Dx 11 benchhmark for ATI guys

Q9450
GTX280
Vista Ultimate 32bit
no OC on anything

All default settings on the benchmark except full screen and 1920x1200

heaven01.png
 
Both tests run on the sig rig and 3x SLI GTX 280's. Performance for the DX11 verson was good but a a lot choppier than the DX 10 run. Also this engine seems to love SLI. Did test single card performance but my SLI inidcators were at max in both runs. So this engine looks to be multi-gpu friendly. AA was off.

unigine_20091024_2253.jpg
unigine_20091024_2259.jpg


Looks like this engine scales VERY well with SLI comparing my results with Whisperfang. He's got nothing overclocked, my 280's are stock my QX9650 is @ 3.63. Still I'm getting almost 2.5 half times the performance with 3 280's vs his one.
 
heatlesssun said:
Both tests run on the sig rig and 3x SLI GTX 280's. Performance for the DX11 verson was good but a a lot choppier than the DX 10 run. Also this engine seems to love SLI. Did test single card performance but my SLI inidcators were at max in both runs. So this engine looks to be multi-gpu friendly. AA was off.

unigine_20091024_2253.jpg
unigine_20091024_2259.jpg


Looks like this engine scales VERY well with SLI comparing my results with Whisperfang. He's got nothing overclocked, my 280's are stock my QX9650 is @ 3.63. Still I'm getting almost 2.5 half times the performance with 3 280's vs his one.
Click to expand...

There are too many flaws in the settings with your bench.
 
heatlesssun said:
Both tests run on the sig rig and 3x SLI GTX 280's. Performance for the DX11 verson was good but a a lot choppier than the DX 10 run. Also this engine seems to love SLI. Did test single card performance but my SLI inidcators were at max in both runs. So this engine looks to be multi-gpu friendly. AA was off.
Click to expand...

Scaling was nice. I would have thought that the DX11 was faster, even though you are using DX10 cards. Did you notice any image improvements in DX11 vs. DX10 that can account for the performance loss?:)

Any reason why you chose not to use AA btw? DX11 AA is supposedly faster, so it would be nice to see if DX10 cards would benifit from it as well.
 
DX10 vs. DX11 on DX10 hardware with 8xFSAA and all options the same.





Using the DX11 API vs. the DX10 API did seem to net me a bit more performance.
 
jimbowdang said:
DX10 vs. DX11 on DX10 hardware with 8xFSAA and all options the same.
Using the DX11 API vs. the DX10 API did seem to net me a bit more performance.
Click to expand...

Thanks. :) DX11 was supposed to be faster also on DX10 hardware, so its nice to see there is some gains, though minor.
 
So for a dx11 card, just run in dx10, then run dx11 w/o tesselation to get the performance comparison between dx10 and dx11?
 
Not that easy, nVidia cards are architecturaly different. They will get different results, no matter what.
 
great benchmark..thought I'd post mine even though they are missing the full DX11 stuff:
00000.jpg

00001.jpg

00002.jpg


So far I'm quite impressed with this benchmark. Even though it's dx10, I love turning up the eye candy to 16x
 
Last edited:
Tamlin_WSGF said:
Scaling was nice. I would have thought that the DX11 was faster, even though you are using DX10 cards. Did you notice any image improvements in DX11 vs. DX10 that can account for the performance loss?:)

Any reason why you chose not to use AA btw? DX11 AA is supposedly faster, so it would be nice to see if DX10 cards would benifit from it as well.
Click to expand...


i still cant believe no one gets it.. its been said many times already in this thread.. that even when the benchmark shows a nvidia card running dx11 it automatically defaults to dx10 when the benchmark starts but will still say it was running dx11 in the final results..


seriously people.. read the entire thread before making the same comments that have been answered a dozen times already..

the card has to fully support dx11 for dx11 to be used.. you cant use a dx10 card to use some of the features in dx11.. its the same way with dx9 cards using dx10.. a dx9 card can not use dx10 in any shape or form.. and dx10 cards can not use dx11 in any shape or form..
 
Last edited:
yeah true it's not dx11..but it's stil a great benchmark that shows relative performance. Very happy.
 
sirmonkey1985 said:
i still cant believe no one gets it.. its been said many times already in this thread.. that even when the benchmark shows a nvidia card running dx11 it automatically defaults to dx10 when the benchmark starts but will still say it was running dx11 in the final results..


seriously people.. read the entire thread before making the same comments that have been answered a dozen times already..

the card has to fully support dx11 for dx11 to be used.. you cant use a dx10 card to use some of the features in dx11.. its the same way with dx9 cards using dx10.. a dx9 card can not use dx10 in any shape or form.. and dx10 cards can not use dx11 in any shape or form..
Click to expand...

To make the comparison to an nvidia card running this bench, just run a dx11 card and disable the tesellation feature. With those settings, my 5850 ran a lot faster than a gtx280, and got similar results to a gtx275 sli setup so far(when comparing results from the members in this thread using their same settings). :)
 
PS3 said:
To make the comparison to an nvidia card running this bench, just run a dx11 card and disable the tesellation feature. With those settings, my 5850 ran a lot faster than a gtx280, and got similar results to a gtx275 sli setup so far(when comparing results from the members in this thread using their same settings). :)
Click to expand...

don't think so, check settings again.:rolleyes:
 
Holy crap your cards are oc'ed. i7 too? Dang, well this bench isn't cpu dependent. I benched at your 1680x1050 16xaf no AA settings...using dx11, all high, etc. except with tesellation disabled. Probably would have gotten closer if you used AA.
 
yeah but like I said need to check again I was 8X AA aswell..lol, otherwise I get 80-100fps
 
Benched again and gtx275 sli definitely has an advantage over a 5850 - here is a good comparison image. Starting to think that a 5850x2 2GB would definitely out perform a gtx295 sli setup since it would scale a lot better then a gtx295 sli setup.

gtx275slivs5850oc.jpg
 
sirmonkey1985 said:
i still cant believe no one gets it.. its been said many times already in this thread.. that even when the benchmark shows a nvidia card running dx11 it automatically defaults to dx10 when the benchmark starts but will still say it was running dx11 in the final results..


seriously people.. read the entire thread before making the same comments that have been answered a dozen times already..

the card has to fully support dx11 for dx11 to be used.. you cant use a dx10 card to use some of the features in dx11.. its the same way with dx9 cards using dx10.. a dx9 card can not use dx10 in any shape or form.. and dx10 cards can not use dx11 in any shape or form..
Click to expand...

This is not true. DX11 have backwards compatibility with DX10 and DX10.1. Some of the features require hardware support, but DX11 is supposed to give benifits to DX10 and 10.1 cards:

The most interesting things to us are more subtle than just the inclusion of a tessellator or the addition of the Compute Shader, and the introduction of DX11 will also bring benefits to owners of current DX10 and DX10.1 hardware, provided AMD and NVIDIA keep up with appropriate driver support anyway.
Click to expand...
http://www.anandtech.com/video/showdoc.aspx?i=3507&p=3

Though DX11 introduced compute shaders on SM 5, it still supports SM4.1 and SM4. Take a look at this:

There is no available DirectX 11 hardware from Nvidia, but DX11 compute shaders version 4.0 are supported by existing DirectX 10 graphics chips
Click to expand...
http://www.xbitlabs.com/news/video/...eases_Drivers_Supporting_Compute_Shaders.html

I hope this clarified things for you, so you understand why I am interested in DX11 performance on DX10 cards. :)
 
PS3 said:
Benched again and gtx275 sli definitely has an advantage over a 5850 - here is a good comparison image. Starting to think that a 5850x2 2GB would definitely out perform a gtx295 sli setup since it would scale a lot better then a gtx295 sli setup.

gtx275slivs5850oc.jpg
Click to expand...

nice, and I think your probably right.
Would be cool if we could get a HD5870 to run the same exact settings.
 
I have a GTX280 and ATI 5870 installed and the benchmark reports that I'm running a GTX280 when it's rendering through the ATI card. Anyway, this is what I get with the system in my sig:

Vsync disabled
Unigine

Heaven Demo v1.0

FPS:25.4
Scores:640

Hardware

Binary:Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system:Windows 7 (build 7600) 64bit
CPU model:Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
CPU flags:3001MHz MMX SSE SSE2 SSE3 SSSE3 HTT
GPU model:NVIDIA GeForce GTX 280 8.16.11.9107 1024Mb

Settings

Render:direct3d11
Mode:2560x1600 fullscreen
Shaders:high
Textures:high
Filter:trilinear
Anisotropy:16x
Occlusion:enabled
Refraction:enabled
Volumetric:enabled

------------------------
Vsync enabled

Unigine

Heaven Demo v1.0

FPS:21.2
Scores:533

Hardware

Binary:Windows 32bit Visual C++ 1500 Release Oct 22 2009
Operating system:Windows 7 (build 7600) 64bit
CPU model:Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
CPU flags:3001MHz MMX SSE SSE2 SSE3 SSSE3 HTT
GPU model:NVIDIA GeForce GTX 280 8.16.11.9107 1024Mb

Settings

Render:direct3d11
Mode:2560x1600 fullscreen
Shaders:high
Textures:high
Filter:trilinear
Anisotropy:16x
Occlusion:enabled
Refraction:enabled
Volumetric:enabled
 
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.
 
@PS3
When you did your GTX275 SLI versus 5850 in DX11 mode, you disabled tessolation on the 5850 eh? to make it a fair-ish comparison?

@Evolucion8
Hmm, your results bring up a kind of interesting point in my mind. How big is the quality difference in your opinion from DX9 versus DX11 mode? Obviously, there will be some people who wiill claim that tessolation only makes a little bit of a difference and there's not much difference between DX10/DX11 blah blah blah. I'm curious if the difference from DX9 to DX11 is obvious in this engine?
 
WBurchnall said:
@Evolucion8
Hmm, your results bring up a kind of interesting point in my mind. How big is the quality difference in your opinion from DX9 versus DX11 mode? Obviously, there will be some people who wiill claim that tessolation only makes a little bit of a difference and there's not much difference between DX10/DX11 blah blah blah. I'm curious if the difference from DX9 to DX11 is obvious in this engine?
Click to expand...

The difference is mainly with the HDR, in DX9 mode it looks a bit more redish and less realistic and the jaggies are less noticeable with DX10 or above. The image quality between DX10 and DX11 is practically identical, the slighly difference is like the blurring is less block in DX11 and the shadows looks slighly better, but nothing else. I cannot use Tessellation because my card doesn't support it, (Its an HD 4870), so probably with an HD 5870, the difference in image quality would be more noticeable thanks to it.
 
Last edited:
WBurchnall said:
@PS3
When you did your GTX275 SLI versus 5850 in DX11 mode, you disabled tessolation on the 5850 eh? to make it a fair-ish comparison?

@Evolucion8
Hmm, your results bring up a kind of interesting point in my mind. How big is the quality difference in your opinion from DX9 versus DX11 mode? Obviously, there will be some people who wiill claim that tessolation only makes a little bit of a difference and there's not much difference between DX10/DX11 blah blah blah. I'm curious if the difference from DX9 to DX11 is obvious in this engine?
Click to expand...

Yes, tesellation was disabled to make things even I would imagine, since nvidia cards don't use tesellation. I only ran that bench to compare gtx275 sli to one 5850. I could probably overclock higher on the core, but I don't want to have to raise the voltage on the gpu.
 
can people still raise the voltage on modern cards? sorry to sound ingnorant, I just don't know much about that. Is that the same thing as 'pencil mod' or something?
 
PlatinumPC said:
can people still raise the voltage on modern cards? sorry to sound ingnorant, I just don't know much about that. Is that the same thing as 'pencil mod' or something?
Click to expand...

I believe with Asus and MSI 5850/5870 cards you can through software. People are also flashing other brands of cards with Asus/MSi firmware to unlock this ability.
 
WBurchnall said:
@Evolucion8
Hmm, your results bring up a kind of interesting point in my mind. How big is the quality difference in your opinion from DX9 versus DX11 mode? Obviously, there will be some people who wiill claim that tessolation only makes a little bit of a difference and there's not much difference between DX10/DX11 blah blah blah. I'm curious if the difference from DX9 to DX11 is obvious in this engine?
Click to expand...

With tessellation on the quality difference is pretty huge imo.
 
Heaven Benchmark v1.0

System: SFF Stock Clocks
Processor: Intel Core i7 860
Memory: G.Skill F3-10666CL7-2GBRH x 2 (4GB)
Mainboard: DP55WB
Display Adapter: ATI Radeon HD 5700 Series
Hard Drive: Western Digital Caviar Blue WD2500AAJS
Windows Version: Microsoft Windows 7 (6.1) Ultimate Edition (Build 7600)
ATI Driver Packaging Version: 8.66.6-091006a-089804E

Benchmark 1:
FPS: 36.8
Scores: 926

Settings:
Render: direct3d11
Mode: 1024x768 fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 4x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

Benchmark 2:
FPS: 29.4
Scores: 740

Settings:
Render: direct3d11
Mode: 1440x900 fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 4x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

Benchmark 3:
FPS: 25.4
Scores: 639

Settings:
Render: direct3d11
Mode: 1440x900 2xAA fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 4x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled

Not bad for a system that was bult more for power use then for speed. I'm happy! :D
 
@Tamlin
Nice screenshots Tamlin. It shows the quality difference pretty well/clearly. The DX10 really looks kind of bad. Like an impossibly-flat-block versus a curved surface.
 
tricky0502 said:
i thought the 4870 had tessellation hardware?
Click to expand...

It does, but it isn't exposed in DX, the developer must code for it. If it isn't because of DX11 support of it, it would die again like before with its previous encarnation called Truform.
 
You must log in or register to reply here.
Back
Top