Unigine DX11 benchmark

[nitromullet]

1920x1200 4xAA/16xAF

Unigine
Heaven Demo v1.0
FPS:
25.6
Scores:
644
Hardware
Binary:
Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system:
Windows 7 (build 7232) 64bit
CPU model:
Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
CPU flags:
3337MHz MMX SSE SSE2 SSE3 SSSE3 SSE41 SSE42 HTT
GPU model:
ATI Radeon HD 5800 Series 8.661.0.0 1024Mb
Settings
Render:
direct3d11
Mode:
1920x1200 4xAA fullscreen
Shaders:
high
Textures:
high
Filter:
trilinear
Anisotropy:
16x
Occlusion:
enabled
Refraction:
enabled
Volumetric:
enabled
Unigine Corp. © 2005-2009

1024x768 4xAA/16xAF

Unigine
Heaven Demo v1.0
FPS:
42.3
Scores:
1065
Hardware
Binary:
Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system:
Windows 7 (build 7232) 64bit
CPU model:
Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
CPU flags:
3337MHz MMX SSE SSE2 SSE3 SSSE3 SSE41 SSE42 HTT
GPU model:
ATI Radeon HD 5800 Series 8.661.0.0 1024Mb
Settings
Render:
direct3d11
Mode:
1024x768 4xAA fullscreen
Shaders:
high
Textures:
high
Filter:
trilinear
Anisotropy:
16x
Occlusion:
enabled
Refraction:
enabled
Volumetric:
enabled
Unigine Corp. © 2005-2009

...not sure why it shows my cpu at 3337Mhz, BIOS and cpu-z say otherwise.

 

[akugami]

Originally posted by: Idontcare
What's up with the intentional "nearsighted" vision effect where everything in the foreground is crisply in focus like the viewer has 20/20 vision but then everything in the background, but still perfectly within resolvable eyesight, its like the viewers vision goes to hell as if they are https://www.google.com/health/ref/Nearsightedness">nearsighted</a> and need glasses to see things in the distant?

Is that supposed to be a cool effect in games these days? I find it annoying. The rest of the demo where they let you see things as if your vision is just fine is pretty cool.

I noticed that too and thought it was way overdone as well. Used sparingly to lightly blur the background would be ok but that was way way overdone. The other issue is that this demo is kind of limited in how large the "world" is.

I'd have been more interested if the sandboxed environment was larger so you can see how performance would be in a game that had a scene where you stare across some plains, perhaps a couple of farms with animal and farming activities happening and all the way to some mountains way off in the background.

As it is, the environment is very small and there really aren't any stuff that is being rendered far in the background nor are there any active NPC/Mobs to stress the system for a better idea of how it will perform.

I likely won't be getting the GT300 or Radeon 5xx0 series because the game engines won't be out for a bit but this is very compelling stuff in terms of graphical image quality.

 

[nitromullet]

Originally posted by: ShadowOfMyself
Very impressive... Looks better than anything Futuremark has put out since 03

Alot of the scenes there look like actual movies, its hard to believe its being rendered in real time

Agreed. I remember when 3DMark actually did look like games that were going to come out in a few years. Vantage was a complete let down after playing Crysis.

The Unigine Heaven benchmark is definitely a lot more pretty than Vantage. Some of the effects are a bit over done probably to make the new tech more recognizable. Had they not done this we'd all be saying, "meh, I don't see what the big deal about DX11 is."

 

[zebrax2]

Can someone with a 5xxx series test it with and without tessellation just for performance comparison

 

[nitromullet]

Originally posted by: zebrax2
Can someone with a 5xxx series test it with and without tessellation just for performance comparison

1920x1200 4xAA/16xAF w/o tessellation

FPS:
39.2
Scores:
988


1920x1200 4xAA/16xAF w/ tessellation

FPS:
25.6
Scores:
644

edit:

I noticed this when running the no tessellation bench... It looks like the artists took tessellation into account when they made this map. It looks like for they stairs they just made a ramp with a texture that looks like stairs. Without tessellation that's what it looks like, but with tessellation they look like stairs.

Stairs w/o tessellation

Stairs w/ tessellation

...really, all the textures look better.

 

[lopri]

Thank you, thilan, for the AMD demo link. Those demos look interesting as well.

I downloaded the Unigine demo and ran it on GTX 280 and it looks stunning without tessellation. Can't imagine what it'd look like with tessellation on @2560x1600. With my E8400 and GTX 280, I got:

1920x1200 4xAA/16xAF DX10: 26.2 FPS (661 Points)
2560x1600 4xAA/16xAF DX10: 15.6 FPS (392 Points)

 

[zebrax2]

Originally posted by: nitromullet

Originally posted by: zebrax2
Can someone with a 5xxx series test it with and without tessellation just for performance comparison

1920x1200 4xAA/16xAF w/o tessellation

FPS:
39.2
Scores:
988


1920x1200 4xAA/16xAF w/ tessellation

FPS:
25.6
Scores:
644

edit:

I noticed this when running the no tessellation bench... It looks like the artists took tessellation into account when they made this map. It looks like for they stairs they just made a ramp with a texture that looks like stairs. Without tessellation that's what it looks like, but with tessellation they look like stairs.

Stairs w/o tessellation

Stairs w/ tessellation

...really, all the textures look better.

Yeah, I noticed those stairs yesterday when i watched the video.

Around 1/3rd performance hit. Thats not bad at all considering how much tessellation is used in this benchmark.

Q: Do any of you think that the extra shadows produced by the tessellated surfaces (eg. those shadows produced by the bricks when tessellated) have some major effect in performance?

Edit:
Worded my sentences better

 

[Compddd]

Why isn't it letting me choose 2560 x 1600 resolution? I'm on a Dell 3007WFP-HC I don't even see the option for my native res in these benchmarks

 

[lopri]

Here

http://img27.imageshack.us/img27/7192/unigine.jpg

Not there?

 

[MarcVenice]

Originally posted by: Compddd
Why isn't it letting me choose 2560 x 1600 resolution? I'm on a Dell 3007WFP-HC I don't even see the option for my native res in these benchmarks

I ran it at 2560 just fine. It kinda put the hammer down on a single HD 5870 though, going to try it with two of them

 

[dug777]

Just out of interest, does anyone know what the difference between the dx9 and dx10 paths is?

There's a relatively significant performance penalty using the dx10 path:

Q6600@stock/P5Q-PRO/4GB DDR2/4850 512mb: Settings - 19x12 16xAF 2xAA everything on or high that I could with a 4850 and Vista 64-bit (so no tessellation ):

dx9

Stock: 22.1fps/556 points
780 core/stock mem: 24.5fps/621 points

dx10

Stock: 20.4fps/515 points
780 core/stock mem: 23.1fps/581 points

 

[evolucion8]

API______DX9______DX10______DX11______OGL
FPS______42.9______29.1_______29.2_______33.1
Score____1081______733_______735________834

All ran at 1280x1024 with everything maxed and 8x FSAA and 16x AF.

 

[njdevilsfan87]

What I don't understand is why is there a ramp for stairs with tessellation off? It's not like tessellation is required to make proper stairs.

 

[akugami]

Originally posted by: njdevilsfan87
What I don't understand is why is there a ramp for stairs with tessellation off? It's not like tessellation is required to make proper stairs.

I think it's just to show off. Some of the effects are overdone to show the differences in the game engine between DX10 and DX11. It does create some excitement for developers because they could make a flat wireframe that would look like real stairs with tessellation. Or they could create actual wireframes for each part of the stairs.

Less time getting the game to look how you want means that resources can be diverted for story and gameplay as well as general polish.