Also known as: Non-Volatile Memory Express

NVMe (Non-Volatile Memory Express) is a protocol for talking to fast solid-state storage directly over a PCI Express link, designed from the start for flash rather than spinning disks.¹

Overview

Older interfaces such as SATA used the AHCI command set, which assumed the slow seek times of a hard disk drive and offered a single shallow command queue. NVMe instead exposes many deep queues that a flash-based SSD can service in parallel, with far lower per-command overhead. Drives commonly appear as an M.2 stick plugged straight into the board’s PCIe lanes, so the CPU reaches the flash without a legacy disk controller in the path.

What it’s for

NVMe exists to stop the interface from being the bottleneck once the medium is flash. A SATA SSD tops out around 550 MB/s; an NVMe drive on a few PCIe lanes runs several times that, with much lower latency under load. For a GopherTrunk node decoding many talkgroups at once, an NVMe SSD ensures that writing decoded frames and IQ snapshots never stalls the pipeline — though for a simple capture node a plain SD card or SATA SSD is usually plenty.

Sources

1. NVM Express — Wikipedia, on the NVMe storage protocol and its advantages over AHCI/SATA. ↩

See also

• Solid-state drive (SSD)
• Flash memory
• PCI Express (PCIe)
• Hard disk drive (HDD)
• Storage (SSD
• Memory hierarchy