New Dx11 Benchmark StoneGiant Released

[justin.kerr]

yep, same thing I found.. seems glitchy as hell

[LagunaX]

Rig in sig.
1920 x 1200, Tes @ high.
Stock 5870 30fps.
900/1250 5870 32fps.

Someone run their 5770 and make me feel better...

[{SAS}TB]

1920x1200 - High Tesselation - DOF Off

GPUS @ Stock

[saaya]

They lost their touch. Have you seen 3D Mark Vantage? Those graphics look like they where made to look impressive for year 2003. Getting low FPS with High-End systems on those ty graphics, was obviously bad optimizations.

i hate vantage as well...
it almost needs as much time to load the tests as to actually run them :P
i think the spaceship thing looks cool, but the jane test is awful, just awful... only 2 game tests too... and there are games out that look much better and run a lot faster than vantage... so...
3dmark has really lost touch with reality... it used to be a good indicator of hw performance, but now... 2k6 scaled way too much with cpu, vantage barely scales with cpu at all...

Wow, bad day, saaya?

hha, no, i just sound like that sometimes ^^
if it would be an audio comment and not text youd know i wasnt pssed off at all

But I'm guessing they're choosing the wrong ambients to showcase tessellation, they're almost dark scenes (AvP, Metro 2033, this techdemo) or fast paced environments like in Dirt 2 where those differences are little noticeable. A game like Oblivion would show better results IMO.

yes, spot on! totally agree... thats what people want, highly detailled character models of their own character or characters they interact with... the game that comes up with tesselated boobs first will def get a LOOOOOOT of attention

so whats the verdict for this bench?
not all that interesting and it performs better on nvidia hardware, but theres a bug that makes it look like ati cards get loooots of fps?

[japamd]

The benchmark numbers shown by the techdemo are flawed, FRAPS tells a very different story

Frames	 Time (ms)	 Min	 Max	 Avg
14160	192702	37	90	73.481

73.5 fps on FRAPS and 147 fps on techdemo.

[W1zzard]

difference is x2 ... maybe the engine does two passes per frame and it counts both

[blindbox]

Tessellation levels compared

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

Those are impressive, thanks. I like to see the DOF at work.

[japamd]

difference is x2 ... maybe the engine does two passes per frame and it counts both

Just saw this video: http://www.youtube.com/watch?v=sM0uRcY6828

For SLI the FRAPS meter seems to match tech demo FPS meter.

BTW, very impressive result (avg 251 fps) and scaling of quad-sli on Stone Giant

Those are impressive, thanks. I like to see the DOF at work.

You're welcome

DOF on/off don't change anything here, maybe it's broken on ATI cards.

[RaZz!]

Tessellation levels compared

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

thanks for the sshots. how did we manage to stand games without tessellation? looks good imo

[Chickenfeed]

DOF on/off don't change anything here, maybe it's broken on ATI cards.

Again willing to bet it is a crossfire bug. The fraps numbers seem more inline with reality if you ask me. Diffusion depth of field didn't work on the pre release but it appears to function for me (single gpu) anyways in the release demo.

Just to make a correction, we aren't comparing depth of field on and off here folks. We are comparing 2 different methods of depth of field (the setting is for diffusion / normal ), with diffusion being the more demanding.

I still suggest crossfire users attempt to run the bench without catalyst AI on as this *should* disable crossfire and hopefully also so the proper framerate as like I mentioned previously with a single card I see the exact same info in fraps and the ingame counter.

[STaRGaZeR]

thanks for the sshots. how did we manage to stand games without tessellation? looks good imo

Because games without tessellation don't look as bad as these DX11 benchmarks do with it deactivated. Look at the Unigine bench for example, roofs, walls and the freaking stairs without TS are just flat with a flat texture on them. To put a recent example, roofs and walls in AC2 are not flat, they look good and could perfectly pass without any kind of tessellation.
The key here is that, in order to make a new feature impressive, you have to magnify it against the old ones. In this case the difference would be minimal in most cases, so you make the old ones look bad instead
You gotta love progress, before we could live with simple geometry and some POM, now for almost the same visual effect we need tessellation, DX11 and a significant speed hit. It would be nice if game developers learn how to properly use this feature, instead of putting an insane amount of polygons that nobody is going to see without wireframe, using crappy textures because tessellation gives the depth feeling and capping the non-tessellated version to 90's quality levels in general.

[Manicdan]

im with stargazer, with one exception, are these benchmarks built to test how well a system handles the environment, or are they tech demos that should show where this tech is headed.

if the screen was black and we couldnt see anything but the end numbers, then its just data, but because the show us pretty effects, we expect their goal is to impress us with the quality of the visuals. but in truth a game that costs 60$ is going to care alot more about it than a free demo. if AvP is a slideshow with tessellation on, then thats a failure, but the stone giant dosnt have to worry about that.

[TurboDiv]

I would say this is DX11 tech demo to showcase the technology and what you can expect in the future. The ability to adjust the amount of details up close and far away is phenomenal with tessellation and the performance hit is not insane too. It deserves a chance imo.

[Calmatory]

2k6 scaled way too much with cpu, vantage barely scales with cpu at all...

Which is better?

3DMark 06 didn't scale with the CPU almost 5 years ago, due to way less powerful GPUs. Same thing with Vantage.

Back in the times when hardware(in this case video card, and sometimes even CPU) reviews were just 2-4 game tests(average fps) + 3DMark score, it was quite easy to just use 3DMark as a "good enough" indicator of performance. Nowadays as a hardware review uses 10+ games, multiple different settings, resolutions, and sometimes even different drivers, so 3DMark score does not really play a big role.

As there are numerous game engines, thousands of different settings and resolution combinations and numerous driver versions, there is no way to make a "Good enough" "gaming performance test". People demand too much. As if the 4 FPS(of 70 FPS) difference between cards makes them non-equal in practical performance. Visibe in graphs yes. Equal in games.

3DMark is just another engine among others.

[Astennu]

This benshmark is useless :S

My 5970 @ 900-1100 was 4x faster then my 5850 @ 950-1225. Thats just not possible.

34 vs 127 FPS avg 1920x1200 High setting.

[THOMO]

I've only a GTS250 atm but ran this anyway.

tesselation & DOF disabled.

47
48
48