DX 11 benchmark for ATI guys - Page 3

RE3FRK · 10-23-2009, 18:01 | posts: 18 | Location: UltimateSweeper

I hope directX11 does a bit better than what 10 was allowed to do as 10 for me was a peak that was never really established except mostly in BIOSHOCK. So I guess 11 has to hit the gas cause I'm on Premium atm and those 5870's look like they need some serious floor grinding.

IPlayNaked · 10-23-2009, 18:05 | posts: 6,559

Just gonna copy paste out of the main thread for everyone.

Some more for you guys at 1024x768 no AA no AF. The rest of the eye candy is maxed though.

One GPU DX11: 35.6 // 896.
Two GPU DX11: 69.7 // I don't have the score anymore, but the FPS is right.

One GPU DX10: 59.5 // 1499
Two GPU DX10: 111.2 // 2802

So actually, at that resolution with no AA, no AF it actually scales pretty much perfectly. 90% or so?

mikeysg · 10-23-2009, 18:06 | posts: 1,790 | Location: Babylon 5

Quote:

Originally Posted by Axil

my result with tesselation

without tesselation I get
a score of 2779
average fpd = 110.3

Man, you almost gave me a heart attack, I thought my rig's not configured right since a single HD5870 equals my score

, then I saw the res - 1024x768 Windowed. Before I saw this, I was thinking your i7 had something to do with your outstanding score.

gerardfraser · 10-23-2009, 18:15 | posts: 1,114 | Location: Canada

Hey I have a DX10 card.Can some post a DX11 screenshot of road,roof,dragon so we can see what DX11 brings with 50% FPS cost.

EDIT:NVM Ertard post #34 has posted some nice screens of tess at work.

lorikano · 10-23-2009, 18:18 | posts: 97

Really good .Will wait and see CF

lorikano · 10-23-2009, 18:20 | posts: 97

Quote:

Originally Posted by mikeysg

Here's my score -

Awesome scaling in CF

SolidBladez · 10-23-2009, 18:25 | posts: 3,721 | Location: CO

Love the demo, a lot more than Vantage *ew*.

Pablo · 10-23-2009, 18:27 | posts: 1,582 | Location: UK

Failed to launch on my 8800gtx.

Will sort out issue when get more time.

nascasho · *Last edited by nascasho; 10-23-2009 at 18:49.*

OC'd back to 3.8Ghz and the cards are still stock 850/1200, but now in CFX. At least it scales well with CFX enabled... microstutter central though... :/

DX11 run:

Unigine
Heaven Demo v1.0
FPS: 52.1
Scores: 1313

Hardware
Binary: Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system: Windows 7 (build 7100) 64bit
CPU model: Intel(R) Core(TM) i7 CPU 965 @ 3.20GHz
CPU flags: 3184MHz MMX SSE SSE2 SSE3 SSSE3 SSE41 SSE42 HTT
GPU model: ATI Radeon HD 5800 Series 8.670.0.0 1024Mb

Settings
Render: direct3d11
Mode: 1920x1080 4xAA fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

This time 950/1240Mhz:
DX11 Run:

Unigine
Heaven Demo v1.0
FPS: 56.2
Scores: 1417

Hardware
Binary: Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system: Windows 7 (build 7100) 64bit
CPU model: Intel(R) Core(TM) i7 CPU 965 @ 3.20GHz
CPU flags: 3184MHz MMX SSE SSE2 SSE3 SSSE3 SSE41 SSE42 HTT
GPU model: ATI Radeon HD 5800 Series 8.670.0.0 1024Mb

Settings
Render: direct3d11
Mode: 1920x1080 4xAA fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

Unigine Corp. © 2005-2009

isidore · 10-23-2009, 18:59 | posts: 5,597 | Location: Timisoara, Romania

Quote:

Originally Posted by Axil

my result with tesselation

without tesselation I get
a score of 2779
average fpd = 110.3

lool, are you serious? the result should have been the other way around, i mean tessellation was made for fps improvement not the other way around.

IPlayNaked · 10-23-2009, 19:04 | posts: 6,559

Quote:

Originally Posted by isidore

lool, are you serious? the result should have been the other way around, i mean tessellation was made for fps improvement not the other way around.

They used it for IQ improvement though. Substantial improvement on all the rocks, all the roofs. If you had put the same effect with dx9 or 10, perhaps it would be significantly slower.

morbias · **Memory**: 6GB HyperX @1600 6-7-6-18

Quote:

Originally Posted by isidore

lool, are you serious? the result should have been the other way around, i mean tessellation was made for fps improvement not the other way around.

If you tried doing the same thing without using tesselation the performance hit would be massive. There is a hit with tesselation enabled, but nowhere near as big as it would otherwise be.

John Dolan · 10-23-2009, 19:28 | posts: 2,161 | Location: uk nottingham

Quote:

Originally Posted by vbetts

From what I've seen, this benchmark doesn't scale well. Now don't know if it scales bad on both crossfire and sli, but what I've seen of it crossfire..

Edit-Another pattern I've noticed. This is a benchmark for the video card, but from what I've seen it's also very cpu heavy too.

First PC- John Dolan
5870 1gb
qx6800 stock
p43
8500 4gb

Second PC- Miller
4870x2 2gb
q6600 3.6ghz
p5q-e
8gb ddr2 800

Now in terms of numbers this might not be the best comparison, so something that would be needed to determine how this software finds the final benchmark score, is have 2 people with the same gpu, say the 5870 or something, but one have like an i7 and the other have say a q9440. It's obvious that the benchmark will take into consideration the rest of the system, but by how much though? Settings for test if 2 people decide to.
Settings
Render: direct3d11
Mode: 1920x1080 4xAA fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

a i7 would score fps in the 50s
a 5870 needs a i7 @ 3.6 to have no cpu bottleneck so my old kentfield stock @ 2.93 will hurt my fps in most if not all games and benchmarks

ibitato · **Memory**: 4x2GB@1080 5.5.5.15 5:6

My run, DX10 native resolution 1920x1200 8xAA 16xAF all enabled / high

For me it's scales like a charm, cats 9.8

John Dolan · 10-23-2009, 19:44 | posts: 2,161 | Location: uk nottingham

This is a very good benchmark imo.Puts 3dmark to shame.

ibitato · **Memory**: 4x2GB@1080 5.5.5.15 5:6

Quote:

Originally Posted by John Dolan

This is a very good benchmark imo.Puts 3dmark to shame.

yep, it is a full 3d environment (heck, you can even move around in there) than just a rendering engine.
It's most like a game engine itself....

neosyoshiran · **only how referecen -** 10-23-2009, 20:33 | posts: 7 | Location: sucre-bolivia

Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

vista 64 sp2
xfx geforce 275, core 2 quad 2.6 8 gigas in ram ddr2
dx9 1680x1050 0xAA fullscreen Fps:47.9 ; 1205
dx9 1680x1050 4xAA fullscreen Fps:37.7 ; 951
dx9 1680x1050 8xAA fullscreen Fps:?? ; ?? testing
dx10 1680x1050 0xAA fullscreen Fps:45.1 ; 1136
dx10 1680x1050 4xAA fullscreen Fps:34.5 ; 870
dx10 1680x1050 8xAA fullscreen Fps:20.9 ; 527

opengl 1680x1050 8xAA fullscreen Fps:18.5 ; 465

isidore · 10-23-2009, 21:18 | posts: 5,597 | Location: Timisoara, Romania

Quote:

Originally Posted by morbias

If you tried doing the same thing without using tesselation the performance hit would be massive. There is a hit with tesselation enabled, but nowhere near as big as it would otherwise be.

yeah i guess you're right..it looked very good from the video, even the little hose on that metal thing had tessellation on it.

mudkips · 10-23-2009, 21:35 | posts: 75

Quote:

Originally Posted by morbias

If you tried doing the same thing without using tesselation the performance hit would be massive. There is a hit with tesselation enabled, but nowhere near as big as it would otherwise be.

I can understand that but regardless of how the tessellation is being done its causing a near 50% performance penalty via hardware. I assumed this was going to happen and I am sure when the 2nd gen DX11 cards are released performance will be adequate for all that eye candy. Tessellation definitely isn't "free".

Alkerion · **Memory**: 12GB Corsair 7-7-7-20

Did you find a way to set the res in 2560x1600 in Heaven Demo ?

Also possible to change the resolution of the AMD tech demo ???

TDurden · *Last edited by TDurden; 10-23-2009 at 23:50.*

WHY is that that DX9 is MUCH faster in this benchmark than DX10?
Unigine benchmarks were never reliable.

edit: it seems aa in dx9 does not look as good as in dx10. looks like dx10 uses transparent aa for vegetation too while dx9 does not.

Mr.StanleyDudek · 10-23-2009, 23:27 | posts: 343 | Location: Denton, TX

IDK, but looking at everyone else, this isn't half bad.

--------

Unigine

Heaven Demo v1.0

FPS: 56.1
Scores: 1414

Hardware

Binary: Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system: Windows 7 (build 7100, Service Pack 2) 64bit
CPU model: Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
CPU flags: 3399MHz MMX SSE SSE2 SSE3 SSSE3 HTT
GPU model: ATI Radeon HD 3870 X2 8.670.0.0 512Mb

Settings

Render: direct3d11
Mode: 1280x1024 fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 16x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

Unigine Corp. © 2005-2009

Zeroko · 10-23-2009, 23:37 | posts: 53 | Location: United Kingdom

Unigine
Heaven Demo v1.0

FPS:
24.0
Scores:
605

Hardware
Binary:Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system:Windows Vista (build 6002, Service Pack 2) 64bit
CPU model: Intel(R) Pentium(R) D CPU 3.00GHz
CPU flags:2999MHz MMX SSE SSE2 SSE3 HTT
GPU model:NVIDIA GeForce GTX 260 8.16.11.9156 896Mb

Settings
Render:direct3d11
Mode:1360x768 8xAA fullscreen
Shaders:high
Textures:high
Filter: trilinear
Anisotropy:16x
Occlusion:enabled
Refraction:enabled
Volumetric:enabled
Unigine Corp. © 2005-2009

-HaRdCoRE- · 10-24-2009, 00:30 | posts: 163 | Location: Burlington, Vermont

My results: Windows7x64, E8400@ 4.3, 2x 4890's @ 950/ 1100

Everything on high, 1920x 1080 8xAA

avg fps: 42.1
score: 1061

same thing no AA, res 1024x768

avg fps: 119.2
score: 3002

No process's killed, internet security up and running

psyside · 10-24-2009, 02:14 | posts: 324 | Location: Macedonia

Quote:

Originally Posted by IPlayNaked

One GPU DX10: 59.5 // 1499
Two GPU DX10: 111.2 // 2802

So actually, at that resolution with no AA, no AF it actually scales pretty much perfectly. 90% or so?

Yes this game/demo seems to scale very very good! it comes to my mind that ati gave 58xxx cards to work on the engine