[OpenGL Benchmark] Fur Rendering - 1280x1024

[JeGX]

be sure to have a fast internet connexion. The contest mode requires an access to oZone3D.Net server.

[largon]

I think this benchmark is the only real "3DMark" out there! Futuremark's dabblings are actually just CPU benchmarks but not "Furmark", no sir... Furmark is a 101% GPU benchmark, CPU doesn't matter at all - and that's the way it should be. Thanks for prog and keep up the good work! 

It would make more sense if each result that get's submitted in the official score list wouldn't create a whole new result, rather each user would get only one result per card type (8800GTS 512MB, HD3870, etc.) and not as many as the user submits with the same card.

And it would be nice if multi GPU detection for the validations listed in the score page would work. As clearly, it doesn't work atm.
For example:

Score: 6623 o3Marks
Submitted by W4RHAMM3R @ March 31 2008, 1:01 am

App Version: oZone3D.Net_FurMark_v1.3.0_Build_Mar 3 2008_at_22:04:00

Renderer: GeForce 8800 GTS 512/PCI/SSE2
Number of Active GPUs: 1
Graphics Drivers: Unknown version

Bench Duration: 60 sec.
Resolution: 1280 x 1024
MSAA samples: 0
Window Mode: fullscreen

CPU: IntelR CoreTM2 Quad CPU Q6600 @ 2.40GHz
CPU Speed: 3200 MHz
Operating System: Windows Vista ver.6.0 build 6001 Service Pack 1 v.668

I know this is a SLi result.

[Fahim]

XFX GeForce 7950 GX2 1GB (520,1300) 175.75



XFX GeForce 7950 GX2 1GB (550,1400) 175.75




Will post my other cards later...

[JeGX]

I think this benchmark is the only real "3DMark" out there! Futuremark's dabblings are actually just CPU benchmarks but not "Furmark", no sir... Furmark is a 101% GPU benchmark, CPU doesn't matter at all - and that's the way it should be. Thanks for prog and keep up the good work! 

Thanks 

It would make more sense if each result that get's submitted in the official score list wouldn't create a whole new result, rather each user would get only one result per card type (8800GTS 512MB, HD3870, etc.) and not as many as the user submits with the same card.

I'll try to improve score submission, you're right.

And it would be nice if multi GPU detection for the validations listed in the score page would work.

Yep there is still some detection bug with Vista. I'll try to fix them.

[Sophosaurus]

I did some testing with different settings and noticed, CPU-speed doesn't matter, which is fine, but also GFX-memory's speed doesn't matter which is not realistic.

So it's a real GPU-only-benchmark. Not bad at all but this should be known.

Sapphire Atlantis 3870 Dual Slot: 3866 (no voltmod, stock heatsink)

[Stefan]

Just noticed an issue with v1.3.0 and Vista 64 bit SP1, i have no 32 bit version to test:
If "aero glass" is enabled, you get a black desktop after benchmark has finished, however you can see the mouse pointer changing when moving over icons.

Furmark 1.0 works fine with "aero glass"

[Sophosaurus]

- found out i posted bullocks, so this posting is empty now -

[JeGX]

Just noticed an issue with v1.3.0 and Vista 64 bit SP1, i have no 32 bit version to test:
If "aero glass" is enabled, you get a black desktop after benchmark has finished, however you can see the mouse pointer changing when moving over icons.

Furmark 1.0 works fine with "aero glass"

not cool, I haven't tested it with Vista when I released v1.3.0. An update is planned soon.

[JeGX]

I did some testing with different settings and noticed, CPU-speed doesn't matter, which is fine, but also GFX-memory's speed doesn't matter which is not realistic.

thanks for this test. It's weird indeed. Maybe the torus mesh and the textures used are not big/dense enough and then they're well cached. I don't know 

[Sophosaurus]

Makes sense. So you now accidentally found out something new and important about GPUs. We do know nothing about their cache size. Am i right?

We know caches make a big difference for CPU performance. We can test shader performance with benches like this, but we still have to guess what makes a good gaming card.

What about measuring cache sizes by using different mesh sizes in a loop to find the point framerate changes?

[indiana_74]

Zotac GeForce GTX280 1GB DDR3 (684/1296/1404) 177.35

[JeGX]

Yeah the first GTX 280 benchmarked with FurMark 
From the overall top-list scores, the GTX 280 is equal to two 8800 GTX in SLI...
I should update that list! Will do asap.

There is still a bug in driver detection under Vista SP1 

[Underground78]

Hello,

It seems that FurMark 1.4 doesn't take advantage of a CrossFire system (of two Radeon HD 2600 XT, yes I know it's quite odd), the second GPU isn't used at all. If I've understood well some older posts, it should normally work. So if you need some beta testing, I would be happy to help (with my poor graphics cards) !

[JeGX]

Crossfire support seems not an easy task   
I'll try to look more serously at this topic but before I have to find a XFire system...

Thanks for your (future) help in beta-testing. That's cool, I stick your name on my beta-testers list (before I forget, send me a pm with your email address).

Seems that guy has managed to run a crossfire system with FurMark: http://www.ozone3d.net/benchmarks/score.php?id=9e026632cffe5414e87c3ac16b3fe944

[winit_a]

score 5660
E2180 @ 3.00GHz
2GB   DDR800
Powercolor HD4850 700/1200