Performance

Measurements were conducted on 04.11.2026 on a single Linux PC. Razdor 1.3.5 and Discord Times Community Update under Wine 11 were compared.

Razdor vs. Original

Comparison on a single GPU (RTX 4060) at 1024x768 resolution:

	Razdor	Original under Wine
frames per second (unlocked)	about 2000	about 176
frames per second (capped, 60 Hz screen)	60	60

Razdor renders a frame approximately 11 times faster on the same GPU. When using software rendering (Mesa llvmpipe), Razdor delivers about 48-73 FPS, while the original delivers about 19-34 FPS (the original’s internal counter showed 13-20).

If the original lags

When running via Wine on the desktop, the original may show around 15 FPS. However, when rendering via the GPU off-screen, it reaches 60 FPS (capped) and 176 FPS (unlocked). The low speed on the desktop is caused not by the game itself, but by launch specifics: 16-bit fullscreen mode at 1024x768, the desktop compositor, and Wine settings. DDrawCompat (ddraw.dll) from the Community Update did not run on this PC under Wine.

How much Razdor spends per frame

Measurements on an RTX 5060 Ti GPU (single CPU core, 1920x1080):

map	frames per second (unlocked)	game’s own load per frame
РК3-Столица (campaign map 3)	about 1950	0.15 ms
Другой берег (200x200, 61 armies)	about 1400	0.29 ms

The budget for a single frame at 60 FPS is 16.7 ms. Razdor uses only about 1-2% of this time on campaign maps. Screen resolution (1920x1080 vs 1024x768) has almost no impact on the load.

Battery saver

The “Battery saver” setting (the “Other” tab, off by default). When nothing happens on the map — no windows are open, the hero neither walks nor waits, no armies are shown, the camera and the pointer stand still, and you press nothing — new frames are drawn 10 times a second instead of at full rate. As soon as something happens, the frames run at full rate again. The game rules do not change.

Map animation

The “Map animation” setting (the “Map” tab, the row before “Mist instead of black”, “On” by default). A click on the row cycles through its values:

“On” — everything on the map moves, as usual: the trees, the armies, the water, the weather, the lights.
“Near only” — the movement stops when the map is zoomed out, and comes back as you zoom in.
“Off” — the map is still: the water, the living map and the ship wakes are off, and new frames are drawn less often, as with “Battery saver”. Less load on the graphics card and the battery.

The game rules do not change.

Graphics quality

The “Graphics quality” row is the first one on the “Map” tab. A click steps through the presets: “Original”, “Low”, “Medium”, “High”, and back to “Original”. One row sets all the settings below at once. The row shows “Custom settings” when one of them has been changed by hand. A new install starts on “Original”, the original’s look with no new effects. An installed game keeps its own settings.

Preset	Map resolution	Light	Shadows	Mist	Water	Weather and living map	Map animation
Original	100%	off	“Painted”	“No”	off	off	“On”
Low	75%	off	“Painted”	“Simple”	off	off	“Near only”
Medium	100%	on, medium light quality	“From the sun”	“Simple”	on, every second frame, medium quality	on	“Near only”
High	100%	on, high light quality	“From the sun”	“Clouds”	on, every frame, high quality	on	“On”

On a weak phone, start with “Low”. Changing any one of these settings by itself makes the row show “Custom settings”.

Map drawing speed

The map draws faster, especially in 2.5D and when zoomed out. The picture is the same and the game rules do not change. What has been sped up:

The map’s objects (trees, buildings, rocks, hills) are kept by areas: each frame draws only the areas on the screen, not every object on the map.
What does not change from frame to frame is worked out once: the sway of trees and plants, the screen size, the turn of objects in 2.5D.
The water does not recount every hill on the map each frame, the fog does not compare the whole map again, and the sprites are drawn without new memory for each sprite.

Large Maps

Before acceleration (Razdor 1.3.5). Synthetic tests on a single CPU core (time for one game tick, when AI moves armies and checks events):

map	10,000 events, no armies	100 wandering armies (typical / worst case)
200x200	0.12 ms	21 ms / 68 ms
400x400	0.12 ms	71 ms / 280 ms
800x800	0.13 ms	420 ms / 1450 ms

What has been accelerated

Measurements of the time required by the AI for 2 game days with 100 wandering armies and 10,000 events:

map	before (single core)	now (8 cores)	faster
200x200	2.1 s	0.7 s	2.9x
400x400	8.1 s	2.6 s	3.1x
800x800	36.9 s	11.5 s	3.2x

Army AI behaves exactly as before: army paths have not changed; this has been verified on all game and community maps.
Pathfinding has become cheaper, and pathfinding for different armies occurs on multiple CPU cores simultaneously (on a single core, it is about a quarter faster than before).
The map no longer freezes while the AI is thinking: the longest frame while moving across an 800x800 map was reduced from 2.4 s to 0.054 s; if the AI cannot finish within a frame, it continues in the next, and the map continues to render.
In the future, an experimental setting with a faster (but not identical) AI for large maps may appear; it is not available currently.
10,000 events do not cause problems on any map size (about 0.12 ms per tick).
The main load is army pathfinding. Every time an army chooses a new target, it searches for a path across the entire map, so the cost grows with the area.
Before acceleration, on a 200x200 map with 100 wandering armies, every tick caused a slight stutter (skipping 1-2 frames). On 400x400, a clear hitch was visible. On 800x800, the hero before acceleration could only move about 3 steps per second while waiting for army calculations to complete.
When the map is static, even 800x800 delivers over 2000 FPS. The bottleneck is the AI, not the rendering.
Real maps in the game do not exceed 200x200 and contain up to 60 armies (many of which are stationary), so current campaign and community maps are far from the limits.

Performance test

In Settings → Other, the “Performance test” row has a “Run” button. The test measures how fast the game runs on this device, on the map “Другой берег” (Another Shore): near (the default zoom) and the whole map at once.

“My settings” (about 30 seconds): your settings and map view.
“All presets” (about 3 minutes): “Original”, “Low”, “Medium” and “High” in 2.5D, and “Original” on the flat map.

During the test the game takes no clicks or taps; Esc stops it. Your settings do not change: the test applies its own values for the run only, then everything comes back. The FPS cap and the battery saver are off during the test; vertical sync stays as the device sets it.

The results table shows FPS, the frame time (median and the 99th percentile) and the number of slow frames (longer than 33 ms). The “Save” button writes the results to a text file in the Razdor folder (next to the saves) and copies the text to the clipboard where that works. You can attach the file to a bug report.

Technical limits of the map format:

Up to 255 armies per map.
Up to 65,535 events.
Up to 64 events in a single point or building.
Map size up to 8192x8192 (up to 67 million tiles).

Tip for map creators: on maps larger than 200x200, keep the wandering armies smaller: they are what cause the load. Guard posts and patrols were not checked in this measurement.

Debug mode

In “Settings” → “Other”, there is a “Debug mode” line. It is disabled by default. If you enable this mode, a performance log (slow frames and loads) will be recorded. A bar will appear at the top of every screen: “Debug mode enabled — FPS may decrease.” This change takes effect immediately. To report a bug, enable the mode, reproduce the issue, and then use “Send the log…” on the same tab. On Android, one file will be saved via a save dialog; on the computer, a folder with the log will open.