All The Spaceships: Incredible (& Free) Star Swarm Demo

Okay, okay. God. OK. I will finally consider upgrading my processor after five years of this Core i7 920 being just fine. Between Steam Home Streaming needing to encode video eversoquickly and now Oxide/Stardock’s Star Swarm offering me the awesome sight of several thousand on-screen spaceships, I am ready to accept that my PC’s futureproofed days might at last be behind it. Fasterfasterfaster.

Star Swarm isn’t a game. It’s a benchmark for a game I want, a space RTS on a massive scale. We saw the video earlier this month, but the actual demo/bench is out now, and it wants to make your PC have a little cry.

The demobench can be grabbed from Steam, but even though it’s free it requires activation and that requires either a Stardock account or a torturous text file uploading process. If you can’t be buggered with all of that, here’s the gist of it:

That there is the Nitrous engine, which will be used for assorted Oxide and/or Stardock games of the future. including a Star Control reboot. I’m going to go out on a space limb and guess that a Sins of a Solar empire sequel using it is in the offing too.

Anyway, for now we must dream, aided by the proud sight of ENORMOUS SPACE WAR on our monitors. Now to spend far too much time trying to establish which motherboard I need/can afford.


  1. marlin says:

    Could anyone enlighten me as to what language that lady was talking? “parallelism”?

    • TheNotchyToad says:

      noun: parallelism 1. the state of being parallel or of corresponding in some way.

      • kris713 says:

        what Patrick said I’m shocked that a mom able to get paid $5552 in four weeks on the internet. did you read this site link

    • SuicideKing says:


    • Nouser says:

      That’s a perfectly fine English word for anything that is parallel.

  2. LionsPhil says:

    A Core i7 being merely “fine”? I’ve yet to find much to stress my Core 2 Quad aside from Planetside 2 and Kerbal Space Program.

    • TheNotchyToad says:

      On High settings PS2 can’t make my computer cry a little when in a massive fight at a large station. I also have the Core 2 Quad, but it is time to upgrade..5 years ago it ran every thing out there on ultra but it’s feeling the passage of time now.

    • Grey Poupon says:

      You’re probably reading your CPU utilization wrong. When one of your cores is capped it’ll read 25% (it’ll probably even keep swapping the load from core to core making them all seem to be running at 25%). There’s plenty of situations where your c2q is insufficient if your GPU is good enough.

      • LionsPhil says:

        Nah, I’m well aware of that, thanks. But more significantly: PS2 and KSP (when dealing with ~80 ton stock part stations) are the only games I’ve found that seem to struggle.

        Oh, wait, and Starbound, too. But everyone’s seems to, so I assume some of that is unfinished software gremlins.

        • Gemsa says:

          i have core2quad ’07 and the only issues i have had are with towns in AssCreed IV and Guild Wars 2. does limit GPU performance by 10-20% but not enough to be a problem

    • SuicideKing says:

      Disagree, have a Core 2 Quad Q8400 (@2.77 GHz), following games have reduced performance:

      Arma 3, Planetside 2, FreeSpace 2 Open with enough ships, Borderlands 2, America’s Army 2 Coop, FC3 occasionally, Receiver, Mafia 2 with PhysX enabled, Sleeping Dogs if you reduce AA, Total War: Rome II, War Thunder to some extent…yeah that’s all i can think of, for now.

      You also have to keep in mind that memory bandwidth has gone from 6-8 GB/s on a Core 2 era platform to 20 GB/s (typical) for a Haswell based system, which makes a significant difference (though the difference b/w 20 and 23, for example, would be less noticeable).

      Other things: AVX/AVX2, an L3 cache, far higher clocks, higher IPC…Intel’s 10% yearly improvement have added up pretty nicely over the last 6 years since the last Core 2 Quads.

      (Arma 3 is borderline unplayable for me, especially ‘Adapt’)

    • MattM says:

      I find that PCSX2 and Dolphin EMU love fast CPUs. I got noticeable improvements switching from a i7-920 @ 3.6 to a i5-3750K @ 4.4

  3. trjp says:

    Well that smashed the living funk out of my PC in a big way – so it’s a nice benchmark but that also means it’s likely customer base can be measured in dozens? :)

    Moreover – it’s not CPU-bound by any means, it really only hit 2 – at most 3 – cores for most of the time I watched it – and it also seems to stutter based on what I think is HDD access (some improvement if you move it to SSD but it still pauses with HDD spikes).

    Also worth noting that it continues to use CPU after the ‘demo’ is over – close it down and your PC will cool down faster! :)

    • SuicideKing says:

      Interesting, i thought it was supposed to be very well threaded…could you log performance with something like HWiNFO and see what’s going on? I don’t want to download 3GB at the start of my billing cycle, as my connection will be throttled to 512Kbps from 8 Mbps after 15GB…

      • trjp says:

        I noodled a bit – results vary depending on the exact test you choose – here’s a typical one tho

        link to

        Note that at the point I took that (cores and GPU are less than 50% used) the FPS had been sub 10fps for a few seconds…

        That would certainly suggest that the bottleneck is getting information between the CPU and GPU

        I didn’t include any HDD/SSD stats because if you run it more than once – it doesn’t seem to touch the HDD/SSD at all – so you can rule that out of your maths…

  4. Surlywombat says:

    This is a demo quite specifically for Mantle isn’t it? Since they haven’t released the mantle drivers, and anyone on older AMD and Nvidea cards will be screwed anyway I really don’t see why people should be concerned?

    • trjp says:

      You don’t NEED Mantle – it offers Mantle and D3D as options…

      Mantle is mainly aimed at getting more from lower-end hardware anyway tho? e.g. it’s for AMD’s ‘A’ processors and their GPUs?

      I could stick an AMD GPU into this PC and test it but I have a massive lazy on today – I can barely be bothered to drink this coffee I made (about an hour ago)…

      • Cognitect says:

        What Mantle does is significantly reduce draw call overhead and allow multiple threads to access GPU command buffers in parallel. In other words, it makes it much easier for the CPU to send a lot of work to the GPU.

        So yes, this is very beneficial if your system is bottlenecked by its (lower-end) CPU. At some point in the future, pairing a cheap CPU with a high-end GPU might actually make sense when you’re choosing components for a new PC, provided enough developers start supporting Mantle.

        However, a more efficient graphics API is a very good thing for everyone, not just the owners of slower CPUs. Even a high-end CPU can be a bottleneck, particularly if it’s tasked with keeping multiple GPUs fed. And having a more efficient API at their disposal might enable game developers to come up with much more ambitious designs.

    • Sharlie Shaplin says:

      I got 15fps avg on DirectX and an AMD 6950 I tested it with, that was without motion blur too. Guess I am going to have to finally upgrade.

  5. McTerry says:

    Great! I hate motion blur and blur of all kinds in games. The engine looks great.

  6. DuneTiger says:

    Yeah, I downloaded this yesterday, not reading the actual description, and then promptly deleted it. I’ll stick to my GTX, thanks.

    • Arcanon says:

      Read it again, it’s for Nvidia cards too. Just disable the damn motion blur in the config file.

  7. nimbulan says:

    Just FYI: This benchmark is made using an extremely resource hungry motion blur shader in order to make Mantle look better. Disabling motion blur in the config file will QUADRUPLE your framerate.

    Motion blur aside, I didn’t find the demo particularly impressive. Can the number of units being simulated really not be done by any other engine? Games like Supreme Commander and the upcoming Planetary Annihilation are designed around massive battles like this and already include similar levels of detail in the units themselves.

    • DatonKallandor says:

      It’s not about the detail of the units (and no, Supreme Commander and PA don’t have nearly as much detail in each unit – it’s much closer to the poly count and texture resolution of your high-quality Dawn of War 2 model), it’s the calculations of each individual projectile (they’re supposedly all simulated, not cheating visual-only FX on top of dice rolls). And even more hungry than the projectiles is the targeting search for every single unit – I suspect they’re doing it per-swarm instead of per-unit since that would melt most CPUs to slag but even that is incredibly hungry. If they’re truly simulating them each capital ship is also the equivalent of a very large fighter group in terms of CPU drain since each turret would be running it’s own search for target and line of fire checks.

      The amount of units is extremely high, and while there’s engines that could run the units, there’s none that could run this amount of individual units pathfinding, searching for targets and simulating each individual (very rapid firing) shot. At least not with this performance. But then I suspect they are “cheating” in several areas to achieve that performance.

      • LionsPhil says:

        Never mind SupComm; Total Annihilation was doing proper projectile modelling, back when a fast CPU was a 200Mhz Pentium, while swarms of fighter jets dogfaught in the skies. It’s not that much of a scale-up given CPU speeds have increased by over a factor of ten, and gained more than one core.

      • SpamHammer says:

        “nd no, Supreme Commander and PA don’t have nearly as much detail in each unit – it’s much closer to the poly count and texture resolution of your high-quality Dawn of War 2 model”

        Uh, I don’t know if we played the same games and benchmark. The units in Supreme Commander are MUCH more detailed, with much, much higher detail in the textures and bump mapping.

        “it’s the calculations of each individual projectile”

        SupCom does that too. It’s one of the things that made the game so resource intensive and so damn good. Unlike StarCraft and CNC, SupCom had a full 3D world where all the projectiles were calculated in real time.

        ” If they’re truly simulating them each capital ship is also the equivalent of a very large fighter group in terms of CPU drain since each turret would be running it’s own search for target and line of fire checks.”

        This is just flat wrong. A few guns on the side does not equal the same CPU drain as a fighter wing of 40+ fighters. In fact, the capital ships are all very, very easy to run. It’s the swarms that slow things down.

        “The amount of units is extremely high”

        Yeah, but in the benchmark, I never got over about 4,500 units. In Supreme Commander, you can have 8 commanders with 1,000 units each. That’s 8,000 units, all self shadowing, fully path and targeting AI’d, with full projectile calculation for each shot fired.

        Nothing in this demo is really impressive, it’s hard to see *why* it lags so much in different scenes.

  8. NZLion says:

    GPU: NVIDIA GeForce GTX 780 Ti
    CPU: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
    Physical Memory: 12875546624 (12GB)
    Average FPS: 24.83

    On the “Follow” scenario, Extreme settings (default)
    The frame rate dipped below 10FPS in places. I wish it listed min and max fps too. Looks like I’m in the same boat as Alec and need to upgrade from the same CPU

  9. Frank says:

    Don’t want. Whew, still no very compelling reason to upgrade. I have to wait for Windows 9, anyway.