What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
A synthetic SSD benchmark measures the workload you configure, not a universal speed rating for the drive. To get a useful result, hold the platform and settings constant, document every condition, test both peak and longer-run behavior, and add an application-level test that resembles what you actually do. Sequential MB/s, random IOPS and latency can reveal different bottlenecks; none alone predicts boot time, game loading or project responsiveness.
Why a high benchmark score can feel slow
An SSD sits inside an I/O stack. An application issues requests; the file system, operating system, drivers, controller, interface and caches can transform those requests before they reach the NAND. The same drive can therefore produce very different results when you change block size, access pattern, read/write mix, queue depth, data pattern, cache policy or test duration. The SNIA Solid State Storage Initiative describes workloads through their access patterns and demand intensity, and explains that synthetic models simplify real workloads rather than reproducing any one application exactly. SNIA SSD Performance: A Primer
As an Amazon Associate I earn from qualifying purchases.
That is why a sequential result of several thousand MB/s is not a general-purpose rating. It describes large, contiguous transfers under one configuration. A desktop that opens many small files at low queue depth may exercise the drive very differently. Conversely, a random-IOPS result measured with many outstanding requests can show how a device behaves under heavy parallel load while saying little about a lightly loaded laptop.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Use synthetic tests for repeatable comparisons, then use a representative workload for practical expectations. Label the latter as a model of your task, not a perfect reproduction. This distinction lets you compare drives fairly without promising an application experience the benchmark never measured.
#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
Start by defining the question you want the benchmark to answer
Write the question before opening a benchmark tool. The test design follows from it.
- Large-file transfer: measure sequential reads and writes with a block size and queue depth appropriate to the transfer. This answers how the drive handles contiguous data, not how quickly it launches an application.
- Small-file or operating-system activity: examine random reads and writes, usually at smaller blocks and lower queue depth. Report the exact block size and queue depth because changing either changes the workload.
- Heavy parallel work: use multiple workers or a higher queue depth to model concurrent jobs, servers or demanding storage queues. Do not present that result as a typical single-user desktop score.
- Long writes: run long enough to expose behavior after any fast write cache is exhausted. A short run can describe a burst, but it cannot establish indefinite sustained speed.
- Application responsiveness: time the activity itself—such as copying a defined folder, compiling a fixed project, loading a repeatable game level or opening a standard work set—under controlled conditions in addition to synthetic tests.
A benchmark is useful only when its workload matches the decision you are making. If you are choosing a drive for video files, a sequential result matters more than a headline random-IOPS number. If you are investigating sluggish project loads, an application-level timing and a small-block test are more informative than a peak sequential write.
Build a controlled test setup
Comparisons are valid only when the variables that matter are held constant. Record the following for every run:
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →| Item | What to record | Why it matters |
|---|---|---|
| Drive identity | Make, model and capacity | Different capacities or firmware revisions can use different NAND layouts and controllers. |
| Connection | SATA or NVMe, motherboard slot or enclosure, negotiated link/interface | The host connection can limit throughput or add latency. |
| Host platform | System model, processor, memory, operating system and relevant storage drivers | The I/O stack and available memory affect what reaches the drive. |
| Drive state | Used capacity, free space, whether the drive is fresh out of the box or previously written | Flash-management behavior can change as a drive fills and accumulates writes. |
| Thermal state | Whether the drive was cool, warmed by earlier tests or thermally constrained | Long transfers can expose heat-related throttling. |
| Software | Benchmark name and exact version | Defaults and result calculations can change between releases. |
| Workload | Test size, block size, random or sequential pattern, read/write mix, queue depth and workers | These parameters define the workload represented by the score. |
| Timing and cache | Warm-up, measurement duration, cache policy and test-data mode | They determine whether you measured a short cache burst or device behavior over time. |
These are practical reporting fields for a reproducible consumer test. They complement, rather than replace, the controlled-platform principle in the SNIA primer, which recommends comparing block-I/O tests on identical platforms.
Choose a tool that exposes the workload
| Tool | Platform and strength | Use it when | Important disclosure |
|---|---|---|---|
| CrystalDiskMark | Windows graphical quick-start benchmark | You need a familiar baseline for a consumer comparison. | Record the test count, test size, selected profile and test-data mode. Its documentation notes that some SSDs respond differently to random data and zero-fill data: CrystalDiskMark main menu documentation. |
| DiskSpd | Windows synthetic benchmark with configurable workloads | You need explicit control over block size, random or sequential access, queue depth, workers, duration and caching. | Capture the version and configuration. The Microsoft repository lists DiskSpd 2.3, released September 3, 2026, and warns that two default changes may require rebaselining: DiskSpd repository. |
| fio | Configurable Linux I/O workload generator | You need repeatable Linux tests with defined I/O size and sequential or random reads and writes. | Record the job configuration, target, block size, access pattern, read/write mix, queue depth and runtime. Microsoft’s benchmark guide describes fio as configurable for these dimensions: Microsoft disk benchmark guide. |
Do not compare a default result from one tool with a heavily customized result from another and call the difference a drive advantage. If tools or defaults differ, rerun both drives under matched settings.
A reproducible SSD benchmarking workflow
1. Prepare the system and record its state
- Identify the SSD, capacity, firmware if available, SATA or NVMe connection, host slot or enclosure and negotiated link.
- Record the operating system, system memory and storage-driver environment.
- Note used and free capacity. Keep the same fill state when comparing two drives, or state clearly that the states differ.
- Let the system settle after boot and close jobs that could create background I/O. If the drive has been working hard, note whether it is warm before the run.
- Choose a target that can be tested without risking important data. Back up anything on a target that a write test could overwrite.
Connection, fill state and thermal condition are not interchangeable details. A drive in an external enclosure, for example, is being measured together with that enclosure and its interface.
Rank #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
2. Capture the benchmark version and defaults
Install one tool and write down its exact version. For DiskSpd, the current repository entry identifies version 2.3 (Microsoft, 2026; release dated September 3, 2026). Because that entry calls out default changes, a result from an older version is not automatically comparable with a new result. Keep the configuration file, command or exported report with your notes. DiskSpd’s customization guidance explains the workload controls: DiskSpd customizing tests.
3. Run a small baseline, but label it as a baseline
On Windows, a default CrystalDiskMark profile is a practical first pass. Record its test count, test size, profile and data mode exactly as shown in the interface. Treat it as a repeatable reference, not as a universal desktop workload. If the utility offers random-data and zero-fill modes, select one deliberately and report it; the two modes can produce different SSD behavior.
A baseline is useful for spotting a wrong link, a severely underperforming drive or a large difference between two otherwise identical systems. It is not enough to answer whether a particular application will feel faster.
4. Add workload-specific tests
Use configurable tests when the question requires more than the baseline:
- Sequential large-block test: reports contiguous-transfer bandwidth. Use the same block size, queue depth and read/write mix for every drive.
- Small random test: reports behavior for scattered requests. State the block size and queue depth; “random IOPS” without those values is incomplete.
- Mixed read/write test: models a workload that reads existing data while writing new data. Record the exact mix, such as the proportion of reads and writes.
- Parallel test: vary workers or queue depth only when that concurrency represents the intended task. A high queue depth can saturate a device that a lightly loaded desktop never does.
Microsoft’s examples use 4 KiB random I/O for IOPS-oriented scenarios and 64 KiB blocks when examining throughput. Those settings are examples for a particular Azure virtual-machine configuration, not universal consumer presets; copy the reasoning and disclosure, not the numbers blindly. Microsoft disk benchmark guide
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 match5. Make the test file large enough to avoid measuring memory cache
If the test file fits comfortably in system memory, the operating system can satisfy part of the workload from cache. You may then be measuring cache behavior rather than the SSD. Microsoft’s example uses a 200 GiB test file and explicitly configures cache behavior to reduce that risk. That is an environment-specific example, not a required desktop value. Choose a size appropriate to your available storage and memory, state it, and use the same size for every drive. Microsoft disk benchmark guide
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]
6. Warm up, measure, repeat and report the spread
Include a warm-up period before recording measurements when the tool supports one. Microsoft’s benchmark examples separate warm-up from the measurement interval. For a consumer comparison, run multiple comparable passes after the system has settled, preserve every result and report the spread rather than selecting only the fastest run. Large variation is a diagnostic signal: investigate background activity, temperature, cache state, power policy, link negotiation or an exhausted write cache.
Do not average together runs made with different settings or different drive states. A mean computed from unlike workloads gives a precise-looking number with no clear meaning.
7. Test long writes separately from burst performance
Many SSDs can accept a short burst at a higher rate than they can maintain after internal write cache is consumed. SNIA distinguishes fresh-out-of-box peak behavior from steady-state measurement and describes preconditioning as part of preparing a drive for steady-state tests. SNIA performance-test specification paper, October 2020
State the duration and the drive state. A 30-second run can characterize a 30-second burst; it cannot prove indefinite sustained speed. If you claim steady state, the procedure must include an appropriate preconditioning phase and evidence that performance has reached the intended stable condition. Otherwise, describe the result as short-run or post-cache behavior.
8. Add a controlled application-level check
For a real-world claim, create a repeatable task with fixed inputs. Examples include copying the same large folder from a specified source to the test drive, compiling the same project from a clean starting state, loading the same project, or starting the same game level. Keep the source drive, operating-system state, background processes, file set and measurement method constant. Run the task more than once and report the range.
This result is still a model: the next project or game may issue different I/O. It is nevertheless closer to the reader’s question than an isolated synthetic score, and it exposes delays caused by the rest of the I/O stack that a block test may not show.
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
How to read the numbers
| Metric | What it describes | What it does not prove |
|---|---|---|
| Sequential MB/s | Bandwidth for large contiguous transfers under the selected block size, queue depth and read/write direction. | It is not a universal rating for boot time, game loading or application responsiveness. |
| Random IOPS | Completed random I/O operations per second for the specified block size, read/write mix and concurrency. | A high value at high queue depth may not resemble a lightly loaded desktop. |
| Average latency | Mean time per operation during the measured interval. | It can hide occasional long pauses that affect interactive work. |
| Latency percentiles | The latency below which a stated proportion of operations completed, such as a high-percentile tail. | A percentile calculated from too few operations is unstable and can collapse to a misleading maximum. |
| Application time | Elapsed time for the defined real task on the tested system. | It does not transfer automatically to a different project, source drive, operating system or background load. |
DiskSpd reports bandwidth and IOPS, and its result-analysis guidance cautions that high-percentile latency estimates require enough operations and an adequate test duration. In its example, tail percentiles collapse to the maximum when the sample is too small. Treat a minimum or a single average as an incomplete latency report. DiskSpd result analysis
Peak, cache and sustained behavior are different measurements
Short bursts
A fresh drive or a drive with available write cache may show a high initial write rate. This is useful if your workload consists mostly of short bursts, but the result should be labeled as such. Do not call it “sustained” merely because the benchmark completed successfully.
Cache-limited tests
A small test file, enabled operating-system caching or a zero-fill pattern can make the host and controller do less physical work than a real data stream. Disclose cache settings, file size and data pattern so another reader can understand what was measured.
Steady-state preparation
Long-running or write-heavy comparisons need a defined preparation phase. SNIA’s October 2020 paper discusses preconditioning and the distinction between fresh-out-of-box peak and steady-state performance. If you do not perform and verify that preparation, publish the run duration and call the result a long-run observation rather than a steady-state specification. SNIA performance-test specification paper
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Compare two SSDs without changing the rules
Use this checklist before declaring a winner:
- Same host, operating system, drivers and physical connection type.
- Same benchmark software version, profile and defaults.
- Same target capacity class, test-file size and data pattern.
- Same block size, access pattern, read/write mix, queue depth and worker count.
- Same cache policy, warm-up interval and measurement duration.
- Comparable used capacity, temperature and preconditioning or fresh-drive state.
- Multiple runs retained, with the range or variation reported.
- Bandwidth, IOPS and latency shown for the workload that matters, not only the most flattering number.
If one drive was tested through an enclosure and the other in a motherboard slot, you tested two storage paths, not just two SSDs. If one result came from DiskSpd 2.3 and another from an earlier release with different defaults, rerun the comparison or rebaseline both sets. The DiskSpd repository specifically warns that default changes can affect comparability. DiskSpd repository
Free tools Windows power users keep installed
One-click scans. No signup required.
Diagnose surprising or inconsistent results
The score is far below the advertised interface speed
- Verify that the SSD is connected through the intended SATA or NVMe link and that the negotiated interface is correct.
- Check whether an enclosure, adapter or expansion slot is limiting the path.
- Confirm that the benchmark is measuring the intended target rather than a cache or another volume.
- Compare a large sequential test with a small random test; a low random result does not necessarily indicate a faulty sequential path.
Results vary widely between runs
- Wait for background updates, indexing, synchronization and other storage activity to finish.
- Keep warm-up, test duration, file size and cache settings identical.
- Record temperature and power mode; investigate throttling after repeated long writes.
- Check whether the drive’s short write cache was exhausted in only some runs.
The first run is fast and later runs collapse
This pattern can indicate a burst cache or a change from fresh to written drive state. Separate short-run results from long-run results, preserve the duration and do not average them into one headline figure.
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).
Latency percentiles look identical to the maximum
Inspect the operation count and measurement interval. DiskSpd’s guidance explains that high-percentile estimates are unreliable with too few samples; lengthen the run or increase the number of operations before interpreting the tail. DiskSpd result analysis
Two tools disagree
First compare their actual workloads: block size, random or sequential mode, queue depth, workers, data pattern, caching, duration and result units. Different defaults can produce valid but non-comparable measurements. Recreate one tool’s settings in the other where possible, or report the tools as separate experiments rather than ranking the drives from unlike tests.
Publish a result readers can reproduce
A useful report puts the conditions next to the numbers. Include:
| Report section | Minimum information |
|---|---|
| Platform | SSD model and capacity; SATA or NVMe path; host system; operating system; memory; enclosure or slot. |
| Drive state | Used/free capacity; fresh or previously written state; temperature or thermal condition when relevant. |
| Software | Tool name and version; profile or configuration; any changed defaults. |
| Workload | Test-file size; data mode; block size; sequential or random access; read/write mix; queue depth; workers; cache policy. |
| Timing | Warm-up duration; measurement duration; number of runs; whether the result is a burst observation or a prepared long-run test. |
| Results | Read and write bandwidth where relevant; IOPS; latency distribution; application-level elapsed time and run-to-run spread. |
DiskSpd can export configuration and system information in XML, which helps preserve the exact setup alongside the output. DiskSpd result analysis Keep the raw output, not just a screenshot of the best score.
A compact decision framework
- Need a quick sanity check? Run the same default profile on the same host, record the profile and data mode, and use it only as a baseline.
- Need to compare drives? Match platform, software version, workload parameters, cache policy, drive state and timing; publish the full context.
- Need to predict a task? Add a controlled application-level test with the actual file set or project, then explain that it models that task rather than every workload.
- Need sustained-write evidence? Use a long run with documented preconditioning or state clearly that the result is a finite-duration observation.
- Need to understand responsiveness? Examine latency distribution and low-queue-depth random behavior, not only peak sequential MB/s or high-queue-depth IOPS.
The most honest conclusion is usually conditional: one drive may lead in sequential reads, another in a particular random workload, and neither difference may produce a visible advantage in the application you care about. State which workload produced the advantage and leave the score attached to that workload.
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.
