Why SSD Conditioning is Vital for Steady-State Benchmarking

This article explains the critical necessity of conditioning Solid State Drives (SSDs) with a full drive write prior to conducting steady-state hardware benchmarks. Brand-new or freshly erased SSDs operate in an artificially fast state that does not reflect real-world, long-term performance. By writing to the entire drive capacity before testing, reviewers and engineers force the SSD's garbage collection, wear-leveling algorithms, and cache mechanisms into their normal operating routines, ensuring that benchmark data accurately represents sustained, steady-state performance.

The "Fresh Out of the Box" (FOB) Illusion

When an SSD is brand new, or has just undergone a Secure Erase command, all of its NAND flash memory blocks are completely empty and ready to accept data. In this Fresh Out of the Box (FOB) state, the drive can write data directly to the flash cells without any preparation.

Benchmarking an SSD in this state yields exceptionally high performance numbers. However, this performance is temporary. Once a user fills the drive with data over normal use, the SSD can never return to this initial speed without wiping the entire drive. Therefore, FOB benchmarks are misleading for evaluating sustained workloads.

The Physics of NAND Flash and Garbage Collection

The primary reason SSDs slow down after a full drive write lies in how NAND flash memory handles data.

  • Writing Data: SSDs write data in small units called pages (typically 4KB to 16KB).
  • Erasing Data: SSDs can only erase data in much larger units called blocks, which consist of hundreds of pages.
  • No Direct Overwrites: Unlike traditional hard drives, NAND flash cannot overwrite existing data directly. The old data must be erased before new data can be written to the same location.

When an SSD is conditioned with a full drive write, every physical page on the drive ends up containing either valid data or invalid (deleted) data. To write new data during a subsequent benchmark, the SSD controller must perform Garbage Collection.

During this process, the controller reads the valid data from a mostly full block, moves it to a new location, erases the entire old block, and finally writes the new data. This process, known as the Read-Modify-Erase-Write cycle, introduces significant latency and reduces write speeds compared to an empty drive.

Exhausting the SLC Cache

Modern consumer and enterprise SSDs utilize a portion of their storage as a high-speed buffer, known as a pseudo-SLC (Single-Level Cell) cache. When data is written to the drive, it goes to this fast cache first, masking the slower native speeds of the underlying TLC (Triple-Level Cell) or QLC (Quad-Level Cell) memory.

Short benchmarks on an unconditioned drive only test the speed of this SLC cache. Conditioning the drive with a full sequential write completely saturates the SLC cache. This forces the SSD to write directly to the slower native NAND while simultaneously trying to flush the cache, revealing the drive's true sustained performance bottlenecks.

Triggering Write Amplification and Wear Leveling

A full drive write also forces the SSD's controller to actively manage drive health and data placement through two mechanisms that impact performance:

  1. Write Amplification: Because the drive must move existing data around to clear full blocks for new data, the actual amount of data written to the physical NAND becomes greater than the amount of data sent by the host computer. This overhead consumes bandwidth and lowers benchmark scores to realistic levels.
  2. Wear Leveling: The SSD controller actively distributes writes evenly across all physical blocks to prevent premature wear. This background management only becomes complex and resource-intensive once the entire drive has been written to at least once.

Achieving True Steady-State

Steady-state is the phase where an SSD's performance stabilizes and becomes predictable under a specific continuous workload. Without pre-conditioning the drive via a full drive write, a benchmark will record the transition from the artificially fast FOB state down to the steady-state, skewing the average results. Conditioning the drive ensures the transition phase is bypassed, allowing hardware testers to measure the exact, consistent performance a user will experience under heavy, long-term workloads.