Unigine DirectX 11 benchmark

I'll stick to shorter posts, and provide links to longer posts for those who want to read into more detail.

Has anyone here tried to install a dedicated Physx board with their ATI card using the 1.02 cake mod to see if there is a performance boost in Heaven?

I am preparing to test that shortly.
 
A Physx board does nothing for the Heaven Benchmark. The program does use Physx, but it limits it to CPU usage. At least they are being fair in that respect.
 
I'll stick to shorter posts, and provide links to longer posts for those who want to read into more detail.

Has anyone here tried to install a dedicated Physx board with their ATI card using the 1.02 cake mod to see if there is a performance boost in Heaven?

I am preparing to test that shortly.

A Physx board does nothing for the Heaven Benchmark. The program does use Physx, but it limits it to CPU usage. At least they are being fair in that respect.

/facepalm :???:
 
A Physx board does nothing for the Heaven Benchmark. The program does use Physx, but it limits it to CPU usage. At least they are being fair in that respect.

Do you mean PhysX PPU board or both PhysX PPU and GPU boards?
 
Question:
So, it's true: You get tessellation under OpenGL with your HD Radeon (I have no idea, what I did, but now I got tessellation with the Heaven 2.0; I installed the OpenGL 4.0 preview drivers, really no idea why). With Tessellation on I have 5-6 fps with my HD4850, if there is nothing to tessellate, the frames go up. But:
http://www.geeks3d.com/20100210/tes...n-opengl-radeon-hd-5000-tessellators-details/
On Radeon HD 2000, 3000 and 4000, the hardware tessellation unit (called tessellator) has a max tessellation factor of 15.0. This is a hardware limitation that can’t be changed by software tweaks. The tessellator of Radeon HD 2000, 3000 and 4000 is a fixed function unit and we can’t program it with a shader. In OpenGL there are only two functions (described HERE) that allow to control it.

On Radeon HD 5000 series (Evergreen family), things are different. The Radeon HD 5000 includes the fixed tessellator of HD 2000, 3000 and 4000 AND a new programmable tessellation unit. This new programmable tessellation unit will be exposed via a new OpenGL extension (available shortly). And this new extension will allow to exploit the maximum tessellation factor supported on HD 5000 graphics cards: 64.0.

The DX11 tessellation and the DX9 tessellation are different pair shoes. Even if under OpenGL it's the same rendering pipeline, the tessellation factors are different. Is Heaven's max tessellation factor 15 or how does it work? Multi-pass?
 
The DX11 tessellation and the DX9 tessellation are different pair shoes. Even if under OpenGL it's the same rendering pipeline, the tessellation factors are different. Is Heaven's max tessellation factor 15 or how does it work? Multi-pass?

In theory they needn't be different, from what I've seen in their DX11 shaders they don't go beyond 15x amplification so the old tessellator should be "compliant". However, it'll be less efficient since you have to perform some R2VB shenanigans to emulate the HS and DS. Also, I don't think their OGL implementation is as robust as the DirectX one to start with, at least not yet, since performance doesn't seem equivalent, at least in my brief experience.
 
Tesselation isn't bound to tesselator itself, either.
For example ATI has demo which does (n-patches?) tesselation via geometry shaders ;)
So while you're "bound" by the limits of the tesselator on HD2-4, you could in theory further tesselate it with shaders if I haven't understood it wrong.
 
I'm a magical troll!

Anyway, we got the EVGA GTX 480 in the office today. The stock variety. It's actually rather good. Not sure about the color quality, but overall not a bad card.

If you already own a HD 5870, don't bother getting the GTX 480 as it is not much of an upgrade. Just get a cheap GT/S 240/250 for dedicated Physx and CUDA and run the Physx Mod 1.02. It works perfectly.

If you have a HD 5970, STICK WITH IT! You are only downgrading when it comes to raw performance. Again, just get a Physx board, if Physx is really that important to you.

If you have a GTX 275/285, or maybe even a GTX 295, OR if you have a Radeon 4XXX series (if you have the X2 card, you are only missing out on Directx 11 and Physx/CUDA), the GTX 480 will be an upgrade to you. If you upgrade from an already very fast Directx 11 card, it's a waste of money in my opinion.

When the GTX 485 comes out with the 28nm fabrication process, it will be MORE worth the upgrade, but still not worth it from a 5870 overclocked a bit, or a 5970.
 
Tesselation isn't bound to tesselator itself, either.
For example ATI has demo which does (n-patches?) tesselation via geometry shaders ;)
So while you're "bound" by the limits of the tesselator on HD2-4, you could in theory further tesselate it with shaders if I haven't understood it wrong.

Yes, you are correct sir. The SP units on the new 5000 series cards can run tessellation to boost, if the software is written to take advantage of Stream.
 
Yes, you are correct sir. The SP units on the new 5000 series cards can run tessellation to boost, if the software is written to take advantage of Stream.

Actually I was thinking more on the lines of removing the limitations of HD2-4's tesselator with geometry shaders, rather than HD5 & STREAM
 
Some time ago Jawed has requested Tesselation tests done with and without shadows.
I would did them sooner but I've been very busy lately with other less important stuff like moving homes, but in the end here they are!

Note: No Ambient; No Refraction; No volumetric
Shaders high, Aniso x4, No AA, 1920x1200
GPU HD5870 stock clock

11:37


19:59


23:59


Massive FPS differences :oops:
 
Last edited by a moderator:
Ooh very interesting, nice work. What's the FPS on precisely that position with everything turned on? Normal and Extreme?

I've just noticed that there's 3 sliders there for tessellation. Blimey, lots of testing possibilities there, e.g. a trade-off of distance and factor.

Seems like you have framerates that represent 3 different sets of one or more passes of tessellation. One with a shadow buffer from the sun, 11:39 only 32fps (approximately 2x the framerate you'd get with everything turned on?); another with minimal effects and seemingly no shadow buffer, 19:59 at 62fps and then 23:59 at 47fps which looks like its got some kind of full-screen "bloomy night lighting" post-process filter, but no shadow buffer - I'm not sure, in this last one, if there really would be an extra tessellated draw call in comparison with the 19:59 result.

I wonder if there's only a single tessellated draw call for the 62fps result?

Hopefully B3D will do something along these lines when investigating tessellation on GF100...

Or maybe the boffins are working on something more custom :p

Jawed
 
Ooh very interesting, nice work. What's the FPS on precisely that position with everything turned on? Normal and Extreme?

I've just noticed that there's 3 sliders there for tessellation. Blimey, lots of testing possibilities there, e.g. a trade-off of distance and factor.

Seems like you have framerates that represent 3 different sets of one or more passes of tessellation. One with a shadow buffer from the sun, 11:39 only 32fps (approximately 2x the framerate you'd get with everything turned on?); another with minimal effects and seemingly no shadow buffer, 19:59 at 62fps and then 23:59 at 47fps which looks like its got some kind of full-screen "bloomy night lighting" post-process filter, but no shadow buffer - I'm not sure, in this last one, if there really would be an extra tessellated draw call in comparison with the 19:59 result.

I wonder if there's only a single tessellated draw call for the 62fps result?

Hopefully B3D will do something along these lines when investigating tessellation on GF100...

Or maybe the boffins are working on something more custom

Jawed

Good that long weekend just started! It means I have some spare time to play with my card again!

I hope this will help even more than previous pictures!
Note that these pictures might be taken from slightly different angle, but it doesn't really affect frame rate at all.

Normal tess.








 
Last edited by a moderator:
Back
Top