Softstream Comparison Benchmark

February 25, 2025

Introduction‍

In today’s virtual world, high-speed streaming performance is a critical requirement for any remote desktop application. At Softdrive, our flagship product, Softstream, was purpose-built to deliver the fastest cloud PC experience in the world. Recognizing the frequent question, “How does your speed compare to [competitor]?”, we took the initiative to develop a benchmark to provide clear, measurable answers.

This test measures the end-to-end latency of a remote desktop application, specifically the round-trip time from initiating an action to its rendering on the local computer (eg. a right click on the desktop should open a context menu - how long does that take from click to seeing the image rendered on the client?).

This metric is valuable because it evaluates the perceived latency when using the application, providing a reproducible, comparable, and impactful measurement. The test results are directly comparable across different systems, unlike internal metrics from applications (e.g., decode time), which may not be inaccurate, but are often calculated differently and cannot be reliably compared between applications.

The end-to-end latency test is built as an external application which allows us to test any remote desktop software (and even test locally with no remote desktop). See our benchmarking results below:

End-To-End Test

⚠ EPILEPSY WARNING - The end-to-end latency test contains quickly flashing and bright colors ⚠

Softdrive aimed to establish an objective and foolproof method for measuring performance across remote desktop products. The goal is to derive data from a repeatable test that eliminates the need for human interpretation, external tools, or any factors that could introduce bias into the results.

How It Works

The client component of EndToEndLatencyTest injects input on local computer.
The remote desktop application captures and sends the input to remote computer.
The input is injected on the remote computer.
The server component of EndToEndLatencyTest switches color based on this input.
The remote desktop application captures the desktop changes and sends it back to the local computer.
The changes are rendered on the local computer.
The client component of EndToEndLatencyTest captures the desktop and calculates the latency between when the input was injected and when we received the new image.

The result is a fast, repeatable test that precisely measures the additional latency a product introduces, beyond baseline variables like refresh rate and ping.

The main components tested in the latency test are:

Encode

Remote desktop video encoding transforms screen content into efficient video streams for network transmission. The technology compresses visual data while maintaining sufficient quality for business use, and automatically adjusts to available bandwidth to ensure continuous operation. This enables reliable remote access across varying network conditions.

Decode

The process of converting encoded data back into its original or usable format for playback or display. In remote desktop applications, decoding involves converting the received encoded data into visual and input elements that the client device can render, impacting the application's performance and responsiveness.

Rendering

The process of generating a visual output from raw data or instructions, converting it into a format that can be displayed on a screen. In remote desktop applications, rendering involves creating the graphical representation of the desktop or application on the client device based on the received data.

Local & Remote Monitor Refresh Rate

The number of times per second a display updates its image, measured in hertz (Hz). For example, a refresh rate of 60Hz means the screen refreshes 60 times per second. Higher refresh rates result in smoother motion and reduced visual artifacts, especially in activities like gaming or video playback.

Desktop Duplication

A process that captures the contents of a computer's screen (or desktop) and makes it available for processing or transmission. It involves duplicating the graphical output of the display, often through a specialized API or feature like DXGI API , enabling applications to access real-time screen data without interfering with the original display. This is commonly used in remote desktop applications, screen sharing, or recording, ensuring accurate and efficient screen representation.

Input Injection

The process of programmatically sending user input (such as keystrokes, mouse movements, or touch events) to a system or application. It is commonly used in automation, testing, or remote desktop applications to simulate user interactions on a target device or interface. This allows seamless control and interaction as if the input were generated locally by the user.

Network Stack

The collection of protocols, technologies, and processes used to manage data transmission between the remote server (hosting the desktop) and the client device. It ensures efficient, secure, and reliable communication by handling tasks like data compression, encryption, packet routing, and retransmission in case of packet loss. Key components typically include transport protocols (e.g., TCP/UDP), session management, and mechanisms for handling latency, bandwidth fluctuations, and error correction, all optimized to deliver a seamless user experience.

Ping

Ping represents the round-trip time for data packets between a remote desktop client and server. This network latency measurement, typically expressed in milliseconds, indicates connection responsiveness and helps determine session quality. Lower ping values generally result in smoother remote desktop experiences. There is then a small amount of latency introduced by the remote desktop application itself, however in the case of high performance software like Softstream and Parsec it's usually less than 10% of the latency.