CrystalDiskMark is the best SSD benchmark tool for most people because it’s free, widely used enough that its results are directly comparable to manufacturer specifications and other reviews, and its default preset covers both sequential and random access patterns in one run. AS SSD Benchmark is a solid second option, particularly useful for its copy-benchmark test simulating real file transfer scenarios that CrystalDiskMark doesn’t directly replicate.
Top 3 picks at a glance
Quick answer
Install CrystalDiskMark, run it on a drive with reasonable free space (not nearly full), and compare the sequential read/write (SEQ1M Q8T1) numbers against your drive’s advertised specs, plus the random 4K Q1T1 numbers for a sense of real-world responsiveness. If you want a second data point or a real-file-copy simulation, add AS SSD Benchmark’s copy-benchmark test to the mix.
As an Amazon Associate we earn from qualifying purchases at no extra cost to you. Product prices and availability are accurate as of the date shown and are subject to change.
| Tool | Strength | Best test to check |
|---|---|---|
| CrystalDiskMark | Industry standard, matches spec sheets | SEQ1M Q8T1 and RND4K Q1T1 |
| AS SSD Benchmark | Real-file copy simulation | Copy-benchmark (ISO/Program/Game presets) |
Why benchmark an SSD at all
A benchmark confirms a drive is actually delivering the performance it was purchased for, which matters because SSDs can underperform their rated specs for reasons that have nothing to do with a defective unit: an M.2 slot sharing PCIe lanes with another device, insufficient cooling causing thermal throttling on sustained transfers, or simply being installed in a PCIe generation slower than the drive supports. None of these show up in Windows’ basic drive properties; a benchmark is the only way to see the actual number.
Benchmarking is also the standard way to validate a fresh install before trusting a system to it, similar in spirit to the memory testing covered in our RAM testing guide. Confirming a new boot drive hits its expected speed in the first hour of ownership is easier than debugging vague slowness months later after other variables have piled up.
For anyone comparing a potential upgrade, published third-party reviews use these same tools, so understanding what the numbers mean lets you interpret review data more critically rather than just trusting a headline number, and lets you reproduce those same tests on your own system to see if you’re getting comparable results to what was reviewed.
Benchmarking pairs naturally with the health monitoring covered in our SSD health checking guide; a drive with a clean SMART status but underperforming benchmark numbers points toward a configuration or thermal issue rather than the drive itself failing, which changes what you should actually troubleshoot.
Setting up and running CrystalDiskMark
CrystalDiskMark is free and available from its official GitHub releases page or the Microsoft Store. On launch, it shows a simple interface with a drive selector, a test count and size selector, and a grid of test types that runs automatically when you click the start icon next to each row, or all at once with the “All” button.
The default test profile (labeled something like “Default” or “NVMe SSD” depending on the version) is appropriate for most gaming PC drives and includes four rows: SEQ1M Q8T1, SEQ1M Q1T1, RND4K Q8T1, and RND4K Q1T1. Each label describes the test pattern: sequential (SEQ) or random (RND) access, block size (1M or 4K), queue depth (Q8 or Q1, meaning how many operations are queued simultaneously), and thread count (T1 or T8).
Test file size options typically range from 1GB up to 64GB or more; a 1GB test is fast and fine for a quick check, but larger test sizes (8GB or more) give more consistent results on drives with a fast SLC cache that can make small tests look artificially faster than sustained performance would actually be. If you specifically want to see sustained speed beyond the SLC cache, use a test size larger than your drive’s cache capacity, which for most consumer NVMe drives falls somewhere in the 4-24GB range depending on the model.
Running the benchmark on a mostly-idle system, closing other disk-intensive applications first, gives more consistent and repeatable numbers, since background processes competing for the same drive can add noise to the results that has nothing to do with the drive’s actual capability.
Interpreting the four core numbers
SEQ1M Q8T1 (sequential, large blocks, high queue depth) represents best-case large file transfer performance, close to what manufacturers usually advertise as a drive’s headline read/write speed. This is the number that matters most for tasks like copying large game installations or media files in bulk.
SEQ1M Q1T1 (sequential, large blocks, single queue) is a more realistic single-file-copy scenario without the parallelism of Q8, and is typically noticeably lower than the Q8 figure, sometimes by a significant margin depending on the drive’s controller. This gap is normal and expected rather than a sign of a problem.
RND4K Q8T1 (random, small blocks, high queue depth) simulates heavily parallel small-file access, relevant for scenarios like running a database or handling many simultaneous small operations, less directly relevant to typical gaming workloads.
RND4K Q1T1 (random, small blocks, single queue) is arguably the most relevant number for actual gaming responsiveness, since loading a game level involves reading many small, scattered files essentially one at a time from the operating system’s perspective. A drive with excellent sequential numbers but comparatively weak RND4K Q1T1 performance may not feel as fast during actual game loading as its headline spec suggests.
AS SSD Benchmark and its copy-benchmark test
AS SSD Benchmark predates CrystalDiskMark’s dominance and remains useful today primarily for its copy-benchmark feature, which simulates copying real file types (an ISO disc image, a set of program files, or a game folder with mixed file sizes) rather than a purely synthetic access pattern. This gives a result in seconds and MB/s that maps more intuitively to “how long will copying my Steam library actually take” than raw sequential numbers do.
The main AS SSD test grid looks similar to CrystalDiskMark’s, with sequential and random tests at different queue depths, though AS SSD deliberately disables a drive’s compression optimizations for some of its tests by using incompressible data, which can make results look slightly different, sometimes lower, than CrystalDiskMark’s for the same physical drive. Neither approach is wrong; they’re simply measuring slightly different things, which is why comparing a single drive across both tools rather than mixing results between different drives tested on different tools gives more meaningful conclusions.
AS SSD also includes a simple overall score combining its test results into one number, useful mainly as a quick comparison if you’re testing the same drive before and after a change (like a firmware update or moving to a different M.2 slot) rather than as an absolute quality measure on its own.
Why real numbers fall short of advertised specs
Manufacturer-advertised speeds are peak figures achieved under specific, often best-case conditions: a particular queue depth, a specific PCIe generation, and sometimes a larger drive capacity within a product line than the one you own, since higher-capacity SKUs in the same drive family often have more NAND channels and slightly higher peak speeds than the smallest capacity option. Comparing your benchmark results against the specific capacity you purchased, not the top-tier model in a marketing image, is a fairer comparison.
PCIe generation and lane allocation matter significantly for NVMe drives; a PCIe 4.0 drive installed in a motherboard slot that only supports PCIe 3.0 speeds (common on some secondary M.2 slots, or when connected through a chipset rather than directly to the CPU) will be capped well below its rated speed regardless of the drive’s own capability. Checking your specific motherboard manual for which M.2 slots run at full speed, and which share bandwidth with other components, explains a lot of otherwise confusing benchmark shortfalls.
Thermal throttling is another common gap source, particularly on drives without a heatsink in cases with limited airflow around the M.2 slot; sustained writes during a longer benchmark test size can trigger throttling that a shorter test wouldn’t reveal, which is part of why testing at a larger file size, as mentioned earlier, gives a more honest picture of real sustained performance.
Comparing before and after a system change
Benchmarking is most informative when used comparatively rather than as an isolated one-time check, particularly around a specific change: moving a drive to a different M.2 slot, updating motherboard BIOS or drive firmware, or adding a heatsink to a previously bare drive. Running the same test with the same settings before and after isolates whether that specific change actually delivered the improvement you expected, rather than relying on assumption.
Keeping a simple record of test size, drive temperature at the time of testing, and the resulting numbers makes this kind of before-and-after comparison meaningfully more useful, since a result taken without matching conditions (a cooler ambient temperature on one test day versus a warmer one, for instance) can create a misleading sense of improvement or regression that isn’t really about the change being tested.
This comparative approach is also the standard practice used in published hardware reviews, including the drive comparisons on this site; testing multiple drives under identical conditions, rather than trusting each manufacturer’s own advertised numbers directly, is what makes a head-to-head comparison meaningful in the first place.
Benchmarking in context of your actual workload
For pure gaming use, the most representative real-world test isn’t actually a synthetic benchmark at all, it’s loading time in an actual game, since real games access files in patterns that don’t map perfectly onto any single synthetic test. If you’re deciding whether a storage upgrade is worth it specifically for gaming, comparing actual load times between drives (using the same game and save point) is more informative than any benchmark score.
For content creators or streamers using a drive as a recording or editing scratch disk, sustained write performance at larger file sizes matters more than the burst SEQ1M Q8T1 number, since recording footage is a long, continuous write rather than a short burst; testing with a larger file size setting in CrystalDiskMark, closer to 16-32GB, better represents this use case.
Boot drive performance is felt most in random access patterns during Windows startup and application launches, making RND4K numbers more relevant there than sequential ones; this is part of why even budget NVMe drives, which often have competitive random 4K performance despite lower sequential numbers than premium drives, can feel similarly snappy for everyday desktop use.
When benchmark numbers don’t tell the whole story
A drive can post excellent benchmark numbers while still being a poor practical choice if it lacks power-loss protection, has a short warranty period, or uses a lower-durability NAND type that wears out faster under heavy write workloads; benchmarks measure speed, not reliability or longevity, which is a separate consideration covered in our SSD health checking guide.
Synthetic benchmarks also can’t capture consistency over time the way extended real-world use does; a drive that benchmarks well when new but has a smaller-than-average SLC cache can show noticeably reduced sustained performance during genuinely large transfers once that cache is exhausted, something a quick default-size benchmark run might not reveal.
How often to rebenchmark a drive
Unlike SMART health monitoring, covered in our SSD health checking guide, which is worth checking regularly, benchmarking doesn’t need to be a routine habit for a drive that’s performing as expected; a single confirmation shortly after setup, plus a rerun any time you suspect a problem or make a relevant hardware change, covers the practical need for most people without turning it into an unnecessary recurring chore.
A reasonable exception is a drive nearing full capacity, since write performance can degrade as free space shrinks on some SSD controllers; rebenchmarking after freeing up space, or before an important workload like a large game install, can confirm the drive is still performing well rather than assuming its original benchmark numbers still apply at a very different fill level.
Troubleshooting unexpected benchmark results
Results significantly below advertised specs, more than roughly 20-30% lower on sequential numbers, should first be checked against which M.2 slot and PCIe generation the drive is actually running at; Windows Device Manager or HWiNFO64 (covered in our HWiNFO64 setup guide) can confirm the negotiated PCIe link speed and width, which is the most common cause of this gap.
Inconsistent results between repeated runs on the same drive, varying by more than roughly 10%, often point to background processes competing for drive access during the test; close unnecessary applications, disable real-time antivirus scanning temporarily if possible, and rerun before assuming a hardware issue.
Unusually low random 4K performance specifically, while sequential numbers look normal, can indicate a nearly-full drive struggling with reduced free space for the controller to work with, or in rare cases outdated drive firmware; checking available free space and the manufacturer’s own storage management tool for a firmware update are both reasonable first steps.
If a drive that previously benchmarked well shows a sudden, significant drop in all test categories, this is worth cross-checking against SMART health data rather than assuming a benchmarking fluke, since a genuine drive-level problem developing over time would show up as declining performance before an outright failure in some cases.
Frequently asked questions
Which is more accurate, CrystalDiskMark or AS SSD Benchmark?
Neither is more “accurate” in an absolute sense since both are synthetic tests measuring different access patterns; CrystalDiskMark is more widely used and its presets are closer to what manufacturers quote, while AS SSD’s own scoring weights things slightly differently, so comparing a drive against itself across both tools gives a fuller picture than trusting one number alone.
Why is my SSD slower than the advertised speed?
Advertised speeds are peak sequential figures reached under ideal conditions with a specific queue depth and file size, which real usage rarely matches exactly; a gap of 10-20% between advertised and benchmarked speed is common and not necessarily a defect.
Do benchmark results matter for gaming performance?
Sequential speeds matter less for gaming than random 4K read performance, since loading game assets involves many small, scattered reads rather than one large continuous file transfer, so a drive’s random 4K QD1 score is often more representative of in-game load time impact.
Does running a benchmark wear out my SSD?
A single benchmark run writes a relatively small amount of data, typically a few gigabytes depending on the test size chosen, which is a negligible fraction of any modern SSD’s total rated write endurance.
Should I benchmark a drive that’s nearly full?
No, benchmark on a drive with meaningful free space if you want results representative of normal use, since many SSD controllers slow down write performance as available free space drops, which would make a nearly-full drive’s results look artificially worse than its typical performance.
Daniel Kovac, Components Editor, runs both CrystalDiskMark and AS SSD’s copy-benchmark on every drive tested with the PSU load tester bench, since the two occasionally disagree just enough to reveal something a single tool would miss, particularly around SLC cache exhaustion on budget drives. For current drive recommendations backed by this kind of testing, see our guides to the best SSD for gaming and best NVMe SSD for gaming.
Related guides
- Best Reviews of Piano in 2026: Specs and Prices Compared: 8 Tested (2026)
- Best Ge Profile Opal Mini Nugget Ice Maker in 2026: Top Picks for Every Budget: 8 Tested (2026)
- Best Chocolate Gift Box in 2026: Specs and Prices Compared: 8 Tested (2026)
- Best Microwave Convection Oven in 2026: Top Picks for Every Budget: 8 Tested (2026)
