Slashdot Mirror
Unigine's Newest Benchmark Features Huge, Open-Space Expanses
jones_supa writes "Unigine announced a new GPU benchmark known as Valley Benchmark. From the same developers who created Heaven Benchmark, the Valley Benchmark is a non-synthetic benchmark that is powered by the Unigine Engine, a real-time 3D engine that supports the latest rendering features. The Valley Benchmark includes massive area of 64 square kilometers of very detailed terrain that includes forest, mountains, green expanses, rocky slopes and flowers. The area can be freely explored by means of walking or flying. All major operating systems are supported."
Are here under Fedora.
I'm starting to think GNU is the problem with "GNU/Linux" these days.
All they have to do is rename it The Elder Scrolls VI and they have themselves a finished game.
Better known as 318230.
-Instancing: For drawing lots of trees without draw call overhead.
-Impostors: For the groups of trees far away
-Vertex Program: For the sway of the trees, probably with per vertex amount of strength
-PSSM: For the shadows
-Godrays: For the sunrays through the trees
-HDR+Bloom with luminance bleeding: For the lighting and skybox
-Instanced Particles: For the clouds
I sure am forgetting some of them, but I think this demo, with huge amounts of instancing, is mainly designed to stress the vertex pipeline of modern videocard.
http://burnedchips.com/results.jpg
not to shabby for a 2009 budget box
-Godrays: For the sunrays through the trees
Nitpick: that's not a technique. Those rays of light are called godrays, it says nothing about the implementation technique.
I sure am forgetting some of them, but I think this demo, with huge amounts of instancing, is mainly designed to stress the vertex pipeline of modern videocard.
I checked out the YouTube - video and, well, I see huge amounts of people complaining about the apparently-poor texture resolution of this benchmark. IMHO, these people are missing the whole point of the demo as the demo is not intended to show exceedingly impressive textures or such. The speeds at which the engine can manage to do so beautiful real-time shadows and lighting, huge, open landscape with loads of foliage, the impressively realistic fogging in certain areas and so on, these are the focus here. I certainly would trade some texture resolution for more realistic lighting and environmental effects in games if it ever came to such a choice.
I think this demo, with huge amounts of instancing, is mainly designed to stress the vertex pipeline of modern videocard.
It actually uses quite a lot of LOD -- even at the highest settings there aren't ever very many triangles on screen. As TFS says, this isn't meant to be a synthetic benchmark. It's not made to stress any one specific thing, and it really doesn't.
Some of the tech it is demoing is pretty cool, even if the resulting image isn't very impressive. In the hands of proper designers, this stuff could be awesome.
Impressive enough, it's always neat to see even programming as utterly performance intensive as a game engine pay off with something so pretty.
A bit of an oversight - it won't run if security is set to reject unknown developers.
Workaround is to right-click on the icon and choose "Open Application" from the popup menu.
I don't care if it's 90,000 hectares. That lake was not my doing.
> Nitpick: that's not a technique. Those rays of light are called godrays, it says nothing about the implementation technique.
Indeed. They could be using "Volumetric Light", "Occlusion Stencil", or as a post-process in Screen Space. Hard to tell which algorithm they are using.
Reference:
* http://http.developer.nvidia.com/GPUGems3/gpugems3_ch13.html
It's propably you wo misses the point. The idea is to bench a realistical workload. Since techniques for replacing far away objects are used in almost every game it makes perfect to include it. Besides, you could not render such scenes without it in real time
Very nice. But I'm really curious: why do video sims have this obsession with pretending to be shot on a physical camera (e.g. rain drops on the "lens", lens flare when looking at the sun)?
I understand it's an aesthetic but it comes over as insecurity: "hey I bet you couldn't tell this wasn't a real camera in a real world". I think sim designers should have more confidence and get over this 'trying to prove we're as good as the real world by simulating failings in cameras'. I think there's some really nice work and they should concentrate on improving their presentation of world rather than trying to reverse engineer the failings of old cameras.
Rain drops on the lens from video shot on real cameras is really annoying. Don't spend energy trying to simulate it, or lens flare. Spend your time improving your new format, do cool stuff the real film makers can't do and take advantage that you're not bound by their limitations. Please don't work on a virtual camera operator's hand cleaning your virtual lens with a virtual disposable tissue when it rains hard....
The thing that impresses me the most is the way that whoever put it together actually knows some geology and biology. The trees, other plants, and rocks are realistically placed. I admit it's kind of geeky, but as a geologist it always bothers me when game designers think any old "random pile of rocks" or "randomly bumpy cliff surface" corresponds to the way geological materials behave in the real world. Same for the shape of mountainsides. They are not randomly steep, planar slopes. Most of them have a graceful exponential kind of curvature. There are similar issues for the distribution of real plants and trees.
If the whole point of a game is to immerse you in an alternate reality, but everything in the "natural" world looks (to the experienced eye) like the building equivalent of walking through a funhouse, it kind of spoils the effect. These people are meticulously observant of nature and actually know what they are doing! Kudos.
Its normally not an issue but when you get a program from a 3rd party such as a mirror or from a torrent, which you would know is exactly this case with this benchmark if you bothered to try to download it, then its nice to have evidence that the program was packaged up using the keys of the developer even if there is some small chance that someone other than the developer might have gotten the keys.
Makes your argument seem a bit silly now, doesnt it?
"His name was James Damore."
I believe the terrain was pulled from GIS data, (the article mentions the creators wanting to show off some of Russia's natural beauty) and the article says they procedurally placed the foliage and rocks, so Its all the work of very well designed math problems. But you are right, they did seem to do a meticulous job getting those math problems to reflect natural patterns.
I've decided to Diversify my Holdings. I've divided my cash between my left and right pockets, instead of all in one.
Rain drops on the lens from video shot on real cameras is really annoying
Heh. You clearly don't wear glasses. Raindrops on the lens is absolutely normal to me, so it doesn't seem even slightly out of place when I see it in a game.
I agree that lens flare is annoying, though.
It still gets a lot of pop-up though. Most engines seem to have this problem where everything far away is okay and every up close is okay, but in the middle ground things just pop into existence and look terrible.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
I did this landscape by a process called "procedural generation":
https://lh4.googleusercontent.com/-lPWscdDzXM8/TjtfFx12uRI/AAAAAAAACqc/gIukXZ8t0Sc/s1600/TerrainFracta+2l.jpg
Rather than place hills and grass individually, it uses fractal formulas to create the shapes and textures. The formulas and textures are height and slope aware, and it uses atmospheric haze to give a distance effect. The software is E-on Vue, which is used in professional movie making, but I just diddled around with it for fun.