/MT[:n] enables multithreaded copying in Robocopy. The valid range is 1 to 128 threads, and writing /MT without a number uses 8 threads. More threads can help when a copy is limited by available bandwidth, but a higher number is not automatically faster. The dependable way to choose is to start conservatively, test a representative run, and adjust for the bottleneck in that particular source, destination, network and server.
What Robocopy’s /MT option changes
Robocopy normally processes copy operations sequentially. Adding /MT asks it to perform multiple file-copy operations concurrently. You can use the switch by itself or specify a count:
As an Amazon Associate I earn from qualifying purchases.
robocopy C:Source D:Destination /E /MT
robocopy C:Source D:Destination /E /MT:16The first command requests the documented default of eight threads. The second requests 16. The number is a concurrency setting: it controls how many copy operations Robocopy can have in flight, rather than changing the rules that decide which files are selected or which properties are copied. Selection, recursion, mirroring and property behavior remain the job of the other Robocopy switches in your command. Microsoft’s Robocopy command reference documents the syntax, limits and examples.
Range, default and meaning
| Syntax or switch | What it does | Documented value or condition |
|---|---|---|
/MT |
Enables multithreaded copying | Uses 8 threads when no number is supplied |
/MT:n |
Enables multithreaded copying with an explicit count | n can be any integer from 1 through 128 |
/E |
Copies subdirectories | Includes empty subdirectories; this is independent of the thread count |
/LOG:path |
Redirects Robocopy output to a log file | Microsoft recommends output redirection with /LOG for better performance |
/IPG |
Inter-packet gap control | Cannot be combined with /MT |
/EFSRAW |
Raw encrypted-file copying mode | Cannot be combined with /MT |
Although 128 is the upper limit, it is not a target. A count is useful only when the source, destination and system can keep those workers supplied with useful work. Microsoft cautions that saturating available bandwidth with more threads does not mean every migration will finish sooner.
#1 Best Overall
- Easily store and access 2TB to content on the go with the Seagate Portable Drive, a USB external hard drive
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Why more threads do not always make a copy faster
Every Robocopy run is constrained by one or more resources. Threads can expose unused capacity, but they cannot create network bandwidth, storage throughput or processor time. If the network link is already full, adding workers mostly increases contention. If the destination disk is busy, more simultaneous writes can make latency worse. If the machine is short of CPU capacity, the extra scheduling and file-system work can reduce useful throughput.
- Network bandwidth: A large initial transfer over a fast link may need concurrency to keep the link busy. A saturated link leaves no bandwidth for additional threads to exploit.
- Processor capacity: Thread management, file enumeration and file-system activity consume CPU. Later runs may need a count closer to the available processor cores and hardware threads, while reserving capacity for other production work.
- Storage I/O: Source and destination disks, storage arrays and network-backed shares can become the limiting resource. More outstanding operations can increase queueing instead of reducing elapsed time.
- File mix: A few large files behave differently from millions of small files. Small files can spend proportionally more time in namespace processing and metadata operations, so simply increasing the count may not produce a proportional gain.
- Competing workloads: Backups, databases, virtual machines, users and other transfers share the same CPU, network and storage. A setting that is fastest in an idle test can be harmful during production hours.
These constraints are why there is no universal “best” value. Treat the thread count as a workload parameter, not as a quality setting that should always be pushed to the maximum.
Choose a count by identifying the kind of Robocopy run
Initial or bulk copy
An initial copy moves a large amount of data and often contains many files that have never reached the destination. Its dominant work is usually data transfer, so enough concurrency to use available network bandwidth can help. Start with the default of 8 unless you already know the environment is bandwidth-limited and has spare CPU and storage capacity. Then compare a modestly higher value under the same conditions.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →For Azure Files migrations, Microsoft reports balanced initial-copy performance between 8 and 20 threads in its tests. The same guidance reports good performance up to 64 threads when the processors can keep those threads active. Those figures describe Microsoft’s Azure Files migration conditions; they are not a general optimum for local disks, another file share or every Windows server. See Microsoft’s Azure Files migration guidance for that context.
Differential or mirror follow-up
A later pass may copy only changed files. When relatively little data changes, the run can spend more of its time enumerating directories, comparing metadata and processing the namespace than moving bytes. Microsoft advises matching thread counts for subsequent runs more closely to processor core count and thread count per core, while accounting for cores required by other production work. Do not assume that the count that won the initial bulk transfer will also be best for every later pass.
Rank #2
- Easily store and access 5TB of content on the go with the Seagate portable drive, a USB external hard Drive
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Many small files
Small-file workloads can be namespace-bound. Increasing concurrency may help when the source and destination can process more requests, but it can also increase metadata contention and storage activity. Benchmark with a file set whose size distribution and directory structure resemble the real job. A test made only from a handful of large files will not predict a small-file migration reliably.
Large files over a fast connection
Large files on a fast network are a case where additional workers may help fill available bandwidth, particularly when one operation does not keep the path busy. Raise the count in measured steps and stop when throughput stops improving, CPU or storage load becomes excessive, or other work is affected.
Outdated 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 matchPC 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 & 11A practical method for selecting the thread count
The following process turns the documented range into a decision you can repeat. It is a practical benchmarking method derived from Microsoft’s cautions about bandwidth, compute and I/O; it is not a Microsoft-prescribed test protocol.
- Define the run you are tuning. Record whether it is an initial copy, a differential pass or a mirror-style update. Note whether the job is dominated by large files, small files or a mixture. A later run with only a few changed files should be tuned separately from the first bulk transfer.
- Record the environment. Identify the source and destination, the network path, the storage type and the other work sharing the server. Keep these conditions visible when comparing values. A result from an idle maintenance window is not directly comparable with a result while backups or database activity are running.
- Start at 8. Eight is the documented default when you write
/MTwithout a number. It is a conservative baseline that avoids guessing at a value before you have observed the workload. - Select a few plausible alternatives. If the baseline appears bandwidth-limited and the machine has spare capacity, test modestly higher values such as 16 or another value appropriate to your environment. Do not jump straight to 128 merely because it is allowed. The useful comparison is between values your hardware and workload can plausibly sustain.
- Use a representative sample or run. Keep source, destination, directory structure, file sizes, network path and competing workloads as consistent as possible. A tiny sample can hide directory-enumeration costs; an unrelated file set can exaggerate or conceal the benefit of concurrency.
- Measure the outcome. Record elapsed time and observed throughput, then watch CPU availability, storage I/O and network utilization. Also note whether the copy disrupts other work. A count that produces a higher peak rate but a longer elapsed run, or that starves a production service, is not an improvement.
- Keep the best workload-specific value. Choose the setting that improves the actual job without exhausting a shared resource. If two values finish within normal run-to-run variation, prefer the lower one when it leaves more capacity for other services.
- Re-test after material changes. A different file mix, storage target, network path, server load or run type can change the bottleneck. Reuse the number as a starting point, not as a permanent promise.
How to run a controlled comparison
Robocopy’s own summary and a retained log give you a record of each trial. Microsoft’s command reference recommends redirecting output with /LOG for better performance and demonstrates logging in its examples. A simple test command is:
robocopy C:Source D:Destination /E /MT:16 /LOG:C:Logscopy.logChange the source, destination and log path to match your environment. In this example, /E includes subdirectories, including empty ones; it does not determine the thread count. Run the same command structure with the other candidate counts, using separate log files or otherwise preserving each result.
Rank #3
- Easily store and access 1TB to content on the go with the Seagate Portable Drive, a USB external hard drive.Specific uses: Personal
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop. Reformatting may be required for Mac
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Keep the trials comparable
- Use the same source and destination and avoid changing the file set between trials.
- Keep the copy mode and non-threading switches identical. Changing selection or retry behavior at the same time makes the result impossible to attribute to
/MT. - Run trials under similar background load. Record whether other transfers, backups or applications were active.
- Compare elapsed time as well as instantaneous throughput. A short burst can look impressive while the complete run remains slower.
- Check CPU, network and storage load while the copy is active. The limiting resource explains why one value wins and helps you avoid overloading a shared server.
When a run contains a mix of changed and unchanged files, compare like with like. A first pass that copies terabytes is measuring a different workload from a later pass that mostly enumerates directories and updates a small number of files.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteInterpreting common results
Throughput rises as you add threads
If elapsed time falls and the network or storage path still has headroom, a higher count is using capacity that was previously idle. Continue only until the improvement becomes negligible or another resource approaches its safe limit. The first value that meets the operational goal is usually more useful than the largest value that happens to improve a lab run.
Throughput stops improving
A flat result usually means another resource has become the bottleneck or that the workload is dominated by per-file processing. More threads cannot overcome a full network link, slow storage or insufficient CPU. Keep the lower count if it gives the same elapsed time with less contention.
Throughput falls when the count increases
A decline indicates that added concurrency is creating contention or consuming capacity needed elsewhere. Reduce the count and check storage latency, CPU availability, network saturation and competing workloads. For many small files, namespace and metadata work can dominate even when the link itself is fast.
The initial copy is fast, but later passes are disappointing
This difference is expected when the workload changes. The initial pass spends more time transferring data; a later differential pass may spend more time scanning and evaluating the namespace. Re-test the later pass independently and choose a count that fits the processor capacity available after other production work is accounted for.
Rank #4
- Easily store and access 4TB of content on the go with the Seagate Portable Drive, a USB external hard drive.Specific uses: Personal
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Azure Files figures: useful evidence, narrow scope
| Microsoft guidance | What the figure describes | How to use it |
|---|---|---|
| 8–20 threads | Balanced initial-copy performance in Microsoft’s Azure Files migration tests | A reasonable Azure Files test range for an initial migration, not a universal recommendation |
| Up to 64 threads | Good performance in those tests when processors could keep the threads active | A conditional ceiling observed in that Azure Files context; validate CPU capacity before trying it |
The page does not establish a publication year for those test figures. They should therefore be read as test observations tied to the Azure Files migration scenario, not as a dated industry benchmark or a promise for every Robocopy destination. Microsoft’s guidance also emphasizes that high thread counts can saturate available bandwidth without making every migration faster.
Compatibility and logging details
Incompatible switches
Do not combine /MT with /IPG or /EFSRAW. If a command requires either of those modes, remove /MT and use the mode’s documented behavior instead. These are compatibility restrictions, not tuning advice; changing the thread count will not make an invalid combination work.
Why use /LOG
Microsoft’s command reference says, “For better performance, redirect your output using /log option.” Logging also gives you a durable record of the command’s summary and any errors, which makes side-by-side trials and later troubleshooting practical. Put the log on a location with sufficient space and permissions, and change the example path to one appropriate for your system.
What /MT does not control
The switch does not select files, include or exclude directories, mirror deletions, preserve properties or choose retry behavior. Those outcomes come from the rest of the Robocopy command. For example, /E controls recursive inclusion of subdirectories, including empty ones, while /MT:16 controls copy concurrency. Keeping these roles separate prevents a thread-count change from being mistaken for a change in copy semantics.
Free tools Windows power users keep installed
One-click scans. No signup required.
Practical starting points
- Ordinary copy with no prior measurements: use
/MT(8 threads), retain a log, and measure the complete run. - Bandwidth-limited initial transfer with spare server capacity: compare 8 with a modestly higher count, such as 16, under consistent conditions.
- Azure Files initial migration: test within the 8–20 range reported by Microsoft, then consider higher values only if processors can keep the threads active and your own measurements justify them.
- Repeated differential or mirror passes: tune separately and keep the count closer to the processor capacity available after reserving resources for other work.
- Small-file or shared-server workload: prioritize elapsed time, namespace behavior and impact on other services over a high peak throughput number.
These are starting points, not guarantees. The deciding evidence is how the complete workload behaves on its actual source, destination and network path.
Best Value
- [Upgraded Version] - This external hard drive features a mirrored logo stripe combined with a striped anti-slip design, and the rounded corners of the casing make it easier to grip. The stripes also have a heat dissipation function, ensuring stable and fast data transfer.
- 【Ultra-thin and quiet】 - The motherboard adopts JMicron 578 noise-free solution, giving you a quiet working environment. Lightweight and portable size designed to fit in your pocket for easy portability.
- 【Ultra-Fast Data Transfers】 - Pairing this external hard drive with JMicron 578 solution USB 3.0 and USB 2.0 interfaces enables blazing-fast data transfer. It boasts theoretical read speeds of up to 125MB/s and write speeds of up to 103MB/s.
- 【Plug and Play】 - With no software to install, just plug it in and the drive is ready to use.The hard disk chip is wrapped with an aluminum anti-interference layer to increase heat dissipation and protect data.
- 【What You Get】 - 1 x Portable Hard Drive, 1 x USB 3.0 Cable, 1 x User Manual, Gift-type shell packaging ,Three-year manufacturer's warranty and free technical support services.
Troubleshooting a problematic thread count
The command is rejected or does not run as expected
Check the number first: /MT:n accepts integers from 1 through 128. Then inspect the command for /IPG or /EFSRAW, both of which are incompatible with /MT. Confirm that the source, destination and log paths are valid and writable before changing performance settings.
The copy is fast but affects other services
Reduce the count and repeat the comparison during a representative production window. A lower value that finishes nearly as quickly while leaving CPU, network and storage capacity available is the safer operational choice.
CPU is busy but the network is not full
More threads are unlikely to solve a processor-bound run. Test a lower count and examine whether file enumeration, metadata work or other applications are consuming the available CPU. For subsequent runs, align the count more closely with processor cores and threads per core, as Microsoft advises, while preserving capacity for essential services.
Recommended Free Tools
The network is full but the copy still does not improve
Once the path is saturated, additional threads cannot add bandwidth. Keep the lowest count that reaches the same sustained throughput and avoids unnecessary storage or CPU contention.
Results vary between runs
Look for changing file sets, background workloads, cache state or storage activity. Repeat the comparison with the same source, destination and conditions, and judge by complete elapsed time rather than one short throughput sample. If the values are effectively tied, choose the more conservative count.
A concise decision rule
Use 8 as the baseline because it is Robocopy’s documented default. Increase the count only when a representative run shows that bandwidth or another parallel resource is underused and the server has capacity to spare. Stop increasing when elapsed time no longer improves, when CPU or storage contention appears, or when other work is harmed. Treat initial transfers and later differential passes as separate workloads, and regard Microsoft’s Azure Files figures—8–20 for balanced initial performance and up to 64 when processors can sustain the threads—as scenario-specific evidence rather than a universal setting.
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.
