v1.0.8

KevqDMA — What's New

Current Version: v1.0.8
Summary Updated: October 13, 2026

Highlights

Fixed and optimized WebRadar, fixed discarded items. Improved stability. Corrections have also been made to the Target menu.

v1.0.7

KevqDMA — What's New

Current Version: v1.0.7
Summary Updated: September 27, 2026

Highlights

• Added support for Rush mode: general overview, individual zones, and corresponding coordinates for Radar and WebRadar.

• Added automatic update checking from GitHub.
• Fixed communication issues with MAKCU and KMBox, activation handling, and the command queue.
• Added Ferrum One adapter via Ferrum App / NET.
• Improved stability of memory reading, background tasks, settings persistence, and WebRadar data streaming.
• Expanded ESP customization options and updated map handling.
• Assets compressed losslessly.

1. Rush, Maps, and Coordinates

• Added general overview for rush_001 map and 20 distinct Rush areas.
• Area selection factors in player position; each area uses its own dedicated coordinates and scale.
• Fixed standard Radar marker projection for Rush areas and corrected marker rotation orientation.
• Map images and coordinates are now updated together from a single pinned source version — avoiding mismatches between new images and outdated layout bounds.
• Removed duplicate obsolete coordinate definitions.
• Added maps: Shelter, Boulder, Debris, El Dorado, Fachwerk, and Poseidon.
• Added extra floors/levels for Nuke, Train, Vertigo, Baggage, and Boulder. Layer selection accounts for elevation with a margin at layer boundaries to prevent rapid flickering from minor position jitter.
• The WebRadar manual map selection list is now generated from the shared registry rather than maintained separately.

2. GitHub Update Checker

• Added release check for the repository cs2-dma/CS2-DMA.
• Current version is compared against published releases by numerical components; update prompt appears when a newer version is available.
• Utilizes the GitHub API with a fallback path via the release page. Requests are executed in parallel with timeout limits and cancellation of redundant work.
• GitHub unavailability will not block the application from launching.
• When an update is detected, a y/n prompt is displayed, where Enter accepts by default.
• If accepted, the release page opens; if a matching ZIP archive is found, it prompts download via the browser. Once opened successfully, delayed application shutdown is initiated.
• If declined, the application proceeds with normal startup.
• Added link validation and graceful handling for malformed responses, timeouts, and GitHub rate limits.

3. MAKCU, KMBox, and Ferrum One

MAKCU

• Refined device and COM port detection, connection routine, and button state stream processing.
• Resolved integration issues between Target activation, the device, and available input sources.
• Clarified states for connection status, movement readiness, and activation key availability: a successful test circle is no longer treated as conclusive proof that the entire input pipeline is operational.

KMBox

• Support for both KMBox Serial and KMBox Network.
• Improved button state polling, Hold/Toggle handling, and interaction with the shared input queue.
• Added ACK waiting and limited timeout retries for network button release. Press actions are intentionally not automatically retried by the transport layer.
• Stale commands are discarded, and test movement cancellation is respected across execution steps and connection changes.

Ferrum One

• Added dedicated Ferrum One (App / NET) option.
• Implemented connection via the compatible Ferrum App network interface.
• Utilizes shared mechanisms for the movement queue, activation, cancellation, and command freshness validation.
• Note: This provides App/NET interface support and does not imply universal support for arbitrary firmware modes or direct COM connections on Ferrum.

4. Target & Action Handling

• Reworked Hold/Toggle activation handling, movement queue, and cancellation of accumulated commands.
• Basic aiming assistance is decoupled from World Geometry readiness; features that require geometry now independently check its availability.
• Added data freshness validation before dispatching clicks, accounting for queue wait time.
• Eliminated unconditional target lockouts following an unconfirmed click until target switch. One bounded retry is permitted if checks pass; infinite retries are prevented.
• Disabling activation clears the corresponding lock state.
• Target health drop alone is no longer treated as definitive proof of our shot hitting.
• Hitchance threshold is never covertly lowered: values below the configured threshold strictly prevent action execution.
• Expanded automated sanity checks for Target and input device interoperability.

5. ESP & Interface

• Expanded health and armor bar settings: side positioning, width, and numerical display mode.
• Added a brief lost health trail, with the main value updating independently from the fading segment.
• Added independent armor placement and compact display mode.
• Fixed HP/Armor number layout issues, including cramped spacing and clipping.
• Expanded Player Flags settings: placement, style, order, and maximum count limit.
• Added a dedicated visual state for "no reliable visibility data" along with customizable element recoloring.
• Expanded Snap Lines configuration: start/end points, thickness, opacity, distance limit, line count, and nearest player indicator.
• Verified menu layouts and preview rendering across various window sizes and DPI scaling levels.

6. Memory Reading & Background Stability

• Improved read queue scheduling: auxiliary data, inventory, and bones now factor in wait time so that a perpetually busy stage cannot starve other tasks indefinitely.
• Strengthened data validity checks to ensure data belongs to the current scene and player.
• For WebRadar local position: validity, player identity, and timestamp of measurement are strictly tracked. Fresh data can supersede stale sources, and missing data is never spoofed with a false zero coordinate.
• Coordinate sampling timestamps are decoupled from transmission times.
• Fixed startup and shutdown states for background threads.
• Offset downloads now enforce timeouts, response size limits, and atomic file replacement checks.
• Diagnostics accurately distinguish between menu idle, data degradation, and genuine read failures. Historical error counters are preserved rather than reset for "cleaner" metrics.

7. Profiles & WebRadar

• Fixed deferred save bug: settings from one profile will no longer overwrite another during rapid switching. Save jobs now capture their own configuration snapshot.
• Streamlined profile serialization and loading, including remote radar configurations.
• Added strict type, range, and string length validation when parsing WebRadar configs.
• Added WebSocket error handling, message size limits, and malformed connection URL checks.
• Added server SSH host key verification prior to password transmission. Unrecognized host keys require confirmation, and unexpected key changes block the connection.
• Fixed Node process startup alignment with systemd constraints.
• Fixed embedded asset naming and path casing when deploying to Linux, covering maps, fonts, and localization files.

8. Asset Optimization & Project Cleanup

• Removed verified dead code paths and redundant legacy algorithms.
• Deprecated settings with no functional impact are excluded from persistence; backward migration is preserved for legacy profiles.
• Existing graphical assets reduced from 6.10 to 3.16 MB — saving 2.94 MB with zero pixel alteration.
• Weapon icon atlas: 2.80 MB → 94 KB, verified with byte-exact RGBA matching after decoding and texture upload.
• Final executable size is 18.79 MB: newly added maps and floor levels added 3.47 MB of image assets, resulting in a slightly larger total build size despite optimization of existing assets.

v1.0.6

{ ... } Direct Memory Access — CS2 DMA

An open-source software framework designed for the research and implementation of Direct Memory Access (DMA) interaction with Counter-Strike 2. The system demonstrates established methodologies for memory reading, game entity parsing, network synchronization, and data visualization. It serves as a practical platform for studying low-level programming, PCIe architecture, and reverse engineering techniques.

{ ! } This project is for educational and research purposes only
{ ! } This is an external method and requires a DMA Card to work.
{ ! } WebSite: https://cs2-dma.github.io/CS2-DMA

        
     

Are you a developer? Suggest improvements to the project!
Telegram: @ne_sravnim  | Write thanks for the project!

🖤 Want to support this research? Feel free to reach out via Telegram!

         Русская локализация     |     English Localization     |     中文本地化

The whole system. In one core. Precision in each layer. Visualization of players and the world, guidance, two radars and hardware input – in a single modular KevqDMA system. Is free. Open source code.

High performance, OOP structure and precision Each module is responsible for its own task. Assemble the appropriate configuration: from the minimum visual layer to the radar on a separate screen.

See how it works on the WebSite

Dual-worker pipeline – Camera at 300 Hz Separate KevqDMA data and camera-matrix workers. State exchange through SnapshotSlot [8] with atomic index publication.

One connected view of the system Reads, snapshot publication, camera, rendering, networking and recovery in one matrix. Select a node to explore its role.

Quick start Prepare the hardware and launch on the secondary PC.

WebRadar

How WebRadar works: DMA reads data from CS2’s memory, sends it to a local backend, and the backend streams player positions to the browser via WebSocket for map display. The address in the browser points to the local web service, for example 192.168.x.x or localhost or 0.0.0.0 Sequence: DMA → memory reading → backend → WebSocket → map display.

ESP

Display	Dichen Fuser V6

How ESP works: The game and the DMA Card are connected to the main PC, while the software runs on a second PC. The visual output is shown by overlaying the second display onto the main monitor using Dichen Fuser v6. Sequence: main PC → DMA data → second PC software → display overlay via Dichen Fuser v6.

Menu UI

Radar UI	WebRadar UI
Settings / Debug UI	Main Start UI
ESP UI

Radar UI Live session overview with tracked data, status, and radar controls.

WebRadar UI
Browser access panel with connection setup, quick actions, and session info.

Settings / Debug UI
Advanced settings and diagnostics for tuning, debugging, and state verification.

Main Start UI
Launch log for connection, version checks, offsets, initialization, and readiness.

ESP UI
ESP visual settings for toggles, previews, layouts, and colors.

Hardware Setup

This project was tested with a dedicated DMA hardware configuration based on the Captain DMA 75T v3.0 card. The card is configured with No-Show M2 SSD Disk (Stealth Firmware), which is used as the firmware profile for this research setup.

{ ! } Hardware behavior, firmware compatibility, and system stability may vary depending on your motherboard, BIOS settings, PCIe layout, and operating system environment.
Tested build: Windows 11 Pro (25H2) + Ryzen 9 9950X3D + X870 AORUS Elite motherboard (F10).

Captain DMA 75T v3.0

No-Show M2 SSD Disk Firmware

Captain DMA 75T v3.0 is used as the external PCIe-based memory access device in this research setup. It provides the hardware layer required for DMA interaction and external memory reading experiments.

The setup also uses several runtime DLL components; updates and documentation can be reviewed through the official project pages: <a href="...