CrystalDiskMark

中文

CrystalDiskMark 是 Windows 磁盘测速工具，用于测量 HDD、SSD、USB 存储和存储卡的读写表现。本文说明 9.0.3 更新及下载、基准测试参数、结果单位与安装方式，并解释如何统一设置进行 SSD 速度对比，以及测试失败或成绩波动时可先检查哪些条件，同时区分性能测试与硬盘健康检测。

CrystalDiskMark 可以测什么

CrystalDiskMark 用读写基准测试观察 HDD、SSD、USB 闪存、存储卡和网络盘的存储性能。它可显示顺序与随机测试结果，并提供 Peak、Real World、Demo 等测试模式，适合在相同条件下比较存储设备的传输表现。它不是 SMART 健康监测、坏道检查或寿命预测工具，单次分数也不能代表所有日常应用的实际速度。

图示是原创概念图，不代表 CrystalDiskMark 的真实界面或实测数据。产品测试支持范围与模式说明见官方产品页及主窗口手册。

测试次数、文件大小和目标盘怎么选

官方手册列出 1–9 次测试次数，默认值为 5；测试文件大小范围从 16 MiB 到 64 GiB，默认值为 1 GiB。较大的测试文件可能更适合观察较长时间的传输表现，但会占用目标盘空间并增加读写量；低速 USB 存储或剩余空间有限的设备，可按设备情况选择较小文件和较少轮次。

开始测试前确认目标盘符和测试设置。CrystalDiskMark 会对选定存储执行读写操作；不要把结果目录、临时文件或重要数据误认为不会受写入影响。对 SSD 或 USB 闪存，官方提醒反复基准测试会增加写入并可能缩短介质寿命，建议只在需要时运行。

数值是手册列出的选项和默认设置，不是软件截图。低速设备可先用较小测试文件排查流程，再决定是否使用更大的工作集。

如何读懂 MB/s、IOPS 和延迟

顺序读写通常用于观察较连续数据的传输表现；随机读写则反映不同块大小与队列/线程设置下的访问情况。结果可能显示 MB/s 或 GB/s、IOPS 和平均延迟（µs）。这些指标各自描述吞吐量、每秒操作数和响应时间，不能只取一个数字来推断硬盘健康或所有程序的加载速度。

测试文件位置与大小、文件碎片、存储控制器、CPU 和数据模式都可能影响结果。对比两块 SSD 时尽量使用相同版本、同一模式与测试大小，并保持测试目标和运行条件一致；不同主版本的结果不可直接当作同一量尺。示意图中的数值为单位占位，不是性能成绩。

CrystalDiskMark 9.0.3 下载、安装与卸载

截至 2026-10-17，CrystalMark 官方下载页列出的版本是 9.0.3，发布日期为 2026-05-24。页面提供 Installer 与 ZIP，并另列 Microsoft Store 项目；商店当前显示 US$9.99。选择时要区分官网直接下载包和商店条目，实际价格与可用性应以所在地区的 Microsoft Store 为准。

1. 打开官方 CrystalDiskMark 下载页，核对版本 9.0.3 和文件类型；不要把榜单中的旧版转载链接当作官方入口。
2. 根据设备架构和系统版本选对应包。官网列出 x86、x64、ARM64；其兼容表特别注明 Installer 与 x64 版本不支持 Windows XP 和 Windows Server 2003。旧系统用户先检查兼容条件，避免盲目安装。
3. 使用 Installer 时运行安装包并按 Windows 提示完成；使用 ZIP 时将压缩包解压到自选目录，再从该目录启动程序。
4. 启动后选择测试次数、测试文件大小和目标磁盘，再运行全部项目或单项测试。开始前确认所选驱动器和剩余空间，避免选错磁盘。
5. 官方安装说明列出：Installer 可重新运行以更新，也可在 Windows“程序和功能”中卸载；ZIP 版可按官方说明覆盖更新或删除解压目录。更新前先保存重要工作。

本地没有下载、运行或扫描任何安装包；安装包大小、签名和哈希也未核验。官网 安装与卸载说明含 ZIP 与 Installer 的具体操作。

CrystalDiskMark 9.0.3 更新了什么

官方版本历史将 9.0.3 标为 2026-05-24，并列出一项更新：改进 DLL 加载过程的安全性（“Improved: Security of DLL loading process”）。本页只概述这条 9.0.3 记录，不把 9.0.2 或 9.0.1 的变更合并进最新版本，也不把“改进安全性”扩写成无法证明的安全保证。

官网历史页还列出 9.0.2（2026-02-17）更新语言文件并移除一项版权持有人记录，以及 9.0.1（2025-06-19）修复某些环境下基准测试无法正常运行的问题。若只想了解当前版更新，请以9.0.3 历史页为准。

常见问题（FAQ）：测速失败、结果波动与安全

为什么两次 SSD 测速差异很大？ 测试文件大小、存放位置、碎片、控制器、CPU、数据模式和运行状态都会影响结果。先统一 CrystalDiskMark 版本、模式和测试设置，并避免同时进行大量磁盘读写；不要把不同主版本的分数直接比较。

网络盘没有出现在列表里怎么办？ 官方 FAQ 提到，网络盘在以管理员身份运行时可能不可见；按官方建议尝试以普通用户权限启动。普通本地磁盘若出现权限错误，可检查目标位置能否创建测试文件，再按官方 FAQ 的权限建议处理。网络盘可见性和本地写入权限是不同问题，不要混为一个设置。

跑分会删除原有文件吗？ 本页没有把软件测试描述为文件恢复或清理操作。运行前确认测试目标和官方设置，重要资料先备份；不要对不确定的磁盘或目录盲目进行写入测试。

反复测试会影响 SSD 或 U 盘吗？ 官网提醒大量重复测试会增加写入，并可能缩短 SSD 或 USB 存储介质寿命。减少不必要的轮次，完成比较后停止即可；不应宣称测试“完全无损”。

下载前的建议

搜索 CrystalDiskMark 下载或 Windows SSD 测速时，请从官方页面确认版本、架构和安装方式。比较读写成绩前统一测试设置，并把基准测试当作性能观察工具，而不是硬盘健康结论。版本与更新记录以 CrystalMark 官方页面为准。

English

CrystalDiskMark tests Windows storage speed. Learn v9.0.3 changes, downloads, test settings, MB/s and IOPS results, plus checks for inconsistent SSD runs.

What CrystalDiskMark measures

CrystalDiskMark benchmarks read and write performance for HDDs, SSDs, USB flash drives, memory cards, and network drives. It reports sequential and random tests and offers Peak, Real World, and Demo modes. Use it to compare storage under matched conditions. It is not a SMART health monitor, bad-sector checker, or lifespan forecast; one score does not predict every application’s everyday performance.

This original diagram is conceptual and does not depict the actual CrystalDiskMark interface or measured data. See the official product page and main-window manual for supported devices and modes.

Choosing test count, file size, and target drive

The official manual lists 1–9 test runs, with 5 as the default. Test sizes range from 16 MiB to 64 GiB, and the default is 1 GiB. A larger test file may help observe longer transfers, but it uses drive space and adds read/write activity. For slower USB storage or a device with limited free space, choose a smaller file and fewer runs based on the device.

Before starting, confirm the target drive letter and test settings. CrystalDiskMark performs read/write operations on the selected storage; do not assume the test area or important data is unaffected by writes. The vendor warns that repeated benchmarks add writes and may shorten the life of SSDs or USB flash media, so run tests only when needed.

The figures summarize documented options and defaults; they are not a product screenshot. A smaller test file can help check the process on slower storage before choosing a larger workload.

Reading MB/s, IOPS, and latency

Sequential tests generally show transfers of more continuous data, while random tests measure access under different block sizes and queue/thread settings. Results can appear as MB/s or GB/s, IOPS, and average latency in microseconds. These describe throughput, operations per second, and response time; none alone indicates drive health or the loading speed of every application.

Test-file location and size, fragmentation, the storage controller, CPU, and data pattern can affect results. To compare SSDs, use the same CrystalDiskMark version, mode, and test size under similar conditions. Results from different major versions are not directly comparable. Any values in this illustration are unit placeholders, not benchmark scores.

Download, install, and remove CrystalDiskMark 9.0.3

As checked on October 17, 2026, the CrystalMark downloads page lists version 9.0.3 dated May 24, 2026. It offers an Installer and ZIP, and lists a separate Microsoft Store item; the Store page shows US$9.99. Treat the direct downloads and Store listing as separate channels, and check local Store pricing and availability for your region.

1. Open the official CrystalDiskMark downloads page and verify version 9.0.3 and the package type. Do not treat an older ranking-site mirror as the official source.
2. Match the package to your device architecture and Windows version. The site lists x86, x64, and ARM64; its compatibility table specifically says the Installer and x64 builds do not support Windows XP or Windows Server 2003. Check legacy-system support before installing.
3. For the Installer, run the package and follow Windows prompts. For the ZIP, extract the archive to a chosen folder and launch the program from there.
4. Choose a test count, file size, and target drive, then run all tests or an individual test. Confirm the drive and free space before starting.
5. The official installation guide says the Installer can be run again to update and can be removed through Windows Programs and Features. A ZIP build can be updated by following the vendor’s overwrite instructions or removed by deleting its extracted folder. Save important work before updating.

No package was downloaded, run, or scanned for this draft; file size, signature, and hash were not checked. See the vendor’s installation and removal guide for exact ZIP and Installer steps.

CrystalDiskMark 9.0.3 release notes and changes

The official version history dates 9.0.3 to May 24, 2026 and lists one change: “Improved: Security of DLL loading process.” This page summarizes only that 9.0.3 entry; it does not fold 9.0.2 or 9.0.1 changes into the current release or turn the security note into an unsupported guarantee.

The history page also lists 9.0.2 (February 17, 2026), which updated language files and removed a copyright-holder entry, and 9.0.1 (June 19, 2025), which fixed benchmark problems in some environments. For the current update, see the official 9.0.3 history.

FAQ: failed tests, inconsistent results, and safety

Why do SSD scores vary between runs? File size and location, fragmentation, the controller, CPU, data pattern, and system activity can affect results. Match the CrystalDiskMark version, mode, and settings, and avoid heavy concurrent disk activity. Do not compare scores across different major versions as if they used one scale.

Why is my network drive missing? The official FAQ notes that a network drive may not appear when the app runs as administrator; try launching it as a normal user. For a local-drive permission error, check whether the target location allows test-file creation and follow the official FAQ’s permission guidance. These are separate issues.

Will a benchmark delete my existing files? This guide does not describe the benchmark as a recovery or cleanup tool. Confirm the target and documented settings before running a write test, back up important data, and do not benchmark a drive or folder you cannot identify.

Can repeated tests affect an SSD or USB drive? The vendor warns that repeated benchmarks add writes and may shorten the life of SSD or USB storage. Avoid unnecessary runs and stop when the comparison is complete; the test should not be called “completely harmless.”

Before you download

For a CrystalDiskMark download or Windows SSD speed test, confirm the version, architecture, and package type on the official page. Keep settings consistent when comparing read/write results, and use a benchmark as a performance observation rather than a drive-health verdict. Follow CrystalMark’s pages for current release information.