New Dx11 Benchmark StoneGiant Released

http://www.stonegiant.se/

Quote:

---

“With DX11 hardware becoming more readily available, we decided now would be a good time to highlight the advanced features of our engine, including DX11 and tessellation support,” said Tobias Persson, Lead Engineer at BitSquid. “Tessellation adds an awful lot to the visual fidelity of the scene and is something that consumers can readily see for themselves. We are looking forward to seeing how games will take advantage of our engine.”

---

http://www.youtube.com/watch?v=DOEDf33AZh4


Keyboard & Mouse mappings
Enter Start/stop cinematic sequence
1-5 Jump between five preset close-up shots
F1 Toggle wireframe (on/off)
F2 Cycle between tessellation settings (off/medium/high)
F3 Toggle depth of field effect (diffusion/regular)
F4 Toggle performance statistics (on/off)

WASD Move camera
Mouse Rotate camera
Mouse Wheel Change camera movement speed

That's two threads about the same topic posted at around the same time :D
Check:http://www.xtremesystems.org/forums/...d.php?t=250185

lol.. i am the one who failed actually :D i wasted some times while searching quotes and videos

its cool when the benches themselves are fun to look at.

Quick, someone get to benchmarking!

pcgh managed to bench a prerelease version in their gtx 480 review.
That version very much favored nvidia cards, so either there was a bug somewhere or theres ALOT of tesselation.

Been waiting for this ! Downloading now ! Their servers seem to be overloaded or something, its slow...:rolleyes:

1.5Hrs till bench time lol

reminds me of Lineage 2

That's what I got

1920x1200 - Tessellation High

http://i43.tinypic.com/315n2iv.jpg

Reminds me of the golem in Dragon Age pimped out with the green rocks.

I hit ENTER but it doesn't bench anything, it just starts to run the demo ?! BTW the DEMO is amazing ! Runs mega smooth on my baby.

PS: NVM, got it. F4 for benchmark measurements.

I did 2 runs at 1680x1050 full-screen, tessellation high: 98 FPS average on both runs. Runs extremely well !!!

here i did another 3 runs, this time i did screenie:

@japamd, you should also get average of your 3 runs, lol.

EVGA GTX 470 SLI stock clocks
Drivers:Nvidia 197.41

@Turbodiv F4 turns on FPS monitor, it updates the sceen after each run with avg FPS.

http://img690.imageshack.us/img690/7...giant73fps.jpg

Quote:

---

Originally Posted by TurboDiv

@japamd, you should also get average of your 3 runs, lol.

---

I just pressed enter 3 times... what else am I supposed to do? :p:

This is my scores with 5870@975mhz/1300mhz + Xeon w3520 (I7 920) @ 195*21 (4095mhz).

Tessellation high on all runs.

1024*768 = 1 Run -> 61fps average
1024*768 = 2 Run -> 62fps average
1024*768 = 3 Run -> 61fps average

2048*1536 = 1 Run -> 48fps average
2048*1536 = 2 Run -> 48fps average
2048*1536 = 3 Run -> 48fps average

When i saw the benchmark installed physX on my computer i knew Nvidia cards would score high..

*edit*

Maybe my drivers is holding me back, using 10.2 atm.

Actually, I did another run with PhysX disabled and I got exactly the same results as above. Check your drivers/system. PhysX has nothing to do with it.

How do you get it to show you the framerates at the end??

You press the F4 button before you hit ENTER.

Average FPS (DOF off): 75
Average FPS (DOF on): 42

This was with a GTX480 @ 800/1600/3700 (197.44 drivers) and Core i7 920 @ 3.6GHz and 6GB DDR3 1600 RAM

1920x1200, 16:10, full screen, tessellation high

Quote:

---

Originally Posted by TurboDiv

Actually, I did another run with PhysX disabled and I got exactly the same results as above. Check your drivers/system. PhysX has nothing to do with it.

---

Yea but dont you agree it means that Nvida have been involved in the making of this benchmark ?

*edit*

Just a typo

I think im going to have to agree cause we don't want to start an argument. And i really mean it, if it installs PhysX its probably developed with nvidia in mind. Just an obvious observation. But than again, that's why we have the thread so other people would test also and we will have a general idea and subsequently proof of the theory. ;)

Quote:

---

Originally Posted by TurboDiv

I think im going to have to agree cause we don't want to start an argument. And i really mean it, if it installs PhysX its probably developed with nvidia in mind. Just an obvious observation. But than again, that's why we have the thread so other people would test also and we will have a general idea and subsequently proof of the theory. ;)

---

Yes agreed, iam sorry if i sounded harsh. Think i just gonna update my driver and see what happen. :yepp:

Cheers

Quote:

---

Originally Posted by Det0x

Yes agreed, iam sorry if i sounded harsh. Think i just gonna update my driver and see what happen. :yepp:

Cheers

---

Yes, Catalyst 10.2 is holding you back.

This seems to be a bit CPU bound too, tried it again with CPU overclocked at 4.2 Ghz

1920x1200 - Tessellation High

http://i44.tinypic.com/24y4bbb.jpg

Fullscreen screenshot
http://img85.imageshack.us/img85/500...042201431c.jpg

Now, with CPU overclock it's a lot more consistent between runs :up:

http://i44.tinypic.com/11b2n9t.jpg

Previous run in previous post was with CPU stock (2.66 Mhz - Turbo off)

That's why anyone using an i7 in stock should burn in hell :D

Ok, so I ran it, started with the stone giant, hit enter went through the bug attack and then back to the stone giant. Hit enter again and it went back to the bug attack. I hit F4 and performance stats was on, but never got a screen that showed the results, what am I doing wrong here?

I took some screenshots showing tessellation levels in wireframe mode, it's amazing to know that so many polygons are being rendered in High tessellation mode !

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

download mirror up at tpu http://www.techpowerup.com/downloads...DX11_Demo.html

bitsquid tech?
fatshark?
never heard of either one of them... :confused:

1. where the FCK is futuremarks dx1 benchmark???? :stick:
2. is this an ati/nvidia demo disguised as a third party benchmark? :para:
3. why would insects attack a stone golem... and why would a stone golem care and fight them off? :confused:
4. his face looks stupid, and the spider thing is weird, besides that, very nice looking!

Quote:

---

Originally Posted by japamd

I took some screenshots showing tessellation levels in wireframe mode, it's amazing to know that so many polygons are being rendered in High tessellation mode !

---

that youd never knew were there unless you use wireframe mode ;)

Quote:

---

Originally Posted by Det0x

This is my scores with 5870@975mhz/1300mhz + Xeon w3520 (I7 920) @ 195*21 (4095mhz).

Tessellation high on all runs.

1024*768 = 1 Run -> 61fps average
1024*768 = 2 Run -> 62fps average
1024*768 = 3 Run -> 61fps average

2048*1536 = 1 Run -> 48fps average
2048*1536 = 2 Run -> 48fps average
2048*1536 = 3 Run -> 48fps average

---

Same hardware settings + cata 10.3

1024*768 = 1 Run -> 62fps average
1024*768 = 2 Run -> 62fps average

1600*1200 = 1 Run -> 49fps average
1600*1200 = 2 Run -> 49fps average

Looks like driver dont matter that much. Anyway iam going back to cata 10.2
I use a samsung 1100mb (21"crt) and this guide dont work with cata 10.3

So iam stuck @ 1600*1200 with 60hz, and with 10.2 i could use 1600*1200 @ 100hz or 2048*1536 @ 85hz. :down:

*edit*

Forgot to say the "Use extended display identification data (EDID)" in calatalyst 10.3 dont work.

Quote:

---

Originally Posted by japamd

I took some screenshots showing tessellation levels in wireframe mode, it's amazing to know that so many polygons are being rendered in High tessellation mode !

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

---

Can you post screenshots without the wireframe? It seems like the graphics look like crap to me. Plus, what's the point if the difference is only visible in wireframe?

It crash, still after installing newest dx11?

nice bench.

51 fps in 3 runs with q6600 @ stock win7 64bits & Ati Radeon HD 5870 cat 10.3. 8 Go DRR2 800. ( all settings by default. )

( running with some apps in same time, i've don't optimized ).

Graphics are nice.

Ridiculous what DOF does to fps, especially with no noticeable IQ gain..
DOF kills Metro as well..

Quote:

---

1. where the FCK is futuremarks dx1 benchmark????

---

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 :banana::banana::banana::banana:ty graphics, was obviously bad optimizations.

Quote:

---

Originally Posted by saaya

bitsquid tech?
fatshark?
never heard of either one of them... :confused:

1. where the FCK is futuremarks dx1 benchmark???? :stick:
2. is this an ati/nvidia demo disguised as a third party benchmark? :para:
3. why would insects attack a stone golem... and why would a stone golem care and fight them off? :confused:
4. his face looks stupid, and the spider thing is weird, besides that, very nice looking!

---

Wow, bad day, saaya? :D

Quote:

---

Originally Posted by saaya

that youd never knew were there unless you use wireframe mode ;)

---

This one has very perceptive differences in normal rendering mode too, I think. I'll do what blindbox is asking below so we can compare those tessellation off/medium/high modes.

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.

Quote:

---

Originally Posted by blindbox

Can you post screenshots without the wireframe? It seems like the graphics look like crap to me. Plus, what's the point if the difference is only visible in wireframe?

---

I'll do that later, besides graphics being not so good, tessellation adds a very good level of detail to them and that's noticeable in normal rendering mode too.

what does depth of field change? seems to be a big perf hit, but are the effects worth it?

Nice demo, but i think the wait for a truly impressive implementation of tessellation continues.

Quote:

---

Originally Posted by ElSel10

Average FPS (DOF off): 75
Average FPS (DOF on): 42

This was with a GTX480 @ 800/1600/3700 (197.44 drivers) and Core i7 920 @ 3.6GHz and 6GB DDR3 1600 RAM

1920x1200, 16:10, full screen, tessellation high

---

Average FPS (DOF off): 69
Average FPS (DOF on): 36

This was with a GTX470 @ 608/1215/3348 (197.44 drivers) and Phenom II @ 3.5 gigand 4GB DDR2 1066 RAM

1680x1050, 16:10, full screen, tessellation high

i liked your format so thought i would just quote it for some comparison between 470/480 and AMD/Intel :up:

I suppose DOF is on by default? How do you guys are turning DOF on?

Ooops... nevermind, found in readme.txt :P

Tessellation levels compared

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

OFF - MEDIUM - HIGH

Quote:

---

Originally Posted by XSAlliN

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 :banana::banana::banana::banana:ty graphics, was obviously bad optimizations.

---

Indeed i think the gfx in 3DMark 06 look much better.

rig in sig 1920x1200 high settings DOF on

http://i18.photobucket.com/albums/b1..._32_38_757.jpg

At 1680x1050 with high tessellation, 3 run average, system in signature :

29fps ( diffusional depth of field on )

42fps ( diffusional depth of field off )

As far as I can tell though turning off diffusional dof ( F3 key ) doesn't completely turn off dof however. In the prerelease version, depth of field worked on ATI hardware but not diffusional depth of field.

your #'s seem low?

Its inline with what I got using the prerelease (normal dof) one so no idea... Peoples numbers look very inconsistent though, they varry way too much system to system so who knows.

Like for example DosDuoNo did his at 1680x1050 on his 470 and got 36 which I'd believe given Fermis prowess at tessellation.

Its possible it doesn't like the 10.4 drivers but no idea, this system still gets 10k on vantage extreme so i know its working properly :shrug:

did you run it in full screen?
I am running the 10.4a's

Yes full screen. I tried windowed mode for the hell of it and it made no difference.

the FPS in game, and what FRAPS say are way different, I just checked that..

I did 3 runs with fraps running along side in bench mode and 1 not in bench mode ( so I could still see the counter ) and both fraps as well as the ingame counter both showed the same numbers for me. When it comes to multi gpu setups though in game frame rates and even fraps can be misleading.

Could you try running with Catalyst AI off and at 1680x1050 perhaps? This should provide a more relavent comparison.

Hi

I ran at 1680x150 - DOF off and High Tesselation and got

http://img691.imageshack.us/img691/15/stocknt.png

Seemed a tad odd and ran the BM taking pics at random times of the FRAPS counter:

http://img249.imageshack.us/img249/2825/wtfr.png

Something looks off to me ?

Nice looking BM though :D

same settings DOF and Tesselation off

http://img27.imageshack.us/img27/8171/systemrun.png

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

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

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

1920x1200 - High Tesselation - DOF Off

GPUS @ Stock

http://img249.imageshack.us/img249/3374/19201200.png

Quote:

---

Originally Posted by XSAlliN

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 :banana::banana::banana::banana: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... :shrug:
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... :confused2

Quote:

---

Originally Posted by japamd

Wow, bad day, saaya? :D

---

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 :D

Quote:

---

Originally Posted by japamd

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 :lol:

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?

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

Code:

---

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

---

73.5 fps on FRAPS and 147 fps on techdemo.

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

Quote:

---

Originally Posted by japamd

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.

Quote:

---

Originally Posted by W1zzard

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 :clap:

Quote:

---

Originally Posted by blindbox

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.

Quote:

---

Originally Posted by japamd

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? :rofl: looks good imo :)

Quote:

---

Originally Posted by japamd

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.

Quote:

---

Originally Posted by RaZz!

thanks for the sshots. how did we manage to stand games without tessellation? :rofl: 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.

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.

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.

Quote:

---

Originally Posted by saaya

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.

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.

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

tesselation & DOF disabled.

47
48
48