Thursday, February 14th 2013

Unigine Announces Valley GPU Benchmark

Valley Benchmark is a new GPU stress-testing tool from the developers of the very popular and highly acclaimed Heaven Benchmark. The forest-covered valley surrounded by vast mountains amazes with its scale from a bird's-eye view and is extremely detailed down to every leaf and flower petal. This non-synthetic benchmark powered by the state-of-the art UNIGINE Engine showcases a comprehensive set of cutting-edge graphics technologies with a dynamic environment and fully interactive modes available to the end user.

Offering a rare chance to experience a breath of untapped, crystal clear air, Valley Benchmark allows you to encounter a morning high up in the mountains when the snow-capped peaks are just barely glittering in the rising sun. Be it flying over the vast green expanses or hiking along rocky slopes, this journey continues as long you wish. Unique in every corner, this open-space world provides a wonderfully relaxing experience under the tranquil music and sounds of nature.
DOWNLOAD: Unigine Valley for Windows, Mac OS X, Linux

"Living in Siberia, we wanted to show how beautiful the nature is here. It is full of contrasts, and flowers can grow through the deadfall right next to brutal mountains," said Andrey Kushner, lead technical artist at UNIGINE Corp. "It was an interesting challenge to create this huge, yet detailed world. Moreover, our engine is so flexible that we could place all objects procedurally and recreate this valley with photorealistic graphics."

Features
  • Extreme hardware stability testing
  • Per-frame GPU temperature and clock monitoring
  • Multi-platform: Windows, Linux, Mac OS X
  • Advanced visual technologies: dynamic sky, volumetric clouds, sun shafts, DOF, ambient occlusion
  • 64,000 square kilometers of extremely detailed, seamless terrain
  • Procedural object placement of vegetation and rocks
  • The entire valley is free to be explored in interactive fly-by or hike-through modes
  • User-controlled dynamic weather
  • Support for stereo 3D and multi-monitor configurations
  • Benchmarking presets
  • Command line automation support
  • Highly customizable reports in CSV format
Alongside a completely free Basic edition, Valley Benchmark provides in-depth performance reviews in the Advanced and Pro editions for hardware manufacturers, graphics driver developers, industry professionals and all individuals involved with video card stability testing. For more information, visit the product page.
Add your own comment

56 Comments on Unigine Announces Valley GPU Benchmark

#1
erocker
by: BigMack70
Did you see that draw distance?
The "pop in" that I'm seeing is with hard shadows and with flowers/ground objects. The actual draw distance of the scenes is huge.
Posted on Reply
#2
BigMack70
by: erocker
The "pop in" that I'm seeing is with hard shadows and with flowers/ground objects. The actual draw distance of the scenes is huge.
Basically everything that's not clouds, a mountain, or a 2D sprite has nasty pop in
Posted on Reply
#3
erocker
by: BigMack70
Basically everything that's not clouds, a mountain, or a 2D sprite has nasty pop in
Yes, the two things I listed.
Posted on Reply
#5
devguy
Demo doesn't let me run at my native 4800x900 like Heaven does (maxes out at 4096 in vertical res). The multiple monitor option is giving me trouble. I got around 32fps average at 1600x900 full screen on one monitor.

Catalyst 13.2 Beta 3 drivers used on Radeon HD 7950 with stock clocks.
Posted on Reply
#6
fullinfusion
1.21 Gigawatts
Yeah my 6990 and 1000Mhz /1500 mem dont work at all on this bench!

A patch will fix this
Posted on Reply
#7
niciuffo
Is there a dedicated thread where you can submit your result? I can't seem to find it anywhere.
Posted on Reply
#8
rickss69
Here is the Q9550 at work...



A small bump to the cpu/gpu's -



Posted on Reply
#9
rickss69
by: niciuffo
Is there a dedicated thread where you can submit your result? I can't seem to find it anywhere.
Feel free to start one...if you want to keep up with it. :p
Posted on Reply
#10
BiggieShady
by: RejZoR
Instead they churn out idiotic ammounts of polygons on that mountain on the other end of the map and leave stuff close tot he viewport ugly as hell. Makes no sense and they all freakin do this. In all games and all benchmarks.
More importantly, they often use only frustum culling and no occlusion culling at all, so CPU makes unnecessary draw calls for occluded objects.
Although it is easy to integrate displacement mapping (tessellation) in LOD system, art production should be zbrush/mudbox style rather than polygon modelling - so if you don't have original object in extreme poly-count version, tessellation is not as straightforward.
Pop-ins should not exist by today's standards, I've seen LOD systems with smooth transitions (and i don't mean that pixelated dissolve shader ubisoft uses)
Posted on Reply
#11
niciuffo
by: rickss69
Feel free to start one...if you want to keep up with it. :p
Eheh, that's the problem :) I just wanted to know if there was some kind of thread where everyone could post their scores, I am pretty sure I've seen a couple of them in various forums but I know they require a lot of work :p
Posted on Reply
#12
patrico
one of the nicest benchmarks ive seen in ages :)
Posted on Reply
#13
bim27142
Here's mine... Beautiful indeed... ;)

Posted on Reply
#14
Vinska
Here's mine. On Linux ;)
Too bad I cannot properly bench it @ 1920x1080 without triggering a [probably a bug in the driver]. *shakes fist*
So I ran it on 1600x900 and 1280x720. "Deal with it, nerds." -Fork Parker

@1280x720


@1600x900


P.S. Yesh. Beautiful. Sux I only noticed this news post "leik, half an hour ago".
Posted on Reply
#15
Prima.Vera
What's up with the scores from nVidia? Even the 460s are faster than the latest 7800 cards...
Posted on Reply
#16
rickss69
by: Prima.Vera
What's up with the scores from nVidia? Even the 460s are faster than the latest 7800 cards...
Don't understand the question. The dual 7800's in post # 24 beat my 460's...all the rest of the 7800 runs are single card.
Posted on Reply
#17
HammerON
The Watchful Moderator
CPU=4.4GHz Mem=2004MHz GPU(s):1125/1375
1920x1080 one 7970:

1920x1080 two 7970's:


2560x1600 one 7970:

2560x1600 two 7970's:


All four tests report 3x GPU, but this is false.
Posted on Reply
#18
rickss69
by: HammerON
CPU=4.4GHz Mem=2004MHz GPU(s):1125/1375
1920x1080 one 7970:
http://img.techpowerup.org/130217/Capture143.jpg
1920x1080 two 7970's:
http://img.techpowerup.org/130217/Capture144.jpg

2560x1600 one 7970:
http://img.techpowerup.org/130217/Capture145.jpg
2560x1600 two 7970's:
http://img.techpowerup.org/130217/Capture146.jpg

All four tests report 3x GPU, but this is false.
That's odd...are they all installed in the mb and you are using jumpers to disable?
Posted on Reply
#19
Vinska
by: Vinska
Here's mine. On Linux ;)
[...]
@1280x720


@1600x900

For comparison, I re-ran it with the same settings on Windows in OpenGL and D3D11 modes.
As expected, D3D11 ran noticeably faster than OpenGL - OGL scored ~17% less on Windows and ~15.5% less on Linux compared to D3D11.
*wonders it that mostly the driver's or the benchmark's fault. Or both.*
*bets on the driver being less optimized for OGL.*

And alas, just like in some previous benchmarks I did, on Linux, OpenGL is faster than on Windows. Although, the gap is much smaller w/ 313.xx drivers - Windows driver IS catching up. (The gap was rather horrible before. Don't ask.)

The results:

@1280x720:
In OGL:

In D3D11:


@1600x900
In OGL:

In D3D11:
Link; thumb broken

P.S. I don't even know why I am actually posting those, knowing that most of You probably !care about this completely.
Posted on Reply
#20
BigMack70
by: Vinska
For comparison, I re-ran it with the same settings on Windows in OpenGL and D3D11 modes.
As expected, D3D11 ran noticeably faster than OpenGL - OGL scored ~17% less on Windows and ~15.5% less on Linux compared to D3D11.
Don't you mean DirectX 11? :roll:
Posted on Reply
#21
Easo
Please dont bash it so hard, I am sure they will fix it. Its still very very pretty.
If they dont, go ahead, i will join you with pitchfork.
Posted on Reply
#22
Vinska
by: BigMack70
Don't you mean DirectX 11? :roll:
NO.
Read this:
by: Vinska

[...]
/* meanwhile... */
To some of You: please, stop saying "DirectX" when it comes to graphics only. DirectX is an API that supplies a very wide range of services. Including, but not limited to: input, sound, networking, graphics, AV decoding, and so on. And the part that deals with graphics is Direct3D. Thus, when talking about the graphics portion of DirectX, please, say "Direct3D".
[sarcasm]...unless maybe Your GPU can do hardware acceleration on mouse input, for example. Which is part of DirectX, but not part of Direct3D.[/sarcasm]
(yes, I am quoting myself :banghead:)
Posted on Reply
#23
MetalRacer
Here are two runs I made for comparison, it looks like CPU clocks and thread count makes little difference in final score.

4.5 ghz HT off.



4.8 ghz HT on.
Posted on Reply
#24
Shihabyooo


Sry for low res pic.

by: Vinska

P.S. I don't even know why I am actually posting those, knowing that most of You probably !care about this completely.
Give it a couple of years, everyone will ;)

I see that the minimum FPS is slightly better on the Linux side while the max goes the other way around. More love for Pilediver on the free side?
Posted on Reply
#25
Vinska
by: Shihabyooo

Give it a couple of years, everyone will ;)

I see that the minimum FPS is slightly better on the Linux side while the max goes the other way around. More love for Pilediver on the free side?
Indeed.
Benchmarks I did (unpublished publicly) shows both Bulldozer and Piledriver get much more love under Tux' wing. And as *better not tell who* commented, that this is also related to microcode issues that are taken care of while on Linux, along with the "usual" Win7 scheduling issues. Also, there's one more possible BD&PD-related issue on Windoze I'd better not speak about.

P.S. On Win7, CPU scheduling is simply horrible when it comes to switching between active tasks.
Posted on Reply
Add your own comment