What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
iTechGuides 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
Benchmark an SSD as a measurement of a specific workload, not as a single score for the drive. Hold the host system, connection, software version, test data, cache policy, block size, queue depth, and run duration constant; then report bandwidth, IOPS, and latency with those conditions. Add a controlled test of the activity you actually care about—such as a large-file copy, a build, project loading, or game startup—because the operating system, file system, cache, and applications can transform I/O before it reaches the SSD.
This approach lets synthetic tests answer the question they are good at—repeatable comparison—without presenting a sequential MB/s number as a promise about every application.
What an SSD benchmark actually measures
An SSD score belongs to a workload definition. The same drive can produce very different results when you change access pattern, block size, read/write mix, outstanding I/O, data pattern, cache handling, or test duration. The SNIA Solid State Storage Initiative primer describes workloads through their access patterns and demand intensity; SSD performance responds to that I/O stream rather than to a device label in isolation.
A synthetic benchmark generates a controlled stream of reads and writes. That control is valuable when comparing two drives, because identical platforms and settings make differences easier to attribute. It is not a literal replay of every application. SNIA explains that synthetic models simplify real workloads: they can provide a useful comparative base without exactly representing any particular real-world workload.
#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
Application I/O also changes over time. A program may issue small random reads, wait, write metadata, flush data, and then request a large sequential transfer. The file system, operating-system cache, storage driver, controller, and enclosure can alter those requests before the NAND media handles them. Therefore, a benchmark result is best read as “performance under this specified workload and setup.”
Peak, burst, and sustained performance are separate questions
A short run can capture fresh-out-of-box or cache-assisted peak behavior. It cannot establish that the same rate will continue through an indefinite write. The SNIA performance-test specification paper (October 2020) distinguishes initial peak behavior from steady-state measurement and discusses preconditioning before a sustained test.
When writing results, use precise labels such as “short-run sequential write,” “burst result,” or “steady-state result after the stated preconditioning.” Do not call a brief initial burst a sustained speed.
Choose the right metric for the question
| Metric or test | What it describes | What it does not prove |
|---|---|---|
| Sequential bandwidth (MB/s or GB/s) | Large, contiguous transfers under the selected block size, queue depth, data pattern, and cache policy. | Boot time, application responsiveness, game loading, or small-file performance. |
| Random IOPS | How many selected-size random operations complete per second at the stated read/write mix and queue depth. | Performance at a different block size, queue depth, or lightly loaded desktop workload. |
| Latency | Time for individual operations, preferably shown as a distribution and percentiles. | A reliable tail estimate when too few operations were collected. |
| Application-level timing | Elapsed time for a defined task, such as copying a specified folder or completing a controlled build. | A universal rating for other applications or systems. |
Report the metric that matches the reader’s decision. A high-queue-depth random result may demonstrate that a device can be stressed or saturated, while a lightly loaded desktop usually generates a different request stream. A sequential result is useful for large transfers, but it should not be used as a general-purpose responsiveness rating.
Prepare a reproducible test environment
Before launching a benchmark, record the conditions that could change the result. These fields are practical reporting recommendations for consumer testing; the cited standards emphasize controlling and documenting the platform rather than prescribing one universal desktop checklist.
- SSD make, model, nominal capacity, and whether it is SATA or NVMe.
- Connection path: motherboard slot or enclosure, negotiated link/interface, and any adapter involved.
- Operating-system version and relevant storage-driver or firmware information.
- Drive fill state, since free space and prior writes can affect behavior.
- Thermal conditions, including whether the drive is actively cooled and whether throttling occurred.
- Benchmark name and exact version.
- Test target, test-file size, test-data mode, block size, access pattern, read/write mix, queue depth, and worker count.
- Cache policy, warm-up interval, measurement interval, and number of repeated runs.
Use the same host and connection when comparing drives. A different PCIe slot, enclosure, operating-system cache state, or thermal condition can become the variable instead of the SSD.
Select a tool and capture its version
Windows options
CrystalDiskMark provides a quick graphical baseline. Its default profile is convenient, but it is not a universal model of every user’s workload. The documentation also notes that some SSDs respond differently to random data and zero-fill test data, so record the selected data mode.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchRank #2
- 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
Microsoft DiskSpd is a configurable command-line tool for synthetic storage workloads. It lets you specify block size, sequential or random access, read/write mix, queue behavior, warm-up, duration, and other parameters. Its repository currently lists DiskSpd 2.3 — Microsoft, 2026, with a release entry dated September 3, 2026. That entry warns of two default changes that may require rebaselining. Results made under changed defaults are not automatically comparable, so include the version and defaults with every published result.
Linux options
fio offers configurable Linux workloads, including I/O size and sequential or random reads and writes. Microsoft’s benchmark guide describes DiskSpd as synthetic and fio as configurable; its examples are designed for particular Azure virtual-machine configurations, not as mandatory consumer presets.
A step-by-step consumer benchmarking workflow
1. Define the decision you are trying to make
Write down the workload before choosing a preset. “Which drive is faster for large video exports?” calls for sustained large writes and an application-level export timing. “Which drive feels quicker for a desktop?” may require small random tests plus controlled application launches. “Which drive handles a backup?” calls for the backup tool or a representative file set, not only a sequential read chart.
2. Record the platform and drive state
Capture the SSD identity, capacity, interface, host connection, operating system, firmware or driver details that are available, fill state, and thermal setup. Note whether the drive is freshly installed or has already received substantial writes. Keep these conditions unchanged for every drive in the comparison.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors3. Establish a small, familiar baseline
Run the benchmark’s default profile when you need a quick consumer comparison, but publish the test count, test size, profile name, and data mode alongside the score. Treat it as a baseline under that profile, not as a universal real-world workload. A default that uses zero-fill data may produce a different result from one using random data.
4. Add tests that match the workload
Use large-block sequential reads and writes to examine large-transfer behavior. Add small-block random tests when the question concerns metadata-heavy or multitasking activity. Choose a read/write mix, block size, and queue depth that model the intended task, and state that the test is a model.
DiskSpd’s customization guide documents these controls. Microsoft’s Azure examples use 4 KiB random I/O for IOPS-oriented scenarios and 64 KiB blocks for throughput-oriented scenarios. Those values are configured for a particular Azure VM and should not be copied as universal desktop presets.
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]
5. Make the test large enough to avoid a cache-only result
A test file that fits comfortably in system memory may measure cache behavior rather than sustained device behavior. Increase the test size when the goal is device-level behavior, and document the choice. Microsoft’s benchmark example uses a 200 GiB test file to minimize memory caching and explicitly configures cache behavior. That is an environment-specific example, not a required value for every desktop.
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 →6. Warm up, repeat, and keep the spread
Allow the system to settle before collecting the measurement interval. Microsoft’s examples include a warm-up period. For a consumer comparison, run multiple comparable iterations after warm-up, report the spread or representative statistic you selected, and investigate large variation rather than publishing only the best run. Keep background updates, indexing, antivirus scans, and other obvious activity from changing between runs where practical.
7. Separate burst from long-run writes
For a short write, report the duration and call it a short-run or burst measurement. For sustained behavior, use a longer procedure with documented preconditioning and a verified steady state. The SNIA paper explains why fresh-out-of-box behavior and steady-state behavior must be treated separately. If your procedure does not reach and verify steady state, do not claim that it does.
8. Save the configuration and results
DiskSpd can export configuration and system information in XML. Preserve that output with the result so another reader can reconstruct the workload. At minimum, publish the software and version, platform, target, test-data mode, test-file size, block size, random or sequential pattern, read/write mix, queue depth and workers, warm-up and measurement durations, cache settings, and measured bandwidth, IOPS, and latency.
How to interpret synthetic results
Sequential MB/s
Sequential bandwidth answers how quickly the drive handled large contiguous transfers under the selected setup. It is useful for a large-file copy or media workflow when the copy itself is controlled, but it is not a general rating for booting, launching games, or opening applications. A system can show a high sequential number while a small-file workload remains limited by latency, queueing, or other layers.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Random IOPS
Random IOPS must always be read with block size, queue depth, read/write mix, and data pattern. Raising queue depth can expose a device’s ability to process many outstanding requests, but a lightly loaded desktop may issue only a few requests at once. Comparing an 8-queue test with a 1-queue test as if they were the same workload is invalid.
Latency distributions and percentiles
Publish a distribution or meaningful percentiles instead of only a minimum or average. Tail latency can reveal occasional stalls that an average hides. The DiskSpd result-analysis guidance warns that high-percentile estimates become unreliable when too few operations are collected; in its example, tail percentiles collapse to the maximum because the sample is too small. Ensure the run performs enough operations and lasts long enough before drawing conclusions about the tail.
Rank #4
- HUGE SPEED BOOST: Get random read/write speeds that are 40%/55% faster than 980 PRO; Experience up to 1400K/1550K IOPS, while sequential read/write speeds up to 7,450/6,900 MB/s reach near the max performance of PCIe 4.0*
- BREAKTHROUGH POWER EFFICIENCY: Use less power and get more performance; Enjoy up to 50% improved performance per watt over 980 PRO, plus optimal power efficiency with max PCIe 4.0 performance**
- SMART THERMAL CONTROL: Samsung's own nickel-coated controller delivers effective thermal control; With its slim size, 990 PRO is a perfect fit for desktops and laptops that meet the PCI-SIG D8 standard***
- THE CHAMPION MAKER: Up to 65% improvement in random performance enables faster loads for an ultimate gaming experience on PS5 and DirectStorage PC games****
- SAMSUNG MAGICIAN SOFTWARE: Get the most out of your SSD with Samsung Magician's advanced yet intuitive optimization tools; Monitor drive health, protect valuable data, and receive important updates for your 990 PRO
Data pattern
State whether the tool used random data, zero-fill data, or another pattern. CrystalDiskMark documents that some SSDs react differently to random and zero-fill input. Two apparently identical profiles can therefore produce different scores if their data modes differ.
Application experience
If the reader cares about an actual activity, measure that activity separately. Define the files, application version, system state, and start and end points; repeat it under the same conditions; and report elapsed time. A controlled project load, compile, game launch, or large-file copy is an additional application-level result, not a conversion of synthetic MB/s into a guaranteed user experience.
How to compare two or more SSDs fairly
Use a matched matrix rather than collecting each drive’s advertised or default score independently.
| Comparison variable | Requirement |
|---|---|
| Host and connection | Same motherboard or enclosure path, negotiated link, operating system, and relevant drivers. |
| Software | Same benchmark version and profile; record defaults, especially when a tool version changes them. |
| Workload | Same test-file size, data pattern, access pattern, block size, read/write mix, queue depth, and worker count. |
| Timing | Same warm-up, measurement duration, and number of repetitions. |
| Cache policy | Same cache settings and a test size appropriate to the question. |
| Drive state | Comparable fill level, thermal condition, and preconditioning or fresh state. |
| Reported outputs | Bandwidth, IOPS, and latency distribution for the workload that matters. |
If the tools or defaults differ, rerun both drives under matched settings. DiskSpd 2.3’s default changes are a practical example of why a version label alone is not enough; preserve the actual configuration and rebaseline when defaults change.
Compare the right result for the right claim
- For large media transfers, compare matched sequential read or write results and a controlled copy or export.
- For multitasking or metadata-heavy work, compare the same small-block random workload at the same queue depth and inspect latency.
- For a write-heavy workflow, publish short-run burst and long-run behavior as separate results.
- For everyday responsiveness, give application timings more weight than a peak sequential headline.
Common benchmarking mistakes and fixes
Mistake: treating one headline score as the drive’s speed
Fix: Name the workload in the result title and include its settings. “Sequential read, 1 MiB blocks, queue depth 4” is informative; “Drive X is faster” is incomplete.
Mistake: comparing different presets
Fix: Match software version, profile, block size, queue depth, data mode, cache policy, duration, and drive state. If any variable differs, rerun the test or qualify the comparison.
Recommended Free Tools
Mistake: using a test file that fits in memory
Fix: Increase the file size when measuring device behavior, configure caching deliberately, and state the test size. Otherwise, label the result as cache-influenced rather than sustained device throughput.
Best Value
- 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).
Mistake: reporting a short burst as sustained speed
Fix: State the run length. Use preconditioning and a verified steady-state procedure for a sustained claim; otherwise publish only the short-run result.
Mistake: overreading high queue-depth IOPS
Fix: Pair the result with a low-queue-depth test or an application workload when the target reader uses a desktop. Queue depth is part of the workload, not a footnote.
Mistake: trusting unstable or tiny latency samples
Fix: Warm up, run long enough to collect adequate operations, and show percentile context. Do not infer tail behavior from a handful of samples.
Mistake: hiding run-to-run variation
Fix: Repeat the same test, report the spread or selected summary statistic, and investigate large changes caused by thermal throttling, background activity, cache state, or drive state.
A publication-ready result template
Use a compact test card before the charts so readers can judge whether a comparison applies to them.
- Drive: model, capacity, interface, firmware if available, and fill state.
- Host: processor or platform identifier where relevant, motherboard or enclosure connection, operating system, link/interface, and thermal setup.
- Tool: name, exact version, profile or command, and any changed defaults.
- Workload: target, file size, data mode, block size, random or sequential access, read/write mix, queue depth, workers, and cache policy.
- Timing: warm-up, measurement duration, repetition count, and whether preconditioning was used.
- Results: read and write bandwidth where relevant, IOPS, latency distribution and percentiles, plus any application-level timing.
- Qualification: whether the result represents a short burst, a modeled workload, or a verified steady-state procedure.
This format makes a number auditable without pretending that a synthetic score predicts every workload.
When a synthetic benchmark is enough—and when it is not
Use synthetic tests when you need a repeatable comparison of storage behavior under controlled conditions, when evaluating an interface or queue-depth response, or when isolating a change in platform configuration. Their strength is repeatability.
Add representative or application tests when the decision concerns user-visible time. A model workload can approximate a task, but it remains a model. For a real claim about copying, compiling, loading, or exporting, measure that operation under controlled conditions and keep the synthetic result as context rather than a substitute.
The most defensible SSD comparison therefore has two layers: controlled synthetic workloads that explain how the devices behave under specified I/O, and a carefully defined task that shows what happened in the application the reader actually uses.
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.
