NAS Throughput Calculator: From GbE to MB/s

The quick answer

To convert a NAS network speed from gigabits per second to megabytes per second:

Theoretical MB/s = Link speed (Gbps) × 1,000 / 8

Or, more simply:

Bandwidth (Gbps) / 8 = theoretical GB/s

Bandwidth (Gbps) × 125 = theoretical MB/s

The most common conversions are:

Network link	Theoretical bandwidth	Practical meaning
100 Mbps	12.5 MB/s	Suitable for light file access and some backups
1 GbE	125 MB/s	Commonly limits transfers to roughly the low hundreds of MB/s
2.5 GbE	312.5 MB/s	Useful for faster single-client transfers
5 GbE	625 MB/s	Provides more headroom for fast storage
10 GbE	1,250 MB/s	Can support high-speed SSD or multi-drive workloads
25 GbE	3,125 MB/s	Usually requires server-class networking and storage

These are link-rate ceilings, not guaranteed file-copy speeds. Ethernet, IP, TCP or SMB, storage protocols, filesystem work, and client operating-system overhead all reduce the result. A 1 GbE connection therefore does not copy files at exactly 125 MB/s.

How to calculate NAS throughput

Step 1: Convert bits to bytes

Network equipment normally advertises speed in bits per second. File-copy tools usually report bytes per second.

Since:

1 byte = 8 bits

The conversion is:

1 Gbps = 1,000 Mbps

1,000 Mbps / 8 = 125 MB/s

This uses decimal units, which are standard for network link speeds and MB/s calculations.

Step 2: Account for protocol and hardware overhead

The theoretical result does not include all the data required to move a file. Actual transfers include overhead from several layers:

• Ethernet framing
• IP and TCP or UDP headers
• SMB, NFS, or another file protocol
• TLS or encryption, where used
• Filesystem metadata and checksums
• NAS and client processing
• Switch and adapter behavior

A useful planning formula is:

Expected MB/s = Theoretical MB/s × measured efficiency

The efficiency factor is not universal. It depends on packet size, protocol, file size, CPU load, storage speed, and configuration. For a rough illustration, if a 10 GbE path achieved 90% of its link rate:

1,250 MB/s × 0.90 = 1,125 MB/s

That is an example calculation, not a guaranteed 10 GbE result. For purchasing decisions, use measured results from the actual NAS, client, switch, protocol, and storage configuration whenever possible.

MB/s versus MiB/s

Network calculations generally use decimal MB/s:

• 1 MB = 1,000,000 bytes
• 1 MiB = 1,048,576 bytes

Some operating systems and benchmark tools report MiB/s while labeling the result as MB/s. The number can therefore appear slightly lower even when the transfer is identical. Check the tool's units before comparing results.

The slowest component sets the result

A NAS transfer has several ceilings. A useful simplified model is:

Actual throughput = lowest of network capacity, NAS storage capacity, client storage capacity, protocol capacity, and processing capacity

This is not a complete performance model, but it helps identify bottlenecks.

Network bottlenecks

The network path includes more than the NAS port:

• NAS network interface
• Client network interface
• Ethernet cable and link negotiation
• Switch ports and backplane capacity
• Router or firewall, if the traffic crosses one
• VLAN, virtual-switch, or bonded-interface configuration

A 10 GbE NAS connected through a 1 GbE switch port still has a 1 GbE path to that client. Similarly, a client with a 2.5 GbE adapter cannot use the full bandwidth of a 10 GbE NAS connection.

Check the negotiated link rate at both endpoints. A connection can fall back to a lower speed because of cabling, adapter settings, a damaged cable, or an incompatible transceiver.

Drive and RAID bottlenecks

The NAS must read or write the data quickly enough to fill the network connection.

A hard-drive array may be fast enough for 1 GbE but become the limiting factor at 2.5 GbE or 10 GbE. SSD storage can provide more bandwidth, but the result still depends on the SSDs, RAID layout, controller, filesystem, workload, and sustained-write behavior.

RAID also introduces trade-offs:

• RAID 1 mirrors data and prioritizes redundancy over usable capacity.
• RAID 5 uses parity and can provide efficient capacity, but writes involve parity work and a degraded array has reduced protection.
• RAID 6 adds another parity block and is often chosen for larger arrays, with additional write and rebuild considerations.
• RAID 10 mirrors and stripes data, often favoring write and random-I/O behavior at the cost of roughly half the raw capacity.

RAID is not a backup. It can keep a NAS online after certain drive failures, but it does not protect against accidental deletion, malware, filesystem damage, theft, fire, or a failed NAS. Maintain separate backups regardless of the throughput target.

Client bottlenecks

The computer receiving or sending the file can be the limit. Common causes include:

• A slower client SSD or hard drive
• A nearly full drive
• CPU overhead from encryption or compression
• Antivirus or endpoint-security scanning
• USB network adapters with limited performance
• Wi-Fi instead of wired Ethernet
• A virtual machine or container network path
• Many small files rather than one large sequential file

A transfer of one large video file is usually easier to sustain than a directory containing thousands of small documents. Small files require more metadata operations and network round trips, so the displayed MB/s can be much lower without indicating a fault.

NAS CPU and software bottlenecks

The NAS processor may limit encrypted SMB, checksumming, compression, deduplication, virtualization, or media services. Plex transcoding and surveillance recording can also consume CPU, storage, and network resources while a file transfer is running.

ZFS adds data-integrity features such as checksumming and copy-on-write behavior, but performance depends on memory, vdev layout, record size, workload, and the underlying drives. ZFS does not automatically make a network link faster, and adding an SSD cache does not guarantee higher throughput.

SSD cache is most useful when the workload benefits from repeated reads or random I/O and the cache is sized and configured appropriately. It may provide little benefit for a single large sequential transfer that the main storage can already serve efficiently. Cache devices also add cost, power use, and another component that needs appropriate endurance and protection.

Single-client versus aggregate throughput

A NAS port's speed is not always the same as the speed available to one user.

Single-client throughput

Single-client throughput is the transfer rate between one NAS and one client. It is usually limited by the slower endpoint or the slowest part of that client's path.

For example, a NAS with 10 GbE does not make a laptop with 1 GbE transfer at 10 GbE speeds. The laptop's connection remains the ceiling:

1 GbE theoretical ceiling = 125 MB/s

Aggregate throughput

Aggregate throughput is the combined traffic from multiple clients, services, or network interfaces.

A NAS might serve:

• One client copying project files
• A second client running a backup
• Plex reading media
• Cameras writing surveillance footage
• Cloud synchronization or replication

Multiple 1 GbE clients can collectively place more load on the NAS than one client, but the total is still limited by the NAS interface, switch uplink, storage subsystem, CPU, and configuration. Link aggregation can distribute traffic across clients, but it does not normally turn one individual connection into a faster connection. Support and behavior depend on the NAS, switch, protocol, and aggregation mode.

Worked example: choosing between 1 GbE and 2.5 GbE

Suppose a NAS and desktop both support 2.5 GbE, and the switch provides a compatible 2.5 GbE path.

Network ceiling

2.5 Gbps × 125 = 312.5 MB/s theoretical

Now assume testing or prior measurements show that the NAS storage can sustain 220 MB/s for the intended large-file workload, while the desktop storage can sustain more than that.

The likely transfer result is capped by the NAS storage:

Actual throughput ≈ lowest of 312.5 MB/s, 220 MB/s, and client storage speed

If the client storage can sustain 500 MB/s or more:

Actual throughput ≈ 220 MB/s before protocol and workload effects

In this case, upgrading from 1 GbE to 2.5 GbE can remove the network bottleneck, but it cannot make the NAS array exceed its own sustained 220 MB/s storage limit.

If the same NAS instead serves data at only 90 MB/s for the workload, 1 GbE may already be sufficient:

• 1 GbE theoretical ceiling: 125 MB/s
• NAS storage result: 90 MB/s
• 2.5 GbE theoretical ceiling: 312.5 MB/s

The additional network capacity would remain unused unless the storage workload changes.

A practical throughput calculator

Use this sequence when estimating a transfer:

1. Convert the link rate

Link speed in Gbps × 125 = theoretical MB/s

1. Find the slowest network endpoint

Compare the NAS port, client port, switch ports, and the complete path.

1. Check storage throughput

Consider the NAS read or write result for the actual RAID, filesystem, drive type, and workload.

1. Check the client disk

The client must sustain the target write or read speed.

1. Allow for overhead

Do not treat the theoretical link rate as the expected file-copy rate.

1. Account for concurrent activity

Include backups, surveillance recording, Plex, synchronization, rebuilds, scrubs, and other users.

1. Test with representative data

Test both a large file and a realistic small-file workload. Measure reads and writes separately.

A compact formula is:

Estimated file-copy MB/s = minimum of (network ceiling, NAS read/write rate, client read/write rate) × protocol/workload efficiency

The exact efficiency cannot be determined from the Ethernet rating alone.

Network-fit checklist

Use this checklist before paying for a faster NAS port or network upgrade:

• [ ] Convert the advertised link speed from Gbps to theoretical MB/s.
• [ ] Confirm the NAS and client both support the target link speed.
• [ ] Confirm every switch port and uplink in the path supports it.
• [ ] Verify the cable or transceiver is appropriate for the selected link.
• [ ] Check the negotiated link rate rather than relying on product labels.
• [ ] Determine whether one client or many clients need the bandwidth.
• [ ] Measure the NAS array's sequential read and write performance.
• [ ] Check the client drive's sustained read and write capability.
• [ ] Consider RAID level, filesystem, ZFS settings, and encryption overhead.
• [ ] Include Plex, surveillance, backups, synchronization, and rebuild activity.
• [ ] Treat SSD cache as workload-dependent, not as an automatic speed upgrade.
• [ ] Remember that RAID improves availability in specific drive-failure scenarios but is not a backup.
• [ ] Plan power, cooling, and expansion for faster networking and additional drives.
• [ ] Test with both large files and many small files before finalizing the design.

If you are comparing hardware for a specific workload, Browse NAS and narrow the choices by storage, networking, and expansion needs. You can also review NAS and storage servers when a faster network or larger drive set changes the type of system you need.