Why do SSDs slow down at 80% capacity? Technical explanation and optimization methods

MarGib September 11, 2026

🌐 🇵🇱 Polski · 🇬🇧 EN

SSDs lose performance when filled beyond 80%. Why does this happen, and can it be fixed? We explain the technical mechanisms, present measurement data, and provide tips on how to optimize your drive's performance.

Ilustracja procesu zapisu i czyszczenia danych na dysku SSD, pokazująca mechanizmy garbage collection i write amplification. — Mechanisms of SSD operation at high capacity.

Why do SSDs slow down at high capacity?

The mechanism behind SSD slowdowns when filled above 80% stems from several key technical processes occurring within the controller and NAND Flash memory. The most important ones are:

Write Amplification (WA): To write new data into a partially filled block, the controller must read the existing data, update it, and write the entire block again. This process, known as write amplification, increases the number of write operations, which lowers performance. At high drive capacity, WA can increase several times over, which directly translates to a drop in speed.
Garbage Collection (GC): The process of automatically cleaning blocks of "dead" data (overwritten or deleted). GC works efficiently only when the drive has enough free blocks. When filled above 80%, the controller has less room to maneuver, which slows down the entire process.
TRIM: A system command that informs the SSD which blocks are unused (e.g., after deleting files). Without TRIM, the controller does not know which blocks can be cleared, which increases the overhead and slows down writes. Operating systems like Windows and Linux support TRIM, but its effectiveness depends on the drive's fill level.
Wear Leveling: An algorithm for evenly distributing writes across all NAND cells to extend the drive's lifespan. At high capacity, the controller has fewer options for distributing writes, which can lead to faster wear of certain blocks.

All these mechanisms are interconnected and influence each other. For example, high write amplification increases the load on garbage collection, which in turn slows down the writing of new data.

Is the 80% limit a universal rule?

Although 80% capacity is a frequently cited threshold, in practice, it depends on several factors:

Controller and NAND Flash type: Drives with newer generation controllers (e.g., PCIe 4.0/5.0) and TLC or SLC memory handle high capacity better than older models with QLC memory. For example, enterprise drives, such as the Samsung PM9A3, are designed to operate even at 90% capacity.
Drive capacity: Larger drives (e.g., 2 TB) have more blocks, which mitigates the slowdown problem. In the case of smaller drives (e.g., 250 GB), the performance drop may be noticeable as early as 70% capacity.
Workload type: Sequential writing of large files (e.g., copying movies) is less susceptible to slowdowns than random writing of small files (e.g., during software installation or database operations).

SSD manufacturers such as Samsung, Crucial, or Western Digital recommend leaving 10–20% of free space for optimal performance. However, it is worth remembering that these are general guidelines, and the actual threshold may vary depending on the drive model.

How much does performance drop at high capacity?

Measurement data from tests conducted by independent technology websites show that the performance drop can be significant. Here are a few examples:

AnandTech tests (2015): The Samsung 850 Pro (TLC NAND) drive showed a drop in 4K random write speed from ~400 MB/s at 50% capacity to ~150 MB/s at 90% capacity. This represents a drop of over 60%.
Tom’s Hardware tests (2021): The WD Black SN850 (PCIe 4.0, TLC) drive lost about 30% of its sequential write speed (from 5000 MB/s to 3500 MB/s) when moving from 50% to 95% capacity. 4K random writes dropped from 800 MB/s to 300 MB/s.
Puget Systems tests (2020): In professional scenarios, such as video editing, rendering time increased by 15–25% at 85% drive capacity.

The largest performance drop concerns random writes of small files, which is critical for the operating system and applications. Sequential writing of large files is less susceptible to slowdowns, but differences can be observed there as well.

How to measure SSD performance?

Several tools can be used to test SSD performance:

crystaldiskmark: A popular tool for testing sequential and random write/read speeds. It allows for quick comparison of results before and after optimization.
AS SSD Benchmark: A tool that evaluates drive performance in scenarios similar to real-world use, such as copying files or launching applications.
fio (Linux): An advanced tool for stress testing that allows for the simulation of various write/read scenarios.

Regularly testing the drive allows you to monitor its performance and detect problems early.

Does the operating system affect SSD performance?

Yes, the way the operating system manages the drive can have a significant impact on its performance, especially at high capacity. Here is how different systems handle this issue:

Windows

TRIM: Enabled by default since Windows 7. The system automatically sends TRIM commands once a week. You can force this manually using the Optimize-Volume -DriveLetter X -ReTrim -Verbose command in PowerShell.
Defragmentation: Disabled by default for SSDs (since Windows 8). Instead, the system performs optimization, which consists of sending TRIM commands.
Caching: Windows uses memory caching (Superfetch), which can mask drive slowdowns but does not solve the write amplification problem.

Linux

TRIM: Supported by the fstrim command (one-time call) or the discard mount option in the /etc/fstab file. Constant discard can lower performance, so it is recommended to use fstrim once a week (e.g., via cron).
File systems: Systems like ext4 or Btrfs work better with TRIM than NTFS. You can also use the F2FS file system, which is optimized for SSDs.
Block allocation: Linux provides greater control over block allocation, which allows for better drive optimization. For example, mounting with the noatime option reduces the number of metadata writes.

Linux offers more flexibility in configuration but requires manual optimization. Windows is more "plug-and-play" but may mask performance issues.

How to optimize an SSD at high capacity?

If an SSD is filled above 80%, there are several methods that can help restore its performance:

1. Freeing up space

The most effective way to improve performance is to free up disk space. Here are some practical tips:

Delete unnecessary temporary files (e.g., %temp% in Windows, /tmp in Linux).
Move rarely used data to another drive (e.g., HDD or external SSD).
Use disk cleaning tools:
- Windows: ccleaner, Windows Disk Cleanup.
- Linux: bleachbit, ncdu.
Disable hibernation in Windows (powercfg /h off) or reduce the size of the page file (pagefile.sys).

2. Forcing TRIM and garbage collection

Windows: Use the Optimize-Volume -DriveLetter X -ReTrim -Verbose command in PowerShell or run optimization via the graphical interface.
Linux: Run sudo fstrim -v / or add a task to cron (@weekly fstrim -a).
MacOS: Enable TRIM with the sudo trimforce enable command.

3. Using manufacturer software

Many companies offer dedicated tools to optimize their SSDs:

Samsung Magician: Allows for optimization, performance testing, and firmware updates.
WD Dashboard: Monitors drive health and offers optimization tools.
Crucial Storage Executive: Enables firmware updates and performance monitoring.

4. Changing system settings

Windows:
- Disable disk indexing (Drive Properties > Uncheck "Allow files on this drive to have contents indexed").
- Move the page file (pagefile.sys) to another drive.
Linux:
- Mount partitions with the noatime option (reduces metadata writes).
- Consider using the F2FS file system, optimized for SSDs.
- Move directories such as /var or /tmp to a RAM disk (tmpfs).

Are modern SSDs less susceptible to slowdowns?

Technological progress has made modern SSDs less susceptible to slowdowns at high capacity, but the problem still exists. Here are the changes introduced in recent years:

PCIe 4.0/5.0 and NVMe: These drives offer significantly higher bandwidth (e.g., 7000 MB/s in PCIe 4.0), which mitigates the perceived performance drop. A higher number of NAND channels (e.g., 16 in Phison E18 controllers) reduces bottlenecks.
SLC Cache: Modern drives use a larger pseudo-SLC buffer (e.g., 200 GB on a 1 TB drive), which allows for maintaining high speeds even at high capacity. Once the buffer is full, the speed drops, but it is less noticeable than in older models.
QLC/PLC NAND: Drives with QLC or PLC memory are cheaper and offer higher capacity, but they are slower and more prone to slowdowns at high capacity than TLC drives. For example, the Crucial P5 Plus (QLC) shows a greater performance drop at 85% capacity than TLC drives.
Enterprise drives: Models such as the Samsung PM9A3 are designed to operate at high capacity (even 90%). They use advanced GC algorithms and larger over-provisioning, which allows for maintaining performance.

Despite these improvements, the slowdown problem still occurs, especially in the case of drives with QLC memory or during intensive random writes. Modern drives are simply less susceptible to extreme performance drops.

What are the long-term consequences of using an SSD >80%?

Using an SSD at high capacity has not only short-term consequences in the form of performance drops but can also affect its lifespan and reliability:

1. Reduced lifespan (TBW)

Write Amplification Factor (WAF): At high capacity, WAF can increase from 1.1–1.3 (optimal) to 3–5. This means that every user write generates 3–5 times more internal writes, which accelerates NAND cell wear.
Example: A TLC drive with a TBW of 600 TB may wear out in 3 years at WAF=3, whereas at WAF=1.5, its lifespan would extend to 5 years.

2. Increased risk of failure

Bad blocks: High capacity increases the risk of bad blocks occurring. The controller must bypass such blocks more frequently, which lowers performance and can lead to errors.
Over-provisioning: Manufacturers reserve 7–15% of drive capacity for wear leveling and garbage collection. When filled above 80%, the user "takes away" this space, which hinders the controller's operation.

3. Market data

An analysis conducted by Backblaze (2022) on a sample of 2500 SSDs showed that drives operating at >90% capacity had a 2x higher risk of failure than those at <70%.

Summary

SSD slowdowns at high capacity are a complex problem resulting from mechanisms such as write amplification, garbage collection, and wear leveling. Although modern drives are less susceptible to extreme performance drops, the problem still exists, especially for drives with QLC memory or during intensive random writes.

To maintain optimal SSD performance, it is worth:

Leaving at least 10–20% of free space.
Regularly running TRIM (fstrim in Linux, Optimize-Volume in Windows).
Avoiding defragmentation (unnecessary for SSDs).
Monitoring drive health using manufacturer tools.
Moving rarely used data to another drive.

In the long term, high drive capacity can shorten its lifespan and increase the risk of failure. Therefore, it is worth ensuring regular optimization and monitoring of the drive's status, especially if it is used intensively.

If you are interested in Linux system optimization, check out our guide How to limit CPU and RAM usage by processes in Linux?. For Windows users, we recommend the post WSL 3 and a new era of programming, which explains how deep integration with Linux is changing Windows.