Next-Gen Tech Allows for Massive Scale RTS Combat

Next-Gen Tech Allows for Massive Scale RTS Combat

Advanced technology from AMD means that strategy games will soon be able to support thousands of individually-simulated units at a time.

Real-time strategy games like Starcraft often have a strange sense of scale, where even the largest battles only have a few dozen combatants. This is partially for balance reasons, but also because our computers can only simulate so much at once. Technology marches on, though, and new innovations are starting to make hardware-induced unit caps a thing of the past. A new engine called Nitrous, using AMD's Mantle tool, is powerful enough to simulate battles with over 5000 units, each with moderately complex AI and physics. And this is only the beginning.

Let's get technical for a moment. Computers have a CPU and a GPU; the CPU generally handles math while the GPU does graphics. We have CPUs and GPUs that are really powerful, but they're two separate units, and the bottleneck comes into play when they need to communicate with each other. Mantle essentially removes (or greatly widens) that bottleneck: all your CPU cores can talk to the GPU at the same time, making it possible to do some really complex simulations.

In terms of games, that means developers will be able to have thousands of AI or physics-driven objects on the screen. An RTS demo called Starswarm shows off the power of this new tech with a massive space battle; the demo ran at a choppy 13 frames per second on conventional hardware, but Mantle more than tripled that. It's not just for strategy games, either: an upcoming update for Battlefield 4 will integrate Mantle tech to increase performance by up to 45 percent. If this catches on, this technology alone could provide developers with a bigger boost in power than any "next-gen" innovations currently available.

Update: Now with video!

Source: Engadget

Permalink

Holy crap that is amazing.

Especially so when you go to that Staswarm link and watch the video they are using to show everything off.

What about the Total War series? That has been able to run battles with upwards of 40000 soldiers by now?

Now I really want Homeworld 3 to be a thing.

I for one look forward to Eve Online being rebuilt using mantle.

Soviet Heavy:
What about the Total War series? That has been able to run battles with upwards of 40000 soldiers by now?

WHile I'm not the expert in Total War series, I've been playing with Shogun 2 lately. And while there are indeed large battles, there aren't individual units that much. There are all those soldeirs, but 100 or 150 of them act at the same time, in almost the same way, doing almost the same actions. Each soldeir is more a particle in a body than anything else, so actually there are things like only 50 or 60 real units (although fairly complex ones).

If this is true, well, these should be each doing it's own thing. And while asweome in some part, I'm not sure it is indeed that important. Why? Because even if the computer can manage it,we humans can't. We use the units as a mass in huge games (like in AoE II or TOtal War) or micromanage in games with smaller counts (Starcraft or Warcraft). SO adding units, well, it is impressive in a technical sense, well I'm not sure it will be altering the RTS scene taht much (well bigger battles, but the playstyle still will look the same). The demo looked amazing, yes, but unless it is multiplayer, those xtra numbers won't add much I think.

I opened this piece thinking it was going to be a console company or console dev spewing bullshit because they're people who scream next-gen the loudest.

Surprised/10.

Let's get technical for a moment. Computers have a CPU and a GPU; the CPU generally handles math while the GPU does graphics. We have CPUs and GPUs that are really powerful, but they're two separate units, and the bottleneck comes into play when they need to communicate with each other. Mantle essentially removes (or greatly widens) that bottleneck: all your CPU cores can talk to the GPU at the same time, making it possible to do some really complex simulations.

To say nothing of the homogenous memory that allows the CPU and GPU to both read a write directly to the same RAM instead of wasting cycles moving data from one to the other, and then back again.

The whole evolution of AMD's APU project has been nice to watch, great to see it paying off.

This will be cool. I'm imagining large scale battles, where where hills and entrenchment are being lost and taken in real time. A strategy games that have lines for troops to fall back and regroup is my dream game.

I love RTS games, but it always bothered me that most encounters ended when one side completely annihilated the other. Finally getting battles that breathe will be great.

Wow... I look forward to that technology getting into consoles in 8 years time!

In the meantime you can go play Cossacks: European Wars in any shitty PC and claim the superiority of sprite based graphics!!!

OH COME ON! I suck so much at RTS games I can't even handle a hundred units, how the flying fuck am I supposed to command five thousand?

I hate how Starcraft keeps being mentioned when Supreme Commander could have a population limit of 1000 each for 8 players (yes buildings counted against it and if you built walls you could really cut into it but that leaves room for 800-900 units per side). In terms of scale Starcraft is nothing, it's tiny armies. Take Men of War Assault Squad where you'll easily have hundreds of independent units or Sins of a Solar Empire with hundreds of ships per side. Starcraft, while a great RTS that I've put a lot of hours into, does not belong in conversations about scale.

my question is this how will this handle on my nividia card are we talking like tress fx in tomb raider or are we talking slight chug?

I think I'm ready for a new Supreme Commander game now.

But seriously, I think this will do greater wonders for the Space Sim genre rather then the RTS side of things.

Sure it sounds awesome to be able to control that many units but they will either have to be insanely tiny or the battlefield has to be stupendously large, either way it doesn't bode well for the player that has to control and direct all that shit. But imagine being a fighter pilot in a space sim that can actually simulate a battle of 5000 spaceships, man that has to be awesome.

Xan Krieger:
I hate how Starcraft keeps being mentioned when Supreme Commander could have a population limit of 1000 each for 8 players (yes buildings counted against it and if you built walls you could really cut into it but that leaves room for 800-900 units per side). In terms of scale Starcraft is nothing, it's tiny armies. Take Men of War Assault Squad where you'll easily have hundreds of independent units or Sins of a Solar Empire with hundreds of ships per side. Starcraft, while a great RTS that I've put a lot of hours into, does not belong in conversations about scale.

Supreme Commander does get rather slow with that many opponent all reaching the upper unit limit, even on modern PC's with good graphics cards.

What AMD is showing here on the other hand doesn't seem to slow down one bit.

Which just means that we need a new Supreme Commander (or simply a remake).

Saw the demo and my was truly amazed (loved the little ships blasting away with both their forwards and rear guns) but while this seems all exciting and I love anything that advances PC gaming I really am not that familiar when it comes to all the tech details.

Could someone who understands it better tell me if this type of stuff will require a massive upgrade in hardware to work or if it just makes what's their more efficent?

kajinking:
Saw the demo and my was truly amazed (loved the little ships blasting away with both their forwards and rear guns) but while this seems all exciting and I love anything that advances PC gaming I really am not that familiar when it comes to all the tech details.

Could someone who understands it better tell me if this type of stuff will require a massive upgrade in hardware to work or if it just makes what's their more efficent?

Primarily, it mainly looks like it's allowing for computers to really use all their cores that processors have been being built with. At least that's my take on it. I know that in general, partly due to just how much more difficult it can be to program for more cores and partly the general lack of computers with multiple cores, most games aren't designed to actually make use of more than one or two cores.

That wouldn't be so bad, except that most of our processor improvements in the last bit of time have been in multi-core processors; getting processors to run much faster has run into many technical problems.

Though, plenty of it is also the video card manufacturers developing the tools to interface with their cards more directly, rather than relying more on other, more neutral companies ala directX or openGL. Which is very cool imo. Those GPUs are amazing if used properly; something that isn't done too much so far. Very complicated things to work with sometimes.

The focus on Mantle here is perhaps a bit misleading..

Worth noting that Nitrous while yes, Mantle pushes the potential a further, Nitrous being a big step forward is not dependent upon Mantle. It was in fact originally developed for DirectX and than ported to Mantle, and is still looking pretty impressive even if you don't have an AMD card.

Here's the only reference to current performance data I could find: http://linustechtips.com/main/topic/98295-amd-tech-day-kaveri-mantle-slides/

In any case, supposedly we'll be seeing Steam Benchmarks at some point this month, looking forward to some solid data.

I can't wait to field an army of 1000 METAL BAWKSES!

Jadak:
The focus on Mantle here is perhaps a bit misleading..

Worth noting that Nitrous while yes, Mantle pushes the potential a further, Nitrous being a big step forward is not dependent upon Mantle. It was in fact originally developed for DirectX and than ported to Mantle, and is still looking pretty impressive even if you don't have an AMD card.

Here's the only reference to current performance data I could find: http://linustechtips.com/main/topic/98295-amd-tech-day-kaveri-mantle-slides/

In any case, supposedly we'll be seeing Steam Benchmarks at some point this month, looking forward to some solid data.

From what was said, with Mantle they expect to not only get higher framrate, but accomplish nearly 10x the amount of computations then they're currently getting do to the code being so poorly optimized and rushed. So I'd say you're under selling it, if there data is correct that is.

Soviet Heavy:
What about the Total War series? That has been able to run battles with upwards of 40000 soldiers by now?

It takes a lot of shortcuts, like a group continuing to act as one even when they get broken up by combat (just look at stragglers trying to rejoin their group). And even then the pathfinding still glitches out a lot. With this new tech they will be able to script every one of those 4000 soldiers (and possibly thousands more)individually

Does this mean we'll actually get a triple a RTS that isnt a total war game? Because the last one I remember is SC2, and then I blank for...years.

Coakle:
This will be cool. I'm imagining large scale battles, where where hills and entrenchment are being lost and taken in real time. A strategy games that have lines for troops to fall back and regroup is my dream game.

I love RTS games, but it always bothered me that most encounters ended when one side completely annihilated the other. Finally getting battles that breathe will be great.

You can do that in company of heroes as the British you can retreat to your commander, not exactly what your looking for but better than baby's first RTS (starcraft). In fact if you do let units keep getting totally annihilated then you are going to lose pretty fast.

For those of you wondering just how you're going to be able to control 5000 or 10,000 units all at once and thinking it'll be impossible, I have a question for you...

...how do you think generals do this in REAL LIFE?

That's what I predict out of this. You may still select squads or groups like in the Total War series... a single lieutenant designating a squad or something of the like... but the battles themselves, the simulations of unit vs. unit, THAT is what will change. THAT is what we will see different. From afar it will be similar to how it was before, but from ground view, it'll look quite different.

Now we just need someone to make an actual strategy game so we can use this. Rome 2 sucked. Starcraft is OK, but hardly designed for such large amounts of units. Other than the Dawn of War series, which was owned by THQ, who else is even making RTS games these days? Shame about THQ in particular, since a Warhammer 40K game with full-scale battles of tens of thousands would be fucking epic.

Crimsonmonkeywar:

Jadak:

From what was said, with Mantle they expect to not only get higher framrate, but accomplish nearly 10x the amount of computations then they're currently getting do to the code being so poorly optimized and rushed. So I'd say you're under selling it, if there data is correct that is.

Yes, I am under-selling it, primarily as I am just trying to switch focus from Mantle to Nitrous itself. This article would have one think that Mantle is the defining point of Nitrous, and does not give nearly enough credit to the guys at Oxide who had made a very impressive engine regardless, one that will serve more than just those using recent AMD cards.

More, even the performance of Nitrous is speculative at this point barring any benchmarking of real world applications, I have no desire to further compound the speculative nature by throwing around figures like 'Nitrous x 10!' that are even further from being real world applicable at this point.

truckspond:
Wow... I look forward to that technology getting into consoles in 8 years time!

The article doesn't explain it properly, but what Mantle really does is allow the software to act closer to the hardware.

But that's already what consoles do. You might have heard this comparison before: "Consoles run more efficiently because the games can be programmed for their specific hardware instead of supporting the many PC configurations".

Mantle aims to bring some of that efficiency to PC.

That article on StarSwarm was terrible.

First off: StarSwarm is a BENCHMARK DEMO not a game. They may be working on a spaceship RTS game but the footage you are seeing is for benchmark purposes not to show off gameplay.

Next: The article claims you need a robust CPU to run the demo but the apparently didn't watch the presentation that Oxide gave on this demo because Oxide specifically stated that they UNDERCLOCKED their CPU and that the demo was GPU bound not CPU bound. In their presentation they talked about how the engine scaled REALLY well with CPU cores so you don't need a super expensive CPU to run the demo you just need a CPU with a lot of cores (So AMD's 8core CPU's would be a good pick) but even then the CPU is not as important as the GPU in this case because the whole benefit of Mantle was it allowed them to fully utilize the GPU without taxing the CPU with hidden extra operations.

I just wish we could finally stop developing graphics further and further and just focus on creating new gameplay concepts.

I'm not all that impressed because it's all about quality, balance, and yes... writing, when it comes to a good RTS experience, not just being able to throw out crazy numbers of units. While I guess some people really liked it, I could never get my mind around the concepts and why the factions were fighting (and the whole idea of one suit building an army and that resolving an entire war, when you'd think each side would have thousands upon thousands of nanite-constucting battle suits, but well... yeah, best not ponder that) in say "Supreme Commander" which as it's selling point was pushing the number of units you could crank out and how it had carefully balanced sea, air, and land components with one faction being able to morph it's units to more than one element as a special ability if I remember.

The point I'm making is that while I suppose the technological achievement is impressive, I think things have remained on a smaller scale largely because it's both easier to control, and to ensure the quality of the experience. Blizzard could have say moved to a larger scale, say Supreme-Commander-like but I don't think it would have had the same quality, and I'd imagine they do also, because they didn't do it at the moment.

At any rate if you want to impress me, come up with a system where you can crank out that many units and provide me with an interface where I feel like I can actually control that many and remain more or less in command of my entire battle. Simply churning out unprecedented numbers of AI bots that might at best wind up following a patrol pattern in of itself doesn't sell me.

 

Reply to Thread

Log in or Register to Comment
Have an account? Login below:
With Facebook:Login With Facebook
or
Username:  
Password:  
  
Not registered? To sign up for an account with The Escapist:
Register With Facebook
Register With Facebook
or
Register for a free account here