In Part 1, we explored what SSD Write Amplification means, why NAND Flash may receive more writes than the host system requests, and how factors such as Garbage Collection, TRIM, SLC Cache, and workload can influence internal write activity.
Write Amplification is also closely connected to NAND endurance.
Every time data is programmed into NAND Flash, the memory cells experience a certain amount of wear. Additional internal writes can make the total NAND workload significantly higher than the amount of data written by the host system.
For SSD manufacturers and OEM designers, understanding this relationship is important when developing storage solutions that balance performance, capacity, reliability, and long-term operating stability.
This article explores how Write Amplification affects NAND endurance and what OEMs should consider during SSD design, testing, and validation.
1. The Relationship Between Write Amplification and NAND Endurance
NAND Flash has a limited number of Program/Erase (P/E) cycles.
A P/E cycle generally refers to programming data into a NAND block and subsequently erasing that block. Actual endurance characteristics depend on the NAND type, flash architecture, operating conditions, controller management, and other design factors.
Whenever an SSD writes data, the NAND cells experience wear over time. Additional internal writes caused by data movement, Garbage Collection, or other management operations can increase the total amount of NAND activity.
예를 들어
- Host writes: 1 TB
- NAND writes: 1.5 TB
- Write Amplification Factor: 1.5
In this simplified example, the NAND receives 50% more written data than the host sends to the SSD.
This does not mean that NAND endurance decreases by exactly 50% in every situation. Wear depends on additional factors, including the distribution of writes, block management, NAND architecture, and controller behavior.
The example nevertheless demonstrates why host-written data and NAND-written data are not always the same.
2. Why Additional NAND Writes Increase Wear
NAND Flash cells can withstand only a finite amount of program and erase activity.
As data is repeatedly written, erased, and managed, the electrical characteristics of the cells can gradually change. Over time, this may affect data retention, error rates, and the ability of NAND to store data reliably.
Higher Write Amplification means that the NAND may experience more internal write activity for the same amount of host data.
Potential effects include:
- Higher total NAND write volume
- Increased consumption of available P/E cycles
- Greater internal data movement
- Additional workload for the controller
- Potentially higher long-term wear
The impact depends on the NAND technology and how the SSD manages data internally.
For this reason, reducing unnecessary internal writes is an important consideration when designing SSDs for long-term use.
3. NAND Type and Endurance Characteristics
Not all NAND Flash has the same endurance characteristics.
Different NAND types store different numbers of bits per cell, which can affect storage density, cost, performance behavior, and endurance.
Common NAND categories include:
- SLC(단일 레벨 셀) — Stores one bit per cell.
- MLC(멀티 레벨 셀) — Stores two bits per cell.
- TLC(트리플 레벨 셀) — Stores three bits per cell.
- QLC(쿼드 레벨 셀) — Stores four bits per cell.
In general, storing more bits in each cell requires more precise control over the cell’s voltage states. This can create different endurance and performance characteristics compared with NAND technologies that store fewer bits per cell.
However, NAND type alone does not determine the final endurance of an SSD.
The controller, firmware, error correction, wear leveling, over-provisioning, workload, and operating environment also influence the overall storage solution.
As a result, OEM designers should evaluate NAND as part of the complete SSD architecture rather than considering it as an isolated component.
4. NAND Endurance and P/E Cycles
P/E cycles are commonly used to describe NAND endurance.
A NAND block can undergo only a certain number of program and erase operations before its characteristics may become less reliable.
The expected number of P/E cycles varies according to the NAND type, manufacturing process, operating conditions, and product specifications.
However, P/E cycle ratings should not be interpreted as a direct prediction of the total usable lifetime of an SSD.
An SSD contains many NAND blocks, and the controller manages how data is distributed across them. Wear leveling attempts to distribute write and erase activity so that some blocks do not wear out much earlier than others.
The actual endurance of an SSD therefore depends on several interacting factors:
- NAND endurance characteristics
- 쓰기 증폭
- Wear-leveling efficiency
- Available spare area
- Workload pattern
- Temperature and operating conditions
- Error correction and data management
A suitable endurance evaluation needs to consider these factors together.
5. How Workload Patterns Affect Write Amplification
The same SSD can experience different levels of Write Amplification depending on the workload.
Large sequential writes, small random writes, mixed workloads, and repeated data updates place different demands on the controller and NAND Flash.
Sequential Workloads
Large sequential writes can often be managed in an organized way. Depending on the controller and firmware, this may reduce certain types of internal data movement.
Even so, sequential writes do not automatically guarantee low Write Amplification. Cache management, Garbage Collection, drive utilization, and other internal operations can still influence the result.
Random Workloads
Random writes can update data across different physical locations.
This pattern may require the controller to move valid data more frequently, especially when small updates are distributed across multiple blocks.
Consequently, random write workloads can create more internal write activity in certain SSD configurations.
Mixed Read/Write Workloads
Many real-world applications perform reading and writing at the same time.
During these operations, the controller must manage incoming writes, read requests, background processes, and data organization concurrently. This can affect latency, performance consistency, and internal write activity.
For OEM validation, using multiple workload types is more informative than evaluating endurance through a single sequential write test.
6. Small Writes and Read-Modify-Write Operations
Small write operations can be particularly demanding for NAND management.
NAND Flash is organized into pages and blocks, and data cannot always be modified in place in the same way as conventional memory.
When a small portion of data changes, the controller may need to preserve existing valid data while writing the updated information elsewhere.
Depending on the implementation, this process can involve additional data movement or Read-Modify-Write operations.
For example, a host may update only a small portion of a larger data structure. The SSD controller may need to manage the updated data together with other valid data in the relevant NAND region.
The resulting internal write volume may exceed the original host write request.
This is why workload granularity matters when evaluating Write Amplification and endurance.
OEM designers should consider not only how much data an application writes, but also how the application writes that data.
7. The Role of Over-Provisioning
Over-Provisioning (OP) refers to NAND capacity reserved for internal SSD management rather than being fully exposed as user-accessible capacity.
This additional space can give the controller more flexibility when managing:
- Garbage Collection
- Wear Leveling
- Bad Block Management
- Data Movement
- Background Operations
- Free Block Preparation
When sufficient spare area is available, the controller may have more options for organizing data and preparing free blocks.
Under certain workloads, this can help reduce internal write pressure and improve performance consistency.
However, the relationship between Over-Provisioning and Write Amplification is not always linear.
Increasing spare area does not automatically guarantee a specific Write Amplification value or a fixed endurance improvement. The result depends on the controller architecture, firmware, NAND configuration, workload, and drive utilization.
For OEM projects, the appropriate balance between usable capacity and reserved capacity should be determined according to the target application.
8. TRIM, Garbage Collection, and NAND Endurance
TRIM and Garbage Collection can influence NAND endurance because they affect how the SSD manages invalid data and reclaims storage blocks.
TRIM provides information about data that the host system no longer needs. This allows the controller to identify invalid logical blocks and manage NAND space more efficiently.
Garbage Collection reorganizes valid data and reclaims blocks that contain invalid data.
When these processes are managed efficiently, the controller may be able to reduce unnecessary data movement in certain situations.
However, Garbage Collection itself can also generate additional internal writes because valid data may need to be moved before a block can be erased and reused.
The overall effect depends on how the firmware manages the balance between:
- Available free blocks
- Valid and invalid data
- Host write requests
- Background operations
- 웨어 레벨링
- NAND block utilization
Therefore, TRIM and Garbage Collection should be evaluated together with Write Amplification rather than considered independently.
9. SLC Cache and Endurance Considerations
SLC Cache is commonly used to improve short-term write performance.
Incoming data may initially be written into an SLC-like area and later reorganized into the final storage format. This can help the SSD handle burst writes more efficiently.
However, additional internal NAND activity may occur when data is moved or reorganized.
The impact on endurance depends on the cache implementation, NAND architecture, controller, firmware, workload, and cache management strategy.
For example, a workload that frequently fills and refreshes the SLC cache may create different internal write behavior from a workload involving occasional short bursts.
This is why SSD testing should not focus only on write speed while the cache is available.
It is also important to evaluate:
- Performance after the SLC cache is exhausted
- Sustained write behavior
- Internal data movement
- Temperature during extended writes
- Performance consistency over time
- NAND write activity where measurable
A well-designed SLC cache should be evaluated as part of the complete SSD architecture.
10. Controller and Firmware Optimization
The SSD controller and firmware play an important role in managing Write Amplification and NAND endurance.
The controller handles data mapping, NAND communication, error correction, wear leveling, Garbage Collection, and other internal operations.
Firmware determines how these processes are coordinated and how the SSD responds to different workload conditions.
Important areas include:
Flash Translation Layer
The Flash Translation Layer (FTL) maps logical addresses from the host system to physical NAND locations.
An effective FTL design can influence how data is organized, updated, and moved internally.
Wear Leveling
Wear leveling distributes write and erase activity across NAND blocks to avoid excessive wear on a limited number of blocks.
Garbage Collection
Garbage Collection reclaims NAND blocks by moving valid data and erasing blocks containing invalid data.
Bad Block Management
The SSD must identify and manage blocks that are no longer suitable for reliable operation.
Error Correction
Error correction helps the controller identify and correct certain types of NAND data errors, supporting data reliability as NAND cells wear.
These functions work together. Optimizing one area in isolation may not produce the best overall result.
11. Error Correction and NAND Wear
As NAND Flash experiences wear, the likelihood of data errors may increase.
Error correction mechanisms such as ECC and LDPC help the controller detect and correct certain errors during data reading.
However, error correction does not eliminate NAND wear or make NAND endurance unlimited.
The controller and firmware need to manage the relationship between:
- NAND wear
- Error correction capability
- Read and write behavior
- Data retention
- Bad block management
- 장기적인 신뢰성
When evaluating an SSD, OEM designers should consider the error correction architecture alongside the NAND type and endurance requirements.
A reliable SSD solution requires more than selecting NAND with a particular endurance rating. The controller, firmware, and error correction system must also be suitable for the target application.
12. Workload Endurance vs. Advertised Endurance
SSD endurance is often represented through metrics such as TBW (Terabytes Written) or DWPD (Drive Writes Per Day).
These metrics provide useful reference points, but they should be interpreted in the context of the test conditions and product requirements.
TBW describes a specified amount of total data written over the expected product life under defined conditions.
DWPD describes how many times the drive’s usable capacity can be written per day over a specified period.
Actual endurance behavior depends on factors such as:
- 쓰기 증폭
- NAND type
- 워크로드 특성
- Temperature
- Drive capacity
- Firmware behavior
- Over-Provisioning
- Data retention requirements
A product’s advertised endurance rating should not be treated as a guarantee that every workload will produce the same result.
For OEM applications, the expected workload and operating environment should be clearly defined before selecting an endurance target.
13. Why SSD Capacity Can Affect Endurance
Different SSD capacity versions may use different NAND configurations.
A higher-capacity SSD may contain more NAND resources, which can provide additional flexibility for data distribution and internal management. However, this does not mean that every higher-capacity version will automatically deliver better endurance or lower Write Amplification.
The actual result depends on the design of the product.
Capacity can influence:
- Number of NAND packages and dies
- Available spare area
- Internal parallelism
- Data distribution
- Controller workload
- Performance consistency
- Endurance behavior
For this reason, SSD manufacturers should evaluate each important capacity version rather than assuming that test results from one model can be applied to all capacities.
Capacity-specific validation is particularly important for OEM projects with different product configurations.
14. Temperature and Long-Term NAND Reliability
Temperature is another factor that needs to be considered during endurance evaluation.
Intensive write workloads can increase controller and NAND activity, which may raise power consumption and operating temperature.
Temperature can influence SSD performance, thermal throttling, and long-term component behavior.
However, temperature and endurance are not independent variables.
An SSD’s thermal behavior depends on:
- Controller power consumption
- NAND characteristics
- PCB layout
- Heatsink and thermal interface
- System airflow
- Ambient temperature
- Workload intensity
- Duration of operation
For example, an SSD may perform well during a short write benchmark but behave differently during a long-duration workload when internal operations and heat generation increase.
OEM validation should therefore include realistic thermal conditions rather than relying only on room-temperature, short-duration testing.
15. How OEMs Should Evaluate Write Amplification
For OEM SSD projects, Write Amplification should be measured under workloads that represent the expected application.
A suitable evaluation process may include the following steps.
Step 1: Define the Target Workload
Identify how the SSD will be used.
Relevant questions include:
- Will the workload be mostly sequential or random?
- How much data will be written each day?
- Will the workload include frequent small updates?
- Will the SSD operate continuously?
- How much free space will normally be available?
- Will the product be used for consumer, industrial, surveillance, or embedded applications?
The workload definition provides a foundation for meaningful testing.
Step 2: Select the SSD Configuration
Define the NAND, controller, firmware, capacity, and PCB configuration.
Different configurations may produce different performance and endurance results.
Step 3: Measure Host Writes
Record the total amount of data written by the host system during the test.
Step 4: Monitor NAND Write Activity
Where appropriate diagnostic information is available, monitor internal NAND writes or other indicators that can help estimate Write Amplification.
The measurement method should be consistent and clearly documented.
Step 5: Test Different Drive Utilization Levels
Evaluate the SSD at different levels of used capacity, because internal data management may change as available space decreases.
Step 6: Run Long-Duration Workloads
Short benchmarks may not fully reveal internal write behavior. Extended workloads can help expose sustained performance changes, background operations, and thermal effects.
Step 7: Evaluate Endurance and Stability
Review the impact of repeated workloads on performance, error behavior, temperature, and long-term stability.
The purpose is to understand the behavior of the complete SSD solution rather than focus on a single measurement.
16. Important Test Scenarios for OEM Validation
A comprehensive SSD validation plan can include several workload scenarios.
Sequential Write Testing
This test is useful for evaluating large file transfers and sustained sequential data processing.
Random Write Testing
Random write testing helps examine SSD behavior under small and distributed data updates.
Mixed Read/Write Testing
This scenario represents workloads in which reading and writing occur simultaneously.
Sustained Write Testing
Continuous writing helps evaluate performance after the SSD has been operating under extended write activity.
High Drive Utilization Testing
Testing at high capacity usage examines how the SSD behaves when much of its storage space is already occupied.
Thermal Testing
Thermal evaluation measures performance and stability under defined temperature and workload conditions.
Long-Duration Testing
Extended operation can help identify changes in performance, error behavior, and internal management over time.
Different test scenarios reveal different aspects of Write Amplification and NAND endurance.
17. Write Amplification Is Not the Only Endurance Factor
Although Write Amplification is important, it should not be treated as the only factor that determines SSD endurance.
Other aspects of SSD design can also affect long-term reliability.
These include:
- NAND quality and consistency
- Controller architecture
- 펌웨어 최적화
- Error correction capability
- Wear-leveling strategy
- Bad block management
- Over-Provisioning
- 열 설계
- Power stability
- Data retention requirements
- Target workload
For example, two SSDs with similar Write Amplification values may still show different long-term results if their NAND characteristics, thermal environments, or firmware implementations differ.
This is why endurance evaluation needs to consider the complete storage system.
18. Balancing Performance, Capacity, and Endurance
SSD design often requires balancing several competing requirements.
Customers may expect:
- High sequential read and write speeds
- Consistent random performance
- Large usable capacity
- Low power consumption
- Long service life
- Stable performance
- Competitive product cost
These requirements cannot always be maximized independently.
For example, increasing usable capacity may reduce the amount of space available for internal management. A more aggressive cache strategy may improve burst performance but introduce additional internal data movement. A higher-performance controller may also create greater thermal or power requirements.
As a result, SSD design should be based on the target application rather than a single performance metric.
The most suitable configuration is the one that provides an appropriate balance of performance, endurance, compatibility, cost, and reliability.
19. What OEM Designers Should Consider
When selecting or developing an SSD solution, OEM designers should evaluate several factors together rather than focusing on a single specification.
Is the NAND Suitable for the Workload?
The selected NAND type and its endurance characteristics should match the expected write volume and operating conditions.
Is the Controller Appropriate?
A suitable controller needs to support the selected NAND configuration while meeting the intended performance and reliability requirements.
Is the Firmware Properly Optimized?
Effective firmware management is essential for data organization, error correction, wear leveling, and Garbage Collection.
Is the Thermal Design Adequate?
Thermal performance should be evaluated under realistic workloads and system operating conditions.
Has Each Capacity Been Validated?
Different capacity versions may use different internal configurations, so individual validation may be necessary.
Has Long-Term Performance Been Evaluated?
Short benchmark results should be complemented by sustained and long-duration testing to understand performance consistency.
Is the Product Suitable for the Target Platform?
Compatibility with the host system, operating environment, and application requirements should be confirmed before mass production.
These considerations help OEM designers move beyond basic specifications and conduct a more complete evaluation of SSD performance, endurance, and reliability.
20. Oreton’s Approach to SSD Design and Validation
에서 Oreton, we focus on the complete SSD architecture when developing storage solutions for different applications.
NAND Flash, controller selection, firmware behavior, capacity, thermal design, and target workload all need to be considered together.
Write Amplification and NAND endurance are important parts of this evaluation process. Rather than focusing only on peak read and write speeds, SSD validation should also consider sustained performance, internal data management, temperature, compatibility, and long-term stability.
For OEM projects, product requirements can vary depending on the target system and application. Different configurations may therefore require different approaches to NAND selection, controller matching, and validation.
Our goal is to provide SSD solutions that balance performance, stability, compatibility, and practical application requirements.
결론
Write Amplification and NAND Endurance are closely connected.
When the NAND writes more data internally than the host system requests, the total NAND workload may increase. Over time, this can influence NAND wear and the long-term endurance behavior of the SSD.
However, Write Amplification is only one part of the overall endurance equation. NAND type, controller, firmware, error correction, wear leveling, Over-Provisioning, temperature, and workload all contribute to the final result.
For OEM designers, SSD endurance should therefore be evaluated through the complete storage architecture and realistic operating conditions.
A reliable SSD is not defined by NAND endurance specifications alone. It also depends on how effectively the controller, firmware, NAND, thermal design, and target platform work together.
By evaluating Write Amplification, sustained performance, workload behavior, and long-term stability, OEMs can make more informed decisions when developing SSD solutions for different applications.