PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
An SSD’s advertised IOPS is a peak result under a defined workload, not a speed guarantee for every computer or benchmark. To find out why your result is lower, reproduce the vendor’s I/O size, read/write mix, access pattern, queueing, cache mode, test duration and connection path, then verify that the benchmark really creates that workload. A low number is meaningful only after those conditions match.
What an advertised IOPS number actually describes
IOPS means input/output operations per second. One operation can be a 4 KiB random read, a 1 MiB sequential write, or another request size and pattern. Those workloads stress different parts of an SSD and produce very different figures. A small-block random test is commonly reported in IOPS; a large-block sequential test is usually more useful when expressed as throughput in MB/s or GB/s.
IOPS is tied to the workload
When a manufacturer lists random-read IOPS, the result normally assumes a particular block size, queue depth or worker count, read/write ratio, test duration and software path. Change any of those variables and you are measuring a different workload. A 4 KiB random-read result cannot be compared directly with a 128 KiB sequential-read result, even if both are labeled “read.”
Benchmark software such as fio exposes these variables explicitly. Its bs option sets the size of each I/O unit, with separate read, write and trim values available. The same drive can show a high IOPS value with small requests and a lower operation count but much higher data throughput with large requests.
#1 Best Overall
- MEET THE NEXT GEN: Consider this a cheat code; Our Samsung 990 PRO Gen4 SSD helps you reach near max performance with lightning-fast speeds; Whether you’re a hardcore gamer or a tech guru, you’ll get power efficiency built for the final boss
- REACH THE NEXT LEVEL: Gen4 steps up with faster transfer speeds and high-performance bandwidth; With a more than 55% improvement in random performance compared to 980 PRO, it’s here for heavy computing and faster loading
- THE FASTEST SSD FROM THE WORLD'S FLASH MEMORY BRAND: The speed you need for any occasion; With read and write speeds up to 7450/6900 MB/s you’ll reach near max performance of PCIe 4.0 powering through for any use
- PLAY WITHOUT LIMITS: Give yourself some space with storage capacities from 1TB to 4TB; Sync all your saves and reign supreme in gaming, video editing, data analysis and more
- IT’S A POWER MOVE: Save the power for your performance; Get power efficiency all while experiencing up to 50% improved performance per watt over the 980 PRO; It makes every move more effective with less consumption
Read, write and mixed workloads are different
Flash translation, garbage collection and the SSD’s controller behave differently for reads and writes. A vendor’s “up to” random-read figure does not predict random-write performance. A mixed workload, such as 70% reads and 30% writes, also has its own result. Record the exact ratio before comparing two numbers.
Peak, sustained and short-term results
An advertised figure may be a peak reached after a short warm-up, while your test may run long enough to expose steady-state behavior, thermal limits or background garbage collection. Treat the duration and preconditioning state as part of the specification. Do not turn a peak figure into a sustained-performance promise.
| Condition | Why it changes the result | What to record |
|---|---|---|
| I/O size | Small requests increase operation count; large requests emphasize data rate. | Block size for reads and writes |
| Access pattern | Random access stresses addressing and parallelism; sequential access allows more efficient transfers. | Random or sequential, plus the span of the test file |
| Read/write mix | Writes can trigger flash-management work and alter read latency. | Read, write or exact mixed percentage |
| Outstanding I/O | More independent requests can keep the controller busy, but may increase latency. | Workers, queue depth and measured in-flight requests |
| Cache and path | Data may be served by system memory, a virtual-disk cache or a storage service instead of the SSD. | Direct I/O, raw device or filesystem, local or remote route |
| Duration and state | Short tests can show burst behavior; longer tests reveal steady-state limits. | Warm-up, runtime and preconditioning procedure |
1. Capture the SSD claim and your test conditions
Start by writing down the complete claim rather than just the headline number. Identify the exact model, capacity and interface, then locate the manufacturer’s table for that model and capacity. Different capacities of one product family can have different NAND configurations and ratings.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Record the published workload. Note whether the figure is read or write, random or sequential, the block size, queue depth or number of workers, and the stated test platform.
- Mark whether it is peak or sustained. “Up to” values and short benchmark runs should not be compared with a long steady-state test.
- Document your system. Record the operating system, filesystem, motherboard slot or adapter, negotiated link speed, firmware, drive temperature and whether the SSD is local, virtualized or attached through a network service.
- Save the complete benchmark output. Keep IOPS, throughput, average and percentile latency, CPU usage, queue information and the exact command or profile.
If the published conditions are missing, you cannot make a device-specific comparison yet. A generic benchmark result beside an unidentified product number is not evidence that the SSD is underperforming.
2. Recreate the workload in fio
Use a benchmark profile whose settings are visible and repeatable. The following Linux example is a template for a disposable test file; adapt the filename, block size and queue settings to the vendor’s stated conditions.
fio --name=randread-test
--filename=/path/to/testfile
--size=20G
--rw=randread
--bs=4k
--ioengine=libaio
--iodepth=32
--numjobs=1
--direct=1
--time_based
--runtime=60s
--group_reportingThis command is an example, not a universal recipe. A test file can measure the filesystem and its free-space behavior rather than the entire raw device. Never point a destructive benchmark at a device containing data unless the test procedure explicitly protects it; use a disposable file or an empty test drive.
Rank #2
- PCIe 4.0 Performance: Delivers up to 7,100 MB/s read and 6,000 MB/s write speeds for quicker game load times, bootups, and smooth multitasking
- Spacious 1TB SSD: Provides space for AAA games, apps, and media with standard Gen4 NVMe performance for casual gamers and home users
- Broad Compatibility: Works seamlessly with laptops, desktops, and select gaming consoles including ROG Ally X, Lenovo Legion Go, and AYANEO Kun. Also backward compatible with PCIe Gen3 systems for flexible upgrades
- Better Productivity: Up to 2x faster than previous Gen3 generation. Improve performance for real world tasks like booting Windows, starting applications like Adobe Photoshop and Illustrator, and working in applications like Microsoft Excel and PowerPoint
- Trusted Micron Quality: Built with advanced G8 NAND and thermal control for reliable Gen4 performance trusted by gamers and home users
Set every material parameter
--bs: set the vendor’s request size, such as 4 KiB. If reads and writes use different sizes, specify them separately.--rw: chooserandread,randwrite,read,writeor the documented mixed ratio.--iodepthand--numjobs: reproduce the advertised outstanding-work level, but do not assume the requested depth was achieved.--ioengine: use an asynchronous engine appropriate to the operating system and device. Engine behavior determines whether multiple requests can actually be in flight.--direct: use direct I/O when the goal is to test the storage device rather than the operating system page cache, if the platform and filesystem support it.- Runtime and warm-up: match the published duration and discard or record warm-up results separately.
For a mixed workload, specify the exact read percentage instead of relying on a default. For example, a profile might use --rw=randrw --rwmixread=70; the percentage is meaningful only when it matches the comparison result.
3. Confirm that queue depth creates real concurrency
A queue-depth setting in a benchmark file is a request to the benchmark, not proof that the SSD received that many simultaneous operations. The I/O engine may issue requests synchronously, the operating system may throttle them, or the device may have reached another limit.
The synchronous-engine trap
NVIDIA’s storage-performance guidance demonstrates the problem: increasing iodepth does not make a synchronous fio engine asynchronous. With a synchronous engine, the next request waits for the previous one, so a nominal depth of 32 can still produce one outstanding operation. Use an engine that supports asynchronous submission and verify its behavior on your operating system.
Observe the device while the test runs
Check the device’s queue activity, not just the command line. Operating-system and vendor tools can show utilization, request size, queue occupancy and latency. If utilization is low and only one request is outstanding, raising the configured depth will not produce the advertised parallel workload. If the queue is full but latency rises sharply, the additional concurrency may be hurting useful performance.
NVIDIA’s 2018 worked example illustrates why this matters. An 850 EVO-era SATA SSD produced about 25,000 random-read IOPS at 4 KiB in a particular Linux and fio setup, while the article cited roughly 92,000 random 4 KiB read IOPS for that drive under its stated conditions. The figures are historical examples, not current or universal SSD results; the point was to investigate whether the workload supplied enough parallel I/O before blaming the drive.
4. Remove cache and path confusion
Your benchmark may measure more than the NAND and controller. Identify every layer between the application and the flash.
Rank #3
- SPEED UP PROJECTS. Launch creator applications fast with uncompromising PCIe 4.0 read speeds up to 7,100MB/s,[2] (1TB and 2TB[1] models) and write speeds up to 6,700MB/s[2] (1TB[1]-4TB[1] models).
- CREATE AND STORE MORE. Make more room for your 4K videos and high-resolution images with capacities from 500GB[1] up to 4TB[1] on M.2 2280 built with our trusted 8th generation SANDISK BiCS QLC 3D CBA NAND.
- IT GOES WHERE YOU GO. With an all-new power efficient design, your drive delivers high performance with low power, giving you more time to be productive while on the go.
- UNCOMPROMISED RELIABILITY. With up to 1,200 TBW[3] (4TB[1] model) endurance rating, your drive is designed for creators.
- KEEP YOUR DRIVE UPDATED. Monitor your SSD’s performance and check for updates with the downloadable SANDISK Dashboard application.[5]
System page cache
Repeated reads can be served from system memory. The first pass may access the SSD, while later passes appear much faster because the data is cached. NVIDIA’s guidance calls out page-cache effects and demonstrates direct-I/O configuration. Use direct I/O when your objective is device performance, or deliberately test cached behavior when that is the real application workload. State which one you measured.
Filesystem versus raw block device
A filesystem adds allocation, metadata and alignment behavior. A raw-device test removes those variables but can destroy data. A file-based test is safer and more representative of applications, yet it may not match a vendor’s raw-device result. Keep the path identical when comparing two runs.
Virtual disks, cloud volumes and network routes
With a virtual machine or cloud volume, the request travels through a guest driver, hypervisor, virtual controller and storage service. The volume may have its own IOPS cap, while the VM or instance imposes another. Microsoft Azure notes that disk-cache settings affect whether reads reach the underlying disk. AWS likewise documents that queue length should be selected for the workload and that platform limits can constrain observed performance. A benchmark of the whole path is not automatically a benchmark of the physical SSD.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Link and adapter limits
Check the negotiated interface and adapter. A drive connected through a slower link, a shared controller or an enclosure can be limited before its internal controller is busy. Record the actual link speed reported by the operating system rather than the slot’s theoretical maximum.
5. Increase concurrency while watching latency
Do not jump directly to a large queue. Run the same profile at progressively higher concurrency, for example 1, 2, 4, 8, 16 and 32 outstanding operations, while keeping all other variables fixed.
| Metric | What an improvement means | Warning sign |
|---|---|---|
| IOPS | More completed operations per second for the selected workload. | IOPS stops increasing as queue depth rises. |
| Throughput | More data transferred per second; useful for large requests. | Throughput is capped by the interface or host. |
| Average and percentile latency | Response time remains acceptable for the application. | Latency climbs while IOPS gains become small. |
| Device utilization | The SSD is receiving enough work to expose its capability. | Low utilization indicates an upstream bottleneck or insufficient parallelism. |
There is no universal queue depth that maximizes useful performance. AWS emphasizes that the optimal queue length depends on the workload and whether it is more sensitive to IOPS or latency. A database may prefer lower latency at a moderate queue, while a batch workload may tolerate deeper queues for more throughput.
Rank #4
- This product has been replaced by our latest generation. Please search for the SANDISK Optimus GX 7100 NVMe SSD
- HIGH-OCTANE GAMING. Experience speeds up to 7,250MB/s read and 6,900MB/s write (1-2TB models), with up to 35% faster performance than previous generation.
- PURPOSE-BUILT. Designed for serious on-the-go gamers, with a PCIe Gen4 interface and SANDISK’s next generation TLC 3D NAND.
- MORE TIME TO CLEAR THAT CHECKPOINT. Built with laptops and handheld gaming devices in mind, with up to 100% more power efficiency over the previous generation.
- DO MORE WITH DASHBOARD. Ensure your drive is optimized for prime performance with the downloadable WD_BLACK Dashboard (Windows only).
6. Control duration, temperature and preconditioning
Run repeatable tests under controlled conditions. NVIDIA recommends preconditioning a new SSD for stable measurements, including sequential and random write workloads, and warns that cache state matters for reads.
Recommended Free Tools
- Prepare the drive consistently. Use the same preconditioning sequence for every comparison, or label a result as an out-of-box burst measurement.
- Allow a defined warm-up. Exclude startup behavior only if the published method does; otherwise report both warm-up and steady-state portions.
- Keep the test long enough. A brief run may measure a burst cache. A longer run can reveal sustained limits and garbage collection.
- Monitor temperature. Record temperature during the run and note any thermal throttling. Do not compare a cool, idle drive with one already hot from unrelated work.
- Stop competing activity. Disable or pause transfers, indexing, virtual machines and background jobs that could consume the same storage path.
- Repeat identical runs. Use the same command, file location, queue settings and timing, then compare the spread rather than one unusually low sample.
7. Use a structured decision process for a low result
After the controlled runs, classify the bottleneck before considering a replacement.
Case A: The workloads do not match
If your block size, read/write mix, pattern, queueing or duration differs from the published test, change the profile first. The result answers a different question and cannot establish a fault.
Case B: The benchmark is not supplying enough work
If the engine is synchronous, the worker count is too low or observed queue activity remains shallow, fix concurrency and repeat. The historical NVIDIA example shows how this alone can explain a large gap between a local result and a headline specification.
Case C: An upstream layer is limiting the path
If direct-I/O testing of a local drive improves the result, page cache or filesystem behavior was involved. If a raw-device or local test is fast but a virtual or cloud test is slow, inspect volume limits, VM limits, cache mode and the virtual controller.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Case D: Performance changes during the run
A falling result accompanied by rising temperature suggests thermal control; a drop after a burst suggests cache exhaustion or steady-state flash management. Record the time series instead of reporting only the first number.
Best Value
- Ideal for high speed, low power storage
- Gen 4x4 NVMe PCle performance
- Up to 6,000MB/s read, 4,000MB/s write
- Includes Acronis cloning software
- 5-year limited warranty
Case E: A matched, repeatable test remains low
Collect the exact model and capacity, firmware, operating system, interface and negotiated link, temperature, health data and complete benchmark output. Check the manufacturer’s model-specific guidance and warranty process. Only after a controlled comparison reproduces the shortfall under matching conditions is a device fault a reasonable conclusion.
How to compare two SSD benchmark results correctly
Place the results side by side and fill every field. If a value is not stated, mark it as “not stated” rather than assuming it.
| Comparison axis | Questions to answer |
|---|---|
| I/O size | Are both tests using the same read and write block sizes? |
| Pattern | Are both random, both sequential, or otherwise equivalent? |
| Read/write ratio | Are both pure reads, pure writes or the same mixed percentage? |
| Outstanding work | How many workers and requests were actually in flight? |
| Benchmark engine | Can both engines issue asynchronous operations in the same way? |
| Cache mode | Were reads cached, direct, or unspecified? |
| Storage path | Was each test on a raw device, filesystem, virtual disk or application path? |
| Duration and state | Were warm-up, preconditioning and steady-state periods equivalent? |
| Latency | Did higher IOPS come with an unacceptable latency increase? |
| Platform limits | Were host, instance, controller or volume caps present? |
The advertised result becomes a useful target only when these axes align. Otherwise, the numbers describe different experiments.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteCommon mistakes and their fixes
- Comparing 4 KiB random IOPS with sequential MB/s: run the same block size and access pattern.
- Raising
iodepthon a synchronous engine: select a suitable asynchronous engine and verify actual queue activity. - Running repeated reads without controlling cache: use direct I/O for a device-focused test or document that the test measures cached reads.
- Testing a cloud volume without checking the VM: check both volume and instance limits using the platform’s benchmark guidance.
- Reporting only IOPS: include throughput, average latency and relevant percentile latency.
- Testing a new drive once: define preconditioning, warm-up, runtime and repeat count.
- Assuming a low score proves failure: first match the vendor’s workload and isolate the storage path.
What to collect before requesting support
A support engineer can act on a reproducible record. Include the exact model and capacity, firmware version, operating system and kernel or build, motherboard slot or adapter, negotiated interface speed, filesystem or raw-device status, benchmark command, fio version, block size, read/write pattern, worker and queue settings, direct-I/O choice, runtime, preconditioning, temperature, health statistics and complete output. State whether the test was local, virtualized or cloud-hosted and identify any cache settings.
This record separates a workload mismatch from a host limitation and gives the manufacturer enough context to assess a possible drive fault. Without it, a single low IOPS value cannot distinguish the SSD from the software and hardware around it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
