TRIM and Garbage Collection in Consumer SSDs: How They Maintain Speed

Consumer SSDs rarely deliver the same performance forever. A new drive may finish a large file copy in seconds, only to slow down once its fast write cache has filled. Later, repeated editing, game installations and file deletions can change sustained write speeds or latency. These shifts are normal, although a drive that never returns to a useful level of performance may have a firmware, cooling or hardware problem.

Understanding TRIM and garbage collection in consumer SSDs explains much of this behaviour. Both processes help the controller keep NAND flash organised, but they work at different levels. TRIM is a command sent by the operating system, while garbage collection is an internal controller function that reclaims blocks and balances wear across the flash.

Why SSD speed changes after use

The first burst of speed usually comes from an SLC cache, a small area of NAND that temporarily accepts writes as if it were faster memory. Manufacturers often quote impressive sequential figures based on this short-lived cache. Once the cache becomes full, the controller must write directly to multi-level NAND, where each cell stores several bits and programming takes longer.

Garbage collection also influences performance as the drive fills. NAND pages can be programmed but not individually erased; an entire block must be erased before it can be reused. The controller may need to copy valid pages out of a partly used block before erasing it. This extra work is called write amplification and can reduce throughput or increase latency, particularly on budget drives with less over-provisioned space.

Thermals are another factor. Australian builders in Brisbane and Sydney can encounter hot summer afternoons, while Melbourne’s variable weather can push warm laptops and compact desktops into higher temperature ranges. SSDs need less cooling than CPUs, but sustained writes can still make some controllers throttle. Case airflow, drive placement and ambient temperature therefore belong in any serious performance assessment.

How TRIM keeps the controller informed

A filesystem does not normally overwrite the same physical NAND location when you replace a file. It allocates another logical block address and updates its own map. The SSD’s flash translation layer then maps that logical address to a different physical page. This protects wear life, but it also leaves behind pages containing old data that must eventually be cleared before their blocks can be reused.

TRIM, also known as the discard operation, tells the controller that particular logical blocks are no longer needed. The controller can mark the corresponding pages as invalid and schedule the relevant blocks for erasure. Without that notification, it may have to inspect occupied blocks, identify which pages remain valid and copy them elsewhere before a block becomes available. Timely TRIM reduces unnecessary housekeeping and helps sustain write performance.

The command is not an instruction to erase one NAND page immediately. Most operating systems batch discard requests, and the controller decides when and how to act on them. TRIM also does not provide a secure wipe: wear levelling and over-provisioning can leave copies in areas the operating system cannot directly address. A drive with encryption should use its manufacturer’s sanitise facility or an appropriately verified secure-erase method when data must be unrecoverable.

What garbage collection does under the hood

Garbage collection is the controller’s internal block-cleaning process. It searches for blocks containing invalid pages, moves any live data into other blocks and erases the original block so that it can enter the free pool. The controller can then use that space for incoming writes without first performing as much work during the next transfer. Static wear levelling may also relocate data that has not changed for a long time.

TRIM improves garbage collection by supplying timely information about unused blocks. It does not replace the process, because the controller must also reclaim pages invalidated through ordinary overwrites and coordinate erase cycles across the entire drive. Internal collection continues whether or not the host sends TRIM, while host-generated discard can make it more efficient and predictable.

Controller design still matters. Some rely on a dedicated DRAM cache for their mapping tables, while DRAM-less models use a small portion of the NAND itself as a host memory buffer. Over-provisioning, wear-levelling algorithms and the amount of spare flash also affect how much work is required. Firmware-specific reports and long-term owner observations around the hardware community resource can add useful context when comparing drives with similar headline specifications.

Operating systems, filesystems and everyday workloads

TRIM support depends on the complete storage path. Windows sends TRIM on NTFS volumes, and its Optimize Drives utility can schedule the operation for solid-state drives. Linux supports discard with common filesystems such as ext4 and XFS, usually through periodic fstrim runs rather than a discard command for every file deletion. macOS uses TRIM with APFS. USB enclosures, RAID controllers and virtualised storage may delay, transform or discard the command, so the drive’s full specifications do not always describe its behaviour in every setup.

Everyday use matters as much as the operating system. Australian PC gamers downloading large game installers over fixed-line fibre or NBN connections, and hybrid workers maintaining growing Steam libraries, create repeated bursts of writes and deletions. Content creators working with 4K timelines can stress sustained throughput even when average benchmark results look strong. Systems with enough memory can also avoid swapping, so Hardware Hounds’ 32GB DDR5 kit guide is relevant for creators whose workloads benefit from keeping active projects in RAM.

Local pricing makes careful comparisons especially useful. Australian listings normally quote capacities in terabytes and prices in Australian dollars, often including GST, so cost per terabyte reveals value more clearly than sticker capacity alone. Warranty terms should be checked alongside the drive’s controller, NAND type and expected write endurance. If a drive develops a fault covered by its warranty, retailers and manufacturers must also comply with the Australian Consumer Law and applicable consumer guarantees.

Practical ways to test and maintain performance

A fresh-drive benchmark measures an optimistic state that may last for only a few minutes. A more useful comparison fills the SSD with representative data, performs a large write workload and then repeats the test. Record sustained throughput as well as latency where possible, since responsive performance under load matters for installations, exports and compiling. Synthetic tools can help, while fdisk, fio and DiskSpd provide more control for technically inclined users.

Before blaming garbage collection for a slowdown, check the interface, driver, filesystem and thermal conditions. A SATA SSD behind a slow USB bridge, a laptop with restricted airflow or a nearly full drive may perform differently from the same product installed natively. Firmware updates can improve power management and workload consistency, although users should follow the manufacturer’s instructions and maintain a current backup of important data.

A short maintenance checklist covers the most useful checks:

Confirm that TRIM is active on the filesystem and ensure any USB or RAID layer passes discard commands through correctly.
Test after a realistic fill level, because results from an empty drive may rely heavily on the SLC cache and spare space.
Compare sustained speed and latency using the same file types and workload whenever possible.
Keep the drive within its intended temperature range, especially during long transfers or media exports.
Track warranty endurance and back up critical files, since housekeeping improves consistency but cannot prevent every hardware failure.