Benchmark

TASTEFTP vs FileZilla
SFTP transfer speed benchmark

On 2026-10-30, GXSOFT timed SFTP transfers with the same PC, the same server and the same files. Values are the median of 5 runs; lower is faster.

More about TASTEFTP Download

Results

TASTEFTP and FileZilla transfer times (seconds, median)
Transfer	TASTEFTP	FileZilla	Result
Download1 GiB × 1 file	13.73 s74.6 MiB/s	23.87 s42.9 MiB/s	about 1.7× faster
Download64 KiB × 1,000 files	26.71 s	47.57 s	about 1.8× faster
Upload1 GiB × 1 file	14.68 s69.8 MiB/s	16.77 s61.1 MiB/s	No clear difference
Upload64 KiB × 1,000 files	11.17 s	34.73 s	about 3.1× faster

Setup: TASTEFTP 0.2.50 (command-line batch transfer) and FileZilla 3.71.1 (transfer queue), SFTP, 2 concurrent transfers, 1 Gbps wired LAN, incompressible synthetic files. Every received file was checked with SHA-256.

Test environment

Client PC: AMD Ryzen 7 8845HS (16 threads) · 31 GiB RAM · Windows 11 Pro (26300) · Realtek USB Ethernet 1 Gbps
Server: AMD Ryzen 5 5600G (12 threads) · 58 GiB RAM · Ubuntu 26.04.1 LTS (Linux 7.0) · OpenSSH 10.2p1 · storage: ext4 image on a SATA SSD
Network: Wired LAN on the same subnet, 1 Gbps on both ends, MTU 1500
Round trip to the server: ping × 50 (same day): average 7 ms, min 1 ms, max 46 ms
Apps compared: TASTEFTP 0.2.50 for Windows (command-line batch transfer, same transfer engine as the GUI) · FileZilla 3.71.1 for Windows (GUI transfer queue)
Files: Incompressible synthetic files from /dev/urandom: 1 × 1 GiB and 1,000 × 64 KiB (62.5 MiB in total)
Transfer settings: SFTP, the same ed25519 key, 2 concurrent transfers (FileZilla's default), speed limits and preallocation off

Method

For each file set, each app got one warm-up run and then 5 measured runs, alternating which app went first. Tables show the median of the 5.
The clock starts when the transfer command is sent. The command-line side includes process start and login; the GUI app starts when the 'Process queue' menu command is sent to its already open window, so its launch time is excluded.
The clock stops when every target file reaches its full size, watched the same way for both apps: every 20 ms on the PC for downloads and every 50 ms on the server for uploads.
Every received file was compared with the original by SHA-256.

Limits and what we did not measure

These numbers come from the maker of the product (GXSOFT). The setup and raw data are public so you can re-run them.
One PC, one server and one LAN. Results may differ with high-latency servers across the internet.
FTP, FTPS and other SFTP apps (WinSCP, Cyberduck and others) were not measured.
Runs vary, so rows whose run ranges overlap are not called for either side.

Raw data and re-running

Per-run raw data (time, hash checks), the scripts and how to re-run them are in the public repository.

See the setup, raw results and scripts FileZilla alternative page

Questions

Which is faster, TASTEFTP or FileZilla?: Under these conditions (same PC, same LAN server, SFTP, 2 concurrent transfers), download of 1 GiB × 1 file was about 1.7× faster, download of 64 KiB × 1,000 files was about 1.8× faster, upload of 64 KiB × 1,000 files was about 3.1× faster, upload of 1 GiB × 1 file showed no clear difference. Results may differ on other networks and servers.
Who measured it, and how?: GXSOFT, the maker of TASTEFTP, measured it. Both apps moved the same synthetic files to and from the same server, 5 alternating runs each, and the raw data and scripts are in a public repository.

FileZilla and related names are trademarks of their respective owners. TASTEDEV is not affiliated with or endorsed by them.

Results

TASTEFTP and FileZilla transfer times (seconds, median)
Transfer	TASTEFTP	FileZilla	Result
Download1 GiB × 1 file	13.73 s74.6 MiB/s	23.87 s42.9 MiB/s	about 1.7× faster
Download64 KiB × 1,000 files	26.71 s	47.57 s	about 1.8× faster
Upload1 GiB × 1 file	14.68 s69.8 MiB/s	16.77 s61.1 MiB/s	No clear difference
Upload64 KiB × 1,000 files	11.17 s	34.73 s	about 3.1× faster

Test environment

Client PC

AMD Ryzen 7 8845HS (16 threads) · 31 GiB RAM · Windows 11 Pro (26300) · Realtek USB Ethernet 1 Gbps

Server

AMD Ryzen 5 5600G (12 threads) · 58 GiB RAM · Ubuntu 26.04.1 LTS (Linux 7.0) · OpenSSH 10.2p1 · storage: ext4 image on a SATA SSD

Network

Wired LAN on the same subnet, 1 Gbps on both ends, MTU 1500

Round trip to the server

ping × 50 (same day): average 7 ms, min 1 ms, max 46 ms

Apps compared

TASTEFTP 0.2.50 for Windows (command-line batch transfer, same transfer engine as the GUI) · FileZilla 3.71.1 for Windows (GUI transfer queue)

Files

Incompressible synthetic files from /dev/urandom: 1 × 1 GiB and 1,000 × 64 KiB (62.5 MiB in total)

Transfer settings

SFTP, the same ed25519 key, 2 concurrent transfers (FileZilla's default), speed limits and preallocation off

Method

For each file set, each app got one warm-up run and then 5 measured runs, alternating which app went first. Tables show the median of the 5.

The clock starts when the transfer command is sent. The command-line side includes process start and login; the GUI app starts when the 'Process queue' menu command is sent to its already open window, so its launch time is excluded.

The clock stops when every target file reaches its full size, watched the same way for both apps: every 20 ms on the PC for downloads and every 50 ms on the server for uploads.

Every received file was compared with the original by SHA-256.

Limits and what we did not measure

These numbers come from the maker of the product (GXSOFT). The setup and raw data are public so you can re-run them.

One PC, one server and one LAN. Results may differ with high-latency servers across the internet.

FTP, FTPS and other SFTP apps (WinSCP, Cyberduck and others) were not measured.

Runs vary, so rows whose run ranges overlap are not called for either side.

Questions

Which is faster, TASTEFTP or FileZilla?

Under these conditions (same PC, same LAN server, SFTP, 2 concurrent transfers), download of 1 GiB × 1 file was about 1.7× faster, download of 64 KiB × 1,000 files was about 1.8× faster, upload of 64 KiB × 1,000 files was about 3.1× faster, upload of 1 GiB × 1 file showed no clear difference. Results may differ on other networks and servers.

Who measured it, and how?

GXSOFT, the maker of TASTEFTP, measured it. Both apps moved the same synthetic files to and from the same server, 5 alternating runs each, and the raw data and scripts are in a public repository.