DEV Community

dev game
dev game

Posted on

Why Space Marine 2 Bet on Swarm Engine: The Horde Tech Behind a 12-Million-Player Power Fantasy

Space Marine 2 runs on Saber's proprietary Swarm Engine, not Unreal Engine 5. That single choice explains why its Tyranid hordes feel different from every other 2024 shooter: hundreds of bodies on screen at once, climbing over each other to reach you, while three Space Marines hold a doorway with chainswords and bolters.

Developer Saber St. Petersburg and publisher Focus Entertainment shipped the game for PlayStation 5, Windows, and Xbox Series X/S on 9 September 2024 as a third-person shooter with hack-and-slash melee, with a campaign playable solo or in 3-player co-op Space Marine 2 release and engine. It attracted more than 2 million players on launch day, sold 4.5 million copies by 16 October 2024, and surpassed 12 million players by 16 April 2026, with a sequel now in development Space Marine 2 sales and sequel.

The search question behind this article is specific: "what engine does Space Marine 2 use" and variants like "what game engine does Space Marine 2 use" and "what graphics engine does Space Marine 2 use." Google Suggestions returned all four phrasings on 15 November 2026, which is rare niche demand for a proprietary engine. Bing organic search for the same query degraded to engine.com dictionary noise, so this analysis relies on direct primary-page retrieval instead of a correlated SERP.

What Swarm Engine actually is

Swarm Engine is Saber's in-house horde engine. It was built to render huge crowds and purposefully handle up to 500 zombies on screen at one time Swarm Engine purpose, with the full game supporting up to 1,000 enemies appearing simultaneously who can climb onto each other to reach players on higher levels 1,000-enemy support.

It did not appear from nowhere. Saber Interactive was founded in 2001 by Andrey Iones, Matthew Karch, and Anton Krupkin, built a 3D engine from scratch with artists in St. Petersburg, then built proprietary Saber3D for TimeShift (2007) and kept updating it for current games Saber founding and Saber3D. The studio later used Saber3D visuals over original gameplay code for Halo: Combat Evolved Anniversary, handled The Witcher 3 Switch port, and launched World War Z to more than 1 million units in its first week Saber history. Swarm Engine is that lineage specialized for one job: crowds as gameplay.

For context on what Saber rejected, Unreal Engine is a C++ 3D engine from Epic Games whose latest generation Unreal Engine 5 launched in April 2022 with Nanite virtualized geometry and Lumen dynamic global illumination Unreal Engine 5 features. Commercial use carries a 5% royalty over $1 million, waived for Epic Games Store exclusives Unreal royalty. By sales, Unreal accounts for about 31% versus Unity 26% and proprietary engines combined 42% engine market share.

What Space Marine 2 asks the engine to do

The campaign puts you as Lieutenant Titus with Brother Chairon and Sergeant Gadriel, playable by co-op partners or AI, plus Imperial Guard allies, fighting Tyranids and Chaos forces of Tzeentch including Thousand Sons Rubric Marines campaign setup.

The combat loop is armor-first, not health-first. Armor protects you, health drops when armor breaks, health only returns via medkits or high-level abilities, and armor regenerates by performing executions on enemies armor and execution system. You can parry and riposte with a gun strike, but some attacks must be dodged parry and dodge.

Modes reuse the same streaming muscle in different shapes:

  • Operations: 3-player co-op PvE missions parallel to the main story
  • Stratagems: Operations variants with random modifiers
  • Siege Mode: endless rogue-like defense of a fortification for as long as possible
  • 6v6 PvP: Annihilation, Seize Ground, Capture and Control, and Helbrute Onslaught between Imperium and Chaos modes

You pick from seven classes — Tactical, Assault, Vanguard, Bulwark, Sniper, Heavy, Techmarine — with cosmetic customization seven classes.

What Swarm gained, and what it cost

Gained: a cast, not a background. Most shooters fake hordes with distant sprites or capped spawns. Swarm treats 200 bodies as one directed actor. Tyranids pour, pyramid, spill over rails, and re-path around a Bulwark shield. World War Z proved the trick with zombies that pile to reach high ground; Space Marine 2 re-skins it as Hormagaunts climbing ruins to reach a Heavy's heavy bolter. No licensed engine gives you that crowd director for free. You would have to build it inside Unreal anyway.

Gained: melee readability at scale. A horde shooter that is also a parry game should not work. Parry needs readable wind-ups, but 50 attackers cannot all telegraph cleanly. Saber's answer is brutal and smart: only a few enemies are ever "duel-active" while the rest are pressure scenery. The armor-execution economy makes this explicit — the crowd is ammunition for your next armor refill. That turns quantity into a resource faucet instead of visual noise.

Gained: Operations as asset amortization. Operations missions run parallel to the campaign, not after it. That is a single-player game using an MMO trick: same biomes, same Swarm streaming, different objectives and modifiers. Stratagems and Siege Mode push it further. When your engine's strength is streaming crowds through corridors and arenas, you do not need twice the art. You need twice the rules.

Cost: no asset store, no hiring pipeline. Unreal gives you Quixel, Marketplace, and thousands of developers who already know Blueprints. Swarm gives you control and a lonely job posting. Every chainsword animation, every Rubric Marine glow, every Xbox Series S memory budget is your problem. Saber can pay that because it has long engine continuity since Saber3D and a catalog — World War Z, Evil Dead, SnowRunner — that amortizes the engine. A 12-person indie cannot.

Cost: beauty on Saber's terms. Nanite and Lumen would have given prettier static bricks. Swarm gives better moving meat. Space Marine 2 looks excellent in motion and merely good in screenshots, which is the correct trade for a game where the floor moves.

Why not Unreal Engine 5? My reading

Everyone else rented. Saber owned. Here is where I depart from the usual "Unreal is better graphics" roundup.

Unreal 5 is a real-estate engine. Nanite sells static density — more rocks, more ruins, more cathedrals per frame. Swarm is a casting engine. It sells directed extras — more bodies moving with purpose per frame. Space Marine 2 does not need a more detailed cathedral. It needs 80 Tyranids to decide, in the same second, to run around the cathedral, climb it, and fall on you. That is AI pathing plus animation LOD plus physics culling, not triangle throughput.

My contrarian take: the Tyranid pyramid is a loading screen in disguise. When zombies or Gaunts climb each other, the engine can freeze far agents into cheap climbing animations, skip their AI, and spend the budget on the five that will actually hit you. Piling looks like excess. Technically it is thrift — a way to show 500 while simulating 20. Unreal could do this, but you would be fighting the engine's defaults. Swarm defaults to it.

Second, armor-by-execution is Swarm's confession. If you can put 1,000 enemies on screen, players will take chip damage no matter how skilled they are. A health-bar game would become unfair. So Saber made armor a melee faucet: get swarmed, execute, refill, keep moving. The horde is not difficulty. It is economy. That is why Operations feels fair at 3 players while World War Z needed 4 — melee executions close the loop faster than headshots at range, so you need one fewer gun.

Third, 3-player co-op is a server budget, not lore. World War Z shipped 4-player co-op for shooting. Space Marine 2 ships 3-player for shooting plus parry plus executions. Parry is latency-sensitive in a way shooting is not. Cutting one player cuts state replication, animation sync, and readability clutter. The missing fourth Marine is not a story decision. It is Swarm admitting that swords are more expensive than guns to synchronize.

Fourth, owning the engine after the Embracer divorce is a hiring filter. Saber left Embracer Group in March 2024 under Beacon Interactive. When you own Swarm, you hire people who want to work on crowds, not people who want Unreal on their resume. That self-selects for the unglamorous work — occlusion, streaming, animation budgets — that makes hordes feel good. Renting Unreal would have made hiring easier and the game more generic.

What indie developers can steal without Swarm

  1. Make the crowd the resource. If you cannot afford 500 agents, afford 30 and make kills refill something essential. Vampire Survivors does this with gems. Space Marine 2 does it with armor. The lesson is not scale. It is faucet.

  2. Parallel missions, not more maps. Operations reuses campaign geography with new modifiers. If your engine streams one biome well, ship two rule sets for it before you build a second biome.

  3. Parry needs a crowd manager. If your melee game adds adds, decide explicitly which enemies are allowed to attack while the player parries. Telegraph only those. Freeze the rest into approach animations. Players read fairness, not count.

  4. Do not rent density to show motion. If your fantasy is "overrun," Nanite will not save you. A cheap instanced horde with good audio and good piling beats a photoreal empty hall. Spend on movement, not bricks.

Train the feel in a browser before you install 100 GB

Space Marine 2's skills split cleanly into melee-horde, clutch shooting, and doorway holding. These lightweight browser games train each slice without copying the game:

If chainsword timing is what hooks you, Assassin Hero Infinity Blade is the closest natural warm-up. It is a roguelike action game about hit-and-run against relentless hordes and bosses Assassin Hero horde action. Practice the same habit Titus demands: dive in, execute, get out before the pyramid collapses on you.

If Operations PvE and 6v6 PvP clutch moments are your goal, Block Contra: Clutch Strike trains the exact micro-skills — crosshair placement, burst control instead of panic spraying, and holding angles under pressure. Its guide stresses controlled fights and reload discipline, which maps directly to Bolter burst control when Gaunts close distance.

If Siege Mode endless defense appeals to you, BodyCamera Shooter is a useful doorway trainer. It is a single-player wave FPS where you clear buildings through a shaking fisheye body camera that punishes open movement BodyCamera wave survival. Its tactics — fire in short bursts, fight from cover, pre-aim doorways, reload only behind concrete — are Siege Mode in miniature: hold the choke, let the horde come to you, do not chase kills into the open.

None of these replace Swarm's 500-body direction. They isolate the player habits that make Swarm readable.

FAQ

Is Space Marine 2 on Unreal Engine 5?
No. Its infobox lists Swarm Engine engine credit, the same proprietary family that powered World War Z's hordes Swarm in WWZ. Confusion comes from how many 2024 shooters did rent Unreal.

Why 3-player co-op instead of 4?
The game supports 3-player campaign co-op and 3-player Operations alongside 6v6 PvP co-op structure. My analysis: melee parry sync plus horde AI is more expensive per player than pure shooting, so 3 preserves readability and server budget.

Does armor regen mean the game is easy?
Armor regenerates via executions, health does not except via medkits or abilities armor rules. That makes aggression mandatory. Passive players bleed out. It is difficulty through forward pressure, not damage sponge.

Will there be a Space Marine 3?
A sequel is in development according to the game's overview sequel status. No engine change has been verified. Expect Swarm to return unless Saber announces otherwise — the horde is the brand.

Swarm Engine is not the prettiest technology of 2024. It is the most honest for this fantasy. Space Marines do not win by outshooting the Tyranids. They win by standing in the flood and refusing to move. An engine built to render 1,000 climbers understands that better than an engine built to render perfect bricks.

Top comments (0)