IOPS (input/output operations per second) counts how many storage requests an SSD completes in one second. It is not a universal speed score. A meaningful IOPS claim must identify the operation size, random or sequential access, read/write mix, queue depth, worker count, drive capacity and test state. A high maximum measured with many requests in flight can show strong concurrent performance while saying little about the responsiveness of a lightly threaded desktop task.
For a real workload, read IOPS alongside latency and the workload’s access pattern. Use random IOPS for small, scattered requests; use throughput for large, sequential transfers; and compare only benchmark recipes that are genuinely alike.
As an Amazon Associate I earn from qualifying purchases.
What IOPS measures—and what it leaves out
An I/O operation is one completed read or write request. IOPS is the count of those completed requests in a second, regardless of how many bytes each request moved. That distinction is why two headline results cannot be compared merely because both are labelled “IOPS.”
For example, 100,000 operations of 4 KiB each move about 409.6 MB/s (using decimal megabytes), while 100,000 operations of 128 KiB each move about 13.1 GB/s. The operation count is identical; the data rate is not. Operation size therefore has to travel with the number.
#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
IOPS also omits how long an individual request waited. A device can complete many requests per second by keeping a deep queue busy, while each request experiences more delay than it would at a shallow queue. Average latency, and especially high-percentile (“tail”) latency, are needed to judge responsiveness.
Read the benchmark recipe before reading the number
Before using an SSD specification, identify every part of its test recipe:
- Access pattern: random requests jump among addresses; sequential requests follow adjacent addresses.
- Transfer size: common examples are 4 KiB for small random I/O and 128 KiB for sequential testing.
- Operation mix: read-only, write-only, or a stated read/write proportion.
- Concurrency: queue depth and the number of workers or threads issuing requests.
- Device state: fresh out of the box (FOB) or sustained/steady-state operation after the drive has been used.
- Capacity and platform: performance can vary by the SSD’s capacity and the host system used for the test.
Micron’s 2400 specification illustrates why those details matter. Its random IOPS tests use 4 KiB transfers at queue depth 128 in fresh-out-of-box state, while its sequential read and write results use 128 KiB transfers at queue depth 32. The page lists results by capacity, so a quoted figure must include the tested capacity as well as these conditions. See the Micron 2400 SSD specifications for the product’s capacity-specific claims.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Micron 2400 test class | Transfer size | Queue depth | Drive condition |
|---|---|---|---|
| Random IOPS | 4 KiB | 128 | Fresh out of box |
| Sequential read/write | 128 KiB | 32 | Product test condition listed by Micron |
Those rows are different workloads, not interchangeable ways of expressing one speed. A random 4 KiB result should be compared with another random 4 KiB result using a similar queue and mix. A 128 KiB sequential result belongs with comparable sequential throughput tests.
Random IOPS versus sequential throughput
Random I/O
Random workloads issue small requests to non-adjacent locations. Operating-system metadata, application files, database indexes and many virtual-machine operations can behave this way. Random IOPS is useful because it counts the large number of small requests such workloads generate. The read/write ratio and block size still matter: a 70/30 read/write test is not equivalent to read-only testing, even if both use 4 KiB operations.
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
Sequential I/O
Sequential workloads move large, contiguous ranges, such as copying a large video or reading a disk-image backup. Throughput in MB/s or GB/s is normally the clearer measure because the useful question is how many bytes arrive per second. A sequential IOPS figure without its transfer size can conceal the actual bandwidth.
The Open Compute Project’s NVMe Cloud SSD Specification separates these cases explicitly: its example random workloads use 4 KiB operations with different queue depths and write proportions, while sequential metrics use 128 KiB transfers. It also defines latency requirements at multiple percentiles. This structure is a practical model for evaluating any SSD specification.
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 →Queue depth: why headline IOPS often uses QD128
Queue depth (QD) is the number of I/O commands outstanding at a time. At QD1, the host generally waits for one request to finish before issuing the next. At QD128, up to 128 requests can be in flight, allowing the controller and flash channels to work on several requests concurrently.
More outstanding work can raise the measured completion rate until the device approaches saturation. It can also increase waiting time: requests line up behind earlier work and compete for internal flash resources. Intel describes SSD scaling as nonlinear; performance eventually flattens as queue depth rises while latency can continue to increase. Its technical explanation is available in the Intel PCIe/NVMe SSD performance paper.
That is why a QD128 maximum is evidence of a drive’s ability to serve highly concurrent work, not a promise that a browser launch or an office application will run 128 times faster than at QD1. Lightly threaded tasks may never create enough simultaneous requests to reach that part of the performance curve.
Rank #3
- 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).
NVMe’s protocol can support up to 65,535 I/O queues, with up to 65,535 commands in each queue, according to the NVM Express specification FAQ. These are architectural limits. They are not a normal desktop workload, nor do they guarantee linear scaling as more queues or commands are added.
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 minuteLatency completes the performance picture
Latency is the elapsed time from issuing an I/O request to receiving its completion. At low queue depth, it often dominates how responsive an application feels. At higher depth, queue waiting can become a substantial part of the end-to-end delay even while aggregate IOPS rises.
Do not rely only on an average. A workload may have a low mean latency but occasional very slow requests that cause visible pauses. Percentiles such as the 95th, 99th or 99.9th show those tail events when a benchmark reports them. The Open Compute Project specification sets latency requirements at several percentiles for this reason.
Amazon Web Services defines EBS latency as the end-to-end interval between sending an I/O and receiving its completion acknowledgement, and notes: “Optimal queue length varies for each workload, depending on your particular application’s sensitivity to IOPS and latency.” That guidance concerns EBS volumes rather than the internal specification of an NVMe SSD, but the queueing principle applies when you interpret storage tests. AWS explains the trade-off in its EBS I/O characteristics and monitoring documentation.
How to compare two NVMe SSDs fairly
Use this order when a product page presents competing numbers:
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
- Match access pattern. Compare random with random and sequential with sequential.
- Match transfer size. A 4 KiB result and a 128 KiB result describe different work. Do not rank them as if they were equivalent.
- Match read/write mix. Record whether each result is read-only, write-only or mixed, including the stated proportions.
- Match concurrency. Align queue depth and worker count. If one result is QD1 and the other QD128, label the difference rather than averaging it away.
- Check test state. FOB results can be higher than sustained results after the drive has filled and garbage collection has occurred.
- Check capacity and host conditions. Capacity-specific results and the test platform belong in the comparison.
- Inspect latency. Prefer a result that includes average and tail latency when responsiveness matters, rather than choosing on peak IOPS alone.
If a specification omits several of these fields, treat its IOPS number as incomplete evidence. It can still indicate the manufacturer’s chosen test result, but it cannot predict every application.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What IOPS means for common real workloads
Everyday desktop use
Booting, launching programs and opening documents combine small random reads with periods of sequential reading. They may be limited by application startup work, CPU processing or a shallow queue rather than by the SSD’s deepest-queue maximum. Low latency and consistent response at modest queue depth are more informative than a single peak random-IOPS line.
Games
Game loading can mix metadata lookups, many small reads and larger sequential asset transfers. A drive with a high random-IOPS claim is not automatically faster for every title; the game’s access pattern and the platform’s software stack determine which part of the SSD’s performance matters.
Compilers, content tools and project files
Build systems and media applications can issue many requests concurrently, so higher queue depths may be relevant. Still, compare the application’s read/write mix and transfer sizes with the test recipe. A sequential-workload result is useful for large media streams, while random results are more relevant to scattered project files and metadata.
Databases and virtualization
These workloads can generate sustained, mixed random I/O and several active workers. Queue depth, write proportion, steady-state behavior and tail latency deserve priority over a fresh-drive peak. A benchmark that resembles the production mix is more valuable than the largest number on a consumer product page.
Best Value
- 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]
How many IOPS do you need?
There is no single target that fits every NVMe SSD buyer. Start with the workload rather than a marketing tier:
- Measure or estimate request size and whether access is random or sequential.
- Determine the read/write ratio and how many concurrent workers your application creates.
- Observe latency sensitivity: an interactive task may benefit more from low tail latency than from extra peak IOPS.
- Choose comparison tests using the same recipe, then verify that the SSD, capacity and interface are compatible with the computer.
For an unknown workload, state the uncertainty openly: peak IOPS alone cannot predict performance across all real applications. A drive with a lower advertised maximum can feel faster in a shallow-queue, latency-sensitive task if its relevant behavior is better.
NVMe, PCIe and current specification context
NVMe is the protocol that defines how host software communicates with non-volatile memory over PCI Express; the NVM Express overview describes this purpose and identifies M.2 among NVMe SSD form factors. NVM Express lists NVMe 2.4 as released on August 4, 2026. Specification revisions can change, so verify the current revision on the NVM Express Base Specification page when a version detail affects a purchase or deployment.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →The protocol’s large theoretical queue architecture helps NVMe handle parallel workloads, but the SSD controller, NAND configuration, firmware, PCIe link and host software determine the performance a particular application sees. Protocol limits should therefore be treated as capability boundaries, not as expected IOPS.
Common IOPS mistakes and their fixes
- Mistake: treating IOPS as bytes per second. Fix: check transfer size and use throughput for large sequential work.
- Mistake: comparing random and sequential results. Fix: keep access patterns separate.
- Mistake: ignoring queue depth. Fix: record QD and worker count; a QD128 peak is not a QD1 promise.
- Mistake: quoting a number without its read/write mix. Fix: include the mix or decline to compare.
- Mistake: assuming FOB equals sustained performance. Fix: look for steady-state testing when the drive will be heavily used.
- Mistake: choosing on peak IOPS while ignoring latency. Fix: inspect average and tail latency, especially for interactive or transactional work.
- Mistake: generalizing one capacity’s result to every capacity. Fix: use the result for the exact capacity and test platform being considered.
A practical reading checklist
When you encounter an NVMe SSD IOPS claim, rewrite it in a complete sentence: “This drive completed [read/write mix] [random/sequential] operations of [transfer size] at [queue depth and workers], in [FOB or sustained] state, for [capacity], with [latency information].” If you cannot fill several blanks, the claim is not a like-for-like performance comparison.
IOPS is most useful as a workload-shaped measurement. Pair it with throughput where bytes matter, queue depth where concurrency matters, and latency where people or applications wait. That combination turns a headline number into evidence you can use.
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.
Recommended Free Tools
