Posted by: Kai_LeRai | August 15, 2009

Gaming Misconceptions: Killing Zombies while killing your Hardware – The Story of Cryostasis

Whenever a game is released to the gaming community it has to stand up to the expectations and judgments of the gamers. In particular that means, that it will be compared to other titles, that are considered equal in visual quality and performance. But once in a while, a game is so completely different, that those standards just don’t work. The problems arise when people don’t realize that and still apply them. One of the victims is Cryostasis.

Released at a time, where gaming technology was still rudimentary, Cryostasis had to face the onslaught of negative opinions that many titles experience that are ahead of their time. Cryostasis is almost perfect within its own realm and that makes it an easy victim for ignorance and prejuduice. Those are the result of never changing judgment patterns of gamers. For this game, they need to adapt. Alas, that’s  a rarely seen quality nowadays and to make things worse, Cryostasis is a first person shooter. Fans of this genre have never been the most tolerant or understanding, let alone interested enough to enjoy a well done plot over lesser attractive gameplay. After all, fans of this genre are entertained merely by shooting things and how many hardcore FPS-titles can you name that offer an outstanding storyline? (No, Half-Life doesn’t count, especially not HL2)

So, additionally to climbing that steep mountain of expectations, Cryostasis also has to carry the burden of FPS-gameplay, but enough of the metaphors and let’s get to the meat of the issue.

Visual Quality: Cryostasis vs. Crysis

A crisis indeed, because every comparison of Cryostasis to any other current game out there is doomed to fail. Cryostasis uses the unique AtmosFear 2.0 engine. Comparing games that are based on different engines makes little sense, unless it comes to gameplay or a direct comparison of similar effects done by both engines. Cryostasis doesn’t give you such a basis though. People usually think that Crysis is visually superior to Cryostasis, but this is arguable. The normal gamer usually doesn’t give much thought to what works behind the scenes, or even pays enough attention to see the important details that eat up performance.

Crysis is a game that shines at polygon pushing. Life-like high-poly enemies, plants, water…you name it. The game wants to deliver superficial graphical excellence and it achieves that. The strain on the hardware when it came out was immense and even today it is still the major benchmark for new high-end gaming cards and that will not change anytime soon.

Cryostasis is different. The polygon count is rather low. NPCs seem stiff and less life-like. Enemies however do look very much alive and very realistic. Still, that doesn’t justify the strain on the hardware. Cryostasis’ performance-chugging black hole is far more subtle, yet very demanding.

Cryostasis animates tens of thousands of particles in certain scenes…all in realtime and all physically accurate. No other game does that so far and also Crysis doesn’t do it. Sure you have some basic physics and in comparison to Cryostasis they are really basic. The number of physically influenced objects at a time in Crysis is fairly low. In Cryostasis you will end up having fights with monsters that will make massive use of physics. Bullet sparks that fly all over the screen, bouncing around on the floor in a realistic fashion are only one factor. Hit an enemy with a bullet and dozens of iceshards fly off him and react accurately with their surroundings. Now imagine attacking him with a tommy gun, which usually results ina  massive amount of sparks and ice shards at the same time. Throw in other physically influenced effects such as water streams from broken pipes, nearby fires, breaking off icicles or stumbling over some of the debris during a fight and you have acquired quite some workload that can trouble many graphics cards without even having applied many of the other visual goodies.

Numerous ice shards fly off when you hit an enemy. This also happens when the bullets hit ice.

If a bullet hits metal, the resulting spark-effect emits light. You can see the yellow light being reflected in the icy surfaces around the spark area.

The lighting effect dies off quickly and sparks are generated. They rain down to the ground in a realistic fashion, bouncing off walls if they hit them.

Sparks raining down. Pay attention to the great visual filter that distorts the view and gives the screen a frozen over look.

Aside from the complex physics Cryostasis also has a higher particle count than any other game out there. Water dripping from the ceiling, water gushing through open hatches water shooting from your water gun, water falls coming through broken ceilings…water all around you. It’s kind of what you would have expected Bioshock to be like, which only let you experience water rather scarcely. Speaking of which, let’s compare the two games’ water. Water has always been a strain on hardware and new games are no different. In Bioshock you only get the better water effects when you play the game with DirectX 10, but even then it is barely comparable to Cryostasis. The water in Bioshock is scripted and therefore puts no load on the CPU physics-wise. Also, Bioshock seems to use a mapping effect for the water, which is an efficient way of pulling the water effect off, but it isn’t realistic. In fact, the water in Bioshock looks downright terrible. It reminds me of the same style 10 year old games used. The only difference is its realistic looking transparancy effect. Cryostasis relies solely on particles to pull off all water effects! Where Bioshock (and all other games, including Crysis) used a texture to simulate a stream of water, Cryostasis uses thousands of particles and all behave physically accurate! This is a quantum leap in terms of realism and considering these different technologies you might see now, why Cryostasis needs much more power than games like Bioshock or Crysis. But it doesn’t end there.

Water running through a broken ceiling. You can see that it consists of many small objects. Even reflections from the light source above are visible in numerous drops of water. Also take note of the motion blurring effect for every water particle, making the whole element look much more realistic.

Cryostasis supports many texture filtering effects to make the surroundings look more real. Apart from your standard run-in-the-mill water reflection effects, The game also reflects the surroundings. The effect can be easily overlooked, because it blends in so well, due to its natural look. Also, the game features impressive water caustics effects. The term caustics describes the behavior of light when it hits a water surface. It is not just merely reflected, but also broken and overlayed and causes complex light patterns on surrounding objects. Cryostasis renders a realistic caustics effect that is also influenced by the water’s movement and additional light sources. While you will find some water in Bioshock or Crysis, you will have a hard time finding caustics.

This is water rendering without environment reflection.

Here environment reflection is turned on. What you can see is the ceiling. The effect is very subtle and can easily be overlooked. Still, without it, the water looks a lot less realistic. Also pay attention to the caustic effect in the upper part of the picture. Compare it with the prior picture and you can see that the caustics change according to the water movement.

Here you can see the beautiful caustic effect.

Here is the scene with caustics turned off.

Cryostasis has a very interesting lighting model. Lighting, as you might know, requires massive computing power. When Stalker came out, it was perfectly playable with static lighting, even on older computers. But when you activated full dynamic lighting, framerates plummeted. The difference in visual quality was immense, though. Try standing on a field at night, surrounded by trees and bushes during a lightning storm and you will know what I mean. You can test your computer’s lighting abilities with 3D Mark. I suggest using 3D Mark 2001. Yes, you read correctly. The benchmark is old, but its lighting test, that uses up to 8 lights in a low poly scene is even capable of bringing high-end graphics cards to its knees. In  Cryostasis you might end up enjoying a minimum of 16 (!!!) light sources in certain levels. While this amount alone is enough to keep your computer busy rendering reflections, there’s also the small matter of shadow rendering. Those add to the amount of light calculations. Compare that to stalker. You had a sun as a light source at day and your flashlight, or even several at night, including fires. How many did you ever see at a time? Additionally to that, weapons in the game are dynamic light sources too. You don’t need a flashlight to find your way through a dark corridor. A full clip from your tommy gun will suffice.

You can see light reflections on all of the icicle shards, that break off if you destroy icicles, or hit an enemy, or ice.

Cryostasis doesn’t stop there. Having shadows helps the look of a game, but to make it look realistic, you also need reflections and Cryostasis is full of them. The game makes heavy use of specular mapping, a technique that reflects light accurately on surfaces according to their reflective abilities. You’ll see blurry sun reflections on snowy ice blocks outside the ship, while you can see sharp light spots on icicles inside.  To make things more real, specular mapping also takes the camera’s position into account. If you move, then the reflections will not remain static, but will change according to your viewpoint.  Cryostasis takes this a step further though. Not just you can change your position, but also the objects. Heat up a room and the ice on the walls will melt. It starts to run down the wall and while it does, the light reflections on it change according to its movement and to its surface structure and thickness. This is a unique effect that no game has showcased so far and it looks extremely realistic.

When you change your position towards objects, that can reflect light...

...then the reflection changes according to your position, while taking the objects surface structure into account.

Another example. You can see how the surface structure of the ice creates an uneven, realistic look of the light reflection...

...and as soon as you move, your position relative to the ice will add another factor to change the reflection. That's as realistic as a game has gotten so far.

The player's position does not just change the appearance of the reflection...

...but also its intensity. Although hard to capture in an image, this side view of the above image is more gloomy. The effect is very apparent within the game itself.

Another clear example how your viewpoint has an influence on the lighting. In this picture the reflection is visible on the right side of the icicles...

...and when you move to the left, it moves with you.

Here it gets really realistic. The ice starts to melt and runs down the wall. May be the most famous effect of the game.

However, I haven't seen a review that paid attention to the reflection change on the moving ice.

It's not only the structure of objects that change the reflection, but also bumpmapped textures, such as those of the ice.

Another major performance factor is the shading, which allows for beautifully looking ice-encrusted walls and doors. Since the game mainly uses these looks, the performance drain just by standing in a room filled with ice is quite heavy.  Of course you also have to take high-resolution textures into account. Enemies are being graced with 4 Megapixel textures (that’s 2048×2048), while objects can have anything from 256×256 up to 1024×1024 and that’s quite some amount of data to handle.

The other Rumor:  Missing multi-core support

I don’t know, who actually started that rumor, or even why, but someone told me, that even the official forum admin of the publisher 1C admitted that the game had no multicore support. Let’s think about that for a second. The assumption alone that a game by a developer, who was capable of pulling off effects that surpass those of current gen games in certain aspects of visual quality, doesn’t honor hardware trends that have been available for half a decade (multicore CPUs were announced in 2004), sounds pretty stupid. In fact, it’s as asinine as it can get. I first stumbled upon this rumor on wikipedia, where it was “proven” with two links that lead to the publisher’s forum, where people bashed the game with their screenshots of proof. Those screenshots showed the workload of the CPU-cores during gameplay and for many of them it was rather low. Those people claimed that if the CPU would have been used more, their framerate would go up. Actually, I have a hard time understanding how someone who practically knows nothing about how the computer components work together, can make such a claim. It’s like telling your doctor how to treat your sickness: “Cut my leg! The loss of blood will help me getting rid of my headache”

The most work is done by the graphics card. However, it is dependend on the CPU to get data to process. Think of it as a dish washer in a restaurant. Someone has to do the main work, which is washing the dishes, but someone also has to bring them to the dish washer. If you have a fast dish washer, then you also need someone who can bring in the dishes fast enough, or he’ll get bored and has to wait. This equals computers with strong graphics cards, but with underpowered CPUs. What you’ll get is a maxed out CPU, while your graphics card would still have some potential to pump out more performance.

However, if your dishwasher has to deal with ages old, encrusted dirty dishes he might slow down to a crawl and the guy who delivers the work to him might get some spare time. The effects in Cryostasis do just that. They are so overwhelmingly massive in comparison to any other game out there, that your graphics card will be maxed out, even if it is a most recent model. Since it can’t take any more work, the CPU cuts back on delivering more data and gains some idle time. So in cases of those people with extremely low core workloads (may be 20% in the shots I saw), it just means that their cards are already maxed out, while their CPUs could most likely also power stronger graphics cards.

Also, there is a clear indicator, whether an application supports multiple cores or not: If it is limited to only a single core during runtime, then it is not multithreading capable. However, the screenshots of those who shunned the game showed equal load balancing among their cores. In other words, their graphical proof disproved what they wanted to prove.  The fact that an admin of the publisher supposedly fueled these ignorant assumptions just shows that you can’t trust even public sources, that should have better access to the developers than you do. For those that I confronted with the above technical facts it didn’t take long to develop new theories to blame the game for their laggy gaming experience. All of a sudden, the performance patch must have had introduced this feature, because suddenly they felt that the game ran better after installing it with better maxed out CPUs.

Well, I tested all retail versions right out of the box and all supported multithreading equally well. In fact I made a screenshot of my CPU-workload while playing the first level of the unpatched russian version (the oldest version there is) and my CPU was maxed out by 75%:

Both cores are clearly in use, even in the oldest unpatched version of the game. Only the balancing is a little off. This was shot with all settings turned down, to show a clear CPU-usage. Maxing the game out pushes the workload on all cores so far down, that the balancing isn't really a problem. With the performance patch, the CPU-usage became even less important for users with compatible PhysX-hardware.

You can check it yourself. Keep in mind that your CPU-load will be higher if your graphics card has more headroom, so turn off all effects. On the other hand, you can see that your card is already maxed out by raising everything to max. This will minimize your CPU-load. Some claimed that you’d need one of the strongest CPU-cores to play the game smoothly, but since it’s the graphics card that is bottleneck in modern games, any CPU will do. In fact, Cryostasis recommends a rather old and weak CPU for recommended specs. This is no surprise, since – as I already said so many times – it has little to do, because most calculations are performed by the gfx-card. CPU-load is even lower for those, who can make use of hardware-physics, because that will take the biggest load off the CPU and put it on the graphics card.

As you can see, Cryostasis needs a strong computer to be enjoyed in all its visual glory, but it is not flawed. It’s just going one step further, just as many titles before it did. Doom, Quake, Quake 2, Quake 3, Doom 3, Half-Life 2, Far Cry and Crysis have all marked big changes in visual brilliance, whether this was achieved by higher polygon counts, or additional visual effects. They all required state of the art computers to be played maxed out, but they also allowed to be played on smaller systems with toned down settings. Cryostasis is no different. It pushes the envelope as far as physics and particle rendering goes, but it can still run on old computers. You just have to know your way around hardware. Just for shits’n giggles I ran the game on my old Radeon x1600, which is worlds below the actual minimal requirement that the manual states (x1900, that one is a lot stronger than an x1600. Check the technical specs). I was able to run the game with all settings turned down at 640×480 and framerates never dropped below 25fps.

If you have a high-end gaming card and can’t play the game, then don’t blame it, but understand it. You want more from the game than your hardware can deliver. The fact that you spent 200, 400, or even 800 dollars on a graphics card, doesn’t mean that you can play any game without limitations. Cryostasis offers room for visual improvement for future graphics card generations (or at least the next one) and that’s a good thing to keep a game fresh. If I had the choice of choosing between a game that was toned down, just so that high-end gamers could get a warm feeling, because they could set everything to max and still drool over smooth framerates, or a game that offered a much bigger potential for visual quality and allowed me to tone it down to a level of playability that my hardware could handle, then I’d always go with the latter.

Take Half-Life 2 as an example. When it came out, it was considered to deliver jaw-droppingly beautiful graphics, but those were surpassed soon enough due to the game’s low texture resolution. Now you can use the cinematic mod, which replaces the mushy textures with high-res multimegapixel ones, giving the game a much newer look. Wouldn’t it have been better if HL2 had come with high-res textures in the first place? Sure you can use the Cinematic Mod, but how many games actually have third party texture packs for them? Doom 3 was advanced enough to offer a texture mode that requires 512MB graphics cards. Unbelievable when it came out, but very nice for gamers 5 years later. Slent Hill 3 allows texture rendering according to the native support of your graphics card. When it came out in 2003, that limit was 2048×2048 for many, but of course no one could play it like that. On current cards, you can quadruple that resolution to 4096×4096 and enjoy visuals that no one could have enjoyed 6 years ago.

So the problem is not Cryostasis, but rather a psychological problem of gamers thinking that they need certain features in a game (like AA), or that a game has to deliver a similar performance, just because others do. It’s all in your head. Get over the cliche that games without AA and AF don’t look good. Check out the game’s options for yourself. Know your hardware’s limitations. Especially the latter is a massive bottleneck, in both the gamer’s head and for the computer.  For example, I encountered numerous people using older high-end cards, who complained about the game’s performance. As it turned out they used the highest texture quality with 512MB Ram cards. I already mentioned above that the game uses very high resolution textures. A single enemy texture can weigh in at 15MB not counting scripts for AI, engine and rendering. Consider other high res textures for your surroundings. How many do you think a level has? How many do you think you can fit into 512MB of VRAM? Do you still think you should play the game with that setting?

Fact is, the game runs…and it runs well. Start out with minimum values and raise only those that you absolutely want. If a Radeon x1600 is capable of running the game, then any other newer card should too, do you actually want to see your gaming rig put to shame by 4 year old mid-range hardware?



%d bloggers like this: