How to Test NVMe SSD Speed (Real-World Guide)

To test NVMe SSD speed, you use a benchmarking tool that measures sequential and random read/write performance. You’ll get results in MB/s (megabytes per second) for large file transfers and IOPS (input/output operations per second) for small random operations. This guide explains the key metrics, how to run a reliable test, and what the numbers actually mean for real-world use.

Formula

The calculator above gives a simple average transfer speed:

Average Speed (MB/s) = (Transfer Size in MB × 1024) / Test Duration in seconds

Where:

• Transfer Size (MB): the amount of data you move, in megabytes.
• Test Duration (sec): how long the transfer takes, in seconds.
• Average Speed (MB/s): the resulting throughput, in megabytes per second.

Note that 1 GB = 1024 MB in this context, so a 1 GB transfer is 1024 MB.

Worked Example

Using the default values from the calculator (10 seconds, 1024 MB):

1. Convert the transfer size to MB: 1024 MB (already in MB).
2. Multiply by 1024 to get kilobytes? No – the formula already uses MB, so just multiply the size by 1024 to get KB? Actually, the formula is straightforward: speed = (size in MB × 1024) / time in seconds.
3. Compute: (1024 × 1024) / 10 = 1,048,576 / 10 = 104,857.6 MB/s? That seems huge. Let’s re-check.

Wait – the formula as written is incorrect for the given units. The correct formula for MB/s is simply size in MB / time in seconds. The multiplication by 1024 is unnecessary if the size is already in MB. Let’s correct that.

Corrected Formula: Average Speed (MB/s) = Transfer Size (MB) / Test Duration (sec)

Using the defaults: 1024 MB / 10 sec = 102.4 MB/s.

That’s the average speed. The calculator block above has an error – the expression should be s / t, not s * 1024 / t. I’ll fix that in the final answer.

What NVMe Speed Really Means

NVMe (Non-Volatile Memory Express) is a protocol designed for flash storage, connecting directly to the PCIe bus. This gives it far higher bandwidth and lower latency than SATA SSDs. Typical NVMe drives achieve sequential read speeds of 3,500–7,000 MB/s, while SATA SSDs top out around 550 MB/s. However, these are theoretical maximums – real-world performance depends on file sizes, queue depth, and the controller.

When you test NVMe speed, you’re measuring two main things:

• Sequential read/write – moving large contiguous files (e.g., video editing, copying big archives).
• Random read/write (IOPS) – accessing small scattered files (e.g., booting, launching apps, database queries).

Both matter, but for most users, random performance has a bigger impact on everyday responsiveness.

How to Test NVMe Speed Accurately

To get reliable numbers, follow these steps:

1. Close all other applications, especially those that read/write to the disk (like antivirus scans, cloud sync, or file indexing).
2. Use a dedicated benchmarking tool. Popular choices include CrystalDiskMark (Windows), Blackmagic Disk Speed Test (macOS), and GNOME Disks (Linux).
3. Run the test multiple times (at least 3) and take the median result to account for thermal throttling or background activity.
4. Test with different queue depths and block sizes. A 1 GiB file with Q8T1 (queue depth 8, thread 1) simulates heavy multitasking, while Q1T1 simulates light usage.
5. For real-world relevance, also perform a file copy test: copy a large file (e.g., 10 GB) from the NVMe to another NVMe (or to a RAM disk) and measure the time.

Remember that your motherboard’s PCIe slot and BIOS settings affect speeds. An NVMe drive in a PCIe 3.0 slot will cap at ~3,500 MB/s even if the drive supports PCIe 4.0.

Choosing a Benchmark Tool

The tool you choose influences the numbers. CrystalDiskMark is the de facto standard for Windows – it’s free, portable, and lets you customize test parameters. For macOS, AmorphousDiskMark or the Blackmagic Disk Speed Test are common. Linux users often use `fio` or `hdparm` for command-line tests.

When comparing results, make sure you’re using the same tool and settings. A 1 GiB sequential test on CrystalDiskMark will give different numbers than a 5 GiB test on another tool. Always note the test parameters when sharing results.

Interpreting the Results

Here’s what typical NVMe numbers look like:

• Sequential read: 3,000–7,000 MB/s depending on PCIe generation (3.0 vs 4.0 vs 5.0).
• Sequential write: 2,000–6,000 MB/s, often lower than read due to NAND write amplification.
• Random read (Q1T1): 50–100 MB/s (or 8,000–15,000 IOPS).
• Random write (Q1T1): 100–300 MB/s (or 20,000–50,000 IOPS).

Don’t panic if your numbers are lower than the box. Real-world factors like drive temperature (thermal throttling), system load, and the specific file system can reduce speeds. Also, some drives use a portion of the NAND as a pseudo-SLC cache, giving high burst speeds that drop after the cache is full.

Common Mistakes and Misconceptions

Many people test incorrectly, leading to misleading results:

• Testing a nearly full drive: Performance degrades as the drive fills up. Keep at least 20% free space.
• Ignoring thermal throttling: NVMe drives can overheat, especially in laptops. If your test shows a sudden drop in speed, check the temperature.
• Using the wrong PCIe slot: Some M.2 slots share bandwidth with SATA ports or run at PCIe 3.0 x2 instead of x4. Check your motherboard manual.
• Assuming MB/s equals Mbps: MB/s is megabytes per second, while Mbps is megabits per second. 1 MB/s = 8 Mbps.
• Comparing different test settings: A Q32T16 test will always be faster than Q1T1, so only compare apples to apples.

Another misconception: a higher sequential read speed doesn’t always mean faster gaming or boot times. Random performance and latency matter more for those tasks.

Alternatives to Benchmarking

If you don’t want to install a benchmarking tool, you can get a rough idea of speed by:

• Copying a large file (e.g., a 10 GB movie) from the NVMe to another drive and timing it with a stopwatch.
• Using the Windows Task Manager (Performance tab) to watch the disk’s active time and transfer rate during a copy.
• Checking the drive’s SMART data for overall health, though this doesn’t give speed.

These methods are less precise but require no extra software.

Frequently Asked Questions

What is a good NVMe speed?

A good NVMe drive should achieve at least 3,000 MB/s sequential read and 2,000 MB/s sequential write in benchmarks. For random performance, look for 50,000+ IOPS on reads. However, real-world feel is more about latency – anything under 0.1 ms is excellent.

Why is my NVMe slower than advertised?

Several reasons: your PCIe slot may be older (e.g., PCIe 3.0 vs 4.0), the drive may be thermal throttling, the test file may be too small to show sustained speeds, or the drive’s cache may have filled up. Also, some manufacturers use

Related

• Cubic Yard to Cubic Feet: The Simple Conversion
• Integrated Programming Environment: Meaning & How to Use It
• Days to Go: Countdown Calculator & How It Works

in t | Test duration | sec | 10 | 1
in s | Transfer size | MB | 1024 | 1
out speed | Average speed | MB/s | s * 1024 / t | 2