Parallel fragmented SGL fetching for hiding host turnaround time

The present disclosure generally relates to reducing latency when fetching Scatter Gather Lists (SGL). Rather than fetching the required SGLs sequentially regardless of what SGL descriptor is needed, the data storage device fetches all of the last entries of each SGL segment in ahead of time after receiving the command, but before the read data is available. The data storage device will still fetch the previous entries in the segment. Once the last entries are fetched, the last entries are stored in a table where the earlier descriptors of each segment are stored as the segments are fetched. In so doing, parallel fetching allows the data storage device to fetch SGL descriptors as needed and reduces the latency.