once the SLC cache is full, write speeds fall off a cliff, to just 200 MB/s, which is extremely low. Filling the whole capacity completes at 307 MB/s on average, one of the worst results that we've ever seen on an M.2 NVMe SSD and only half that of the NV2.