View Single Post
Old 07-13-10, 12:06 AM   #7
ragejg
nV News Alumni
 
ragejg's Avatar
 
Join Date: Sep 2002
Location: Finger Lakes of Hell, NY
Posts: 5,399
Send a message via AIM to ragejg
Default Page 7

Benchmark Results



Unigine Heaven
  • DX11, shaders at high
  • 1920x1080, 4xAA/16xAF
  • Tessellation at Moderate


  • 1920x1080, 4xAA/16xAF
  • Tessellation at Normal



This is an unexpected result. It would have seemed at first glance that the GTX 465's 256-bit memory bus would have out-tessellated the GTX 460, but the little 768mb card actually beat its larger sibling in the moderate test, and stayed within 1 fps in the normal test. The HD 5830 tried to keep pace but ran out of gas once a little more tessellation was used.





Stone Giant
  • 1920x1080, 4xAA/16xAF
  • Tessellation at Medium


  • 1920x1080, 4xAA/16xAF
  • Tessellation at High



Once again, the EVGA GTX 460 SC keeps pace with the 465, while outclassing the Radeon. The reconfiguration of the core and optimization of the SM units must be playing a role here.





3DMark Vantage
  • Performance Preset (1280x1024)


  • 1920x1080, 4xAA/16xAF, Extreme textures, shadow shaders, shadows and shader quality, post processing scale at 1:5



Traditionally the Radeons have done better in this benchmark. It is interesting to see the 460 eclipsing the 465, again, especially in the GPU score for the 1920x1080 benchmark, as it also beats the HD 5830.

I don't understand why Radeons do so well in Feature Test 6. If someone can enlighten me, please do.

  • 3DMark with PhysX Enabled Performance



It appears that although the GTX 460 is strong at PhysX, the 465 is a little bit stronger for the most part.





D3D11 SDK Detail Tessellation Demo
  • 1920x1080, 4xAA/16xAF
  • Detail Tessellation
  • Texture: Rocks
  • Displacement selected
  • Tessellation Factor (edges): 11.00
  • Tessellation Factor (inside): 11.0
  • Link Tess. factors selected
  • Rotating camera selected





In this test, things line up a little closer to how the specs line up, with the GTX 460 lining up firmly in between the GTX 465 and HD 5830.





D311 SDK SubD11 Demo
  • 1920x1080, 4xAA/16xAF
  • Patch Division: 10
  • Bumpheight: 0.15
  • Toggle Materials





This demo makes heavy use of tessellation as well. The bump height adjustment adjusts the amount of displacement mapping (no real effect on performance), and the patch divisions setting adjusts tessellation. Worth noting is that on both the GTX 460 and 465, adjusting tessellation all the way up to setting 31.0 only brings performance down into the thirties, whereas the Radeon is already in the thirties at 0.15.





NVIDIA Human Hair Demo
  • 1920x1080, 4xAA/16xAF
  • Presets 3
  • Animation 3
  • Curly Hair, shadows, HW tessellation, Dynamic LOD selected
  • Play Animation
  • Loop Animation
  • Shadows selected
  • LOD rate at default
  • Add wind force, wind strength at max, wind randomness at max
  • Smoothing interval at default
  • Compute shader and simulationLOD selected
  • Show floor, show face, show floor shadows selected






... Another tessellation-heavy demo that is just too much for the Radeon. The GTX 460 and 465 run this test almost identically.






Gameplay Benchmarks -





Metro 2033
  • SP Mission: "Bridge"
  • "High" Preset, 1280x720, 4xAA/16xAF, DX11 on, Advanced DOF off


  • "High" Preset, 1920x1080, 4xAA/16xAF, DX11 on, Advanced DOF off



This game shows tessellation in action, amongst a host of other system-taxing effects. What's VERY surprising is that the EVGA GTX 460 SC manages to perform on-par with the GTX 465 in this game, sometimes beating it. It's reasonable to infer that at these settings the game would be very fluid at 1920x1080with no AA and a little bit of AF. Predictably the Radeon falls short in this game.





DIRT 2
  • "Ultra" Preset, 1280x720, 4xAA/16xAF
  • Custom track run without AI opponents


  • 1920x1080, 4xAA/16xAF



This game uses a little bit of tessellation as well, not enough to over-tax the Radeon. All three cards remain competitive in all tests, but surprisingly the GTX 460 pulls a solid win in both resolutions.





Battlefield: Bad Company 2
  • Single player mission 1
  • All settings at highest, DX11, HBAO on
  • 1280x720, 4xAA/16xAF


  • *]1920x1080, 4xAA/16xAF



The only DirectX 11 feature used in this game is soft shadow, which aren't very system-intensive. Since the release of this game, however, Radeon cards have always run this game quite a bit better. NVIDIA has told me that they are currently working on performance improvements for this game.





Dark Void
  • Jetpack jungle runthrough
  • All settings at highest, Physx OFF
  • 1280x720, 4xAA/16xAF


  • 1920x1080, 4xAA/16xAF



I have tested and retested this with all cards, getting the same result every time. This could just be a matter of where I'm at in the game; it might be a spot where the Radeon has an edge somehow. I may just drop this benchmark, or find another gameplay example that doesn't show such disparate performance.





Far Cry 2
  • 1280x720, 4xAA/16xAF
  • Demo(Ranch Medium), 1280x720 (60Hz), D3D10, Fixed Time Step(No), Disable Artificial Intelligence(Yes), Full Screen, Anti-Aliasing(4x), VSync(No), Overall Quality(Custom), Vegetation(Very High), Shading(Ultra High), Terrain(Ultra High), Geometry(Ultra High), Post FX(High), Texture(Ultra High), Shadow(Ultra High), Ambient(High), Hdr(Yes), Bloom(Yes), Fire(Very High), Physics(Very High), RealTrees(Very High)


  • 1920x1080, 4xAA/16xAF
  • Demo(Ranch Medium), 1920x1080 (60Hz), D3D10, Fixed Time Step(No), Disable Artificial Intelligence(Yes), Full Screen, Anti-Aliasing(4x), VSync(No), Overall Quality(Custom), Vegetation(Very High), Shading(Ultra High), Terrain(Ultra High), Geometry(Ultra High), Post FX(High), Texture(Ultra High), Shadow(Ultra High), Ambient(High), Hdr(Yes), Bloom(Yes), Fire(Very High), Physics(Very High), RealTrees(Very High)



All of the cards remain competitive through all the tests for this game, with (surprise!) the GTX 460 coming out on top.




Resident Evil 5
  • RE5 Variable Benchmark, all settings at highest
  • 64-bit HDR (ARGB16F) tweak
  • 1280x720, 4xAA/16xAF


  • 1920x1080, 4xAA/16xAF


This MT Framework-engined game looks great, runs wonderfully on the NVIDIA hardware, and runs adequately on the ATI card. Once again the GTX 460 is right at the heels of the GTX 465.





Lost Planet
  • DX10, all settings at highest
  • 1280x720, 4xAA/16xAF


  • 1920x1080, 4xAA/16xAF



Using the same engine as Resident Evil 5, I thought this game would show a similar result with ATI bringing up the rear. That didn't happen, and all three cards ran the game exceptionally well in most spots except for some serious choppiness in some of the snow areas at 19x10. This game is still quite taxing even to a newer PC.



Gameplay Summary


I did not expect the EVGA GTX 460 to do so well vs the GTX 465. In almost every game and benchmark it either comes close to matching it, or it wins altogether. Being the least expensive card of the group, the GTX 460 represents the best value. NVIDIA claimed that the GTX 460 was powerful in DirectX 9/10 games, and that it could play the toughest DirectX 11 titles. They made good on their word.


NOTE: I will be looking into the Dark Void issue over the following week, and despite being a total PhysX noob, I'm going to try to look into some of the PhysX aspects of some of the games I have (Dark Void, Mirror's Edge, Metro 2033) and see if I can add anything worthwhile to this review regarding this API.

Last edited by ragejg; 07-15-10 at 02:43 PM.
ragejg is offline