ATI Crossfire & Eyefinity and NVidia SLI - any change from FSX?

Simon Lyon · Post by **Simon Lyon** » Thu Dec 16, 2010 4:49 pm

John,

I second a recent poster's request for a sub-forum on multi-monitor and graphics driving in general.

I've just ordered a new ATI card to try out Eyefinity as an alternative to Matrox solutions and it and my motherboard are also Crossfire compatible.

With FSX this has always been a contentious issue - with some people claiming to have got Crossfire (or SLI) to work and benefited from increased framerates while others swear blind FSX can't make use of second GPUs.

It appears you've made some improvements to rendering and shading, just from side by side visual comparisons with FSX, so what's the official line on Crossfire - is it worth me ordering a second card and linking them up or would it be a wasted effort?

Regards,

Simon

Legacy Support · Post by **Legacy Support** » Thu Dec 16, 2010 5:01 pm

The Hardware topic was created. http://www.prepar3d.com/forum-5/?mingle ... rum&f=24.0

There is no official line from us on crossfire. We are working with AMD next year, so we can try stuff out for sure. Perhaps in the meantime, others could post their experiences.

Post by **Beau Hollis** » Fri Dec 17, 2010 9:15 am

From what I've read, Crossfire and SLI are designed to emulate a single GPU interface, so there should be a performance gain even with one monitor without applications doing anything special under the hood. You may actually see more gains with non-SLI solutions where two graphics cards are used and one is assigned to each monitor at the OS level. DirectX9 doesn't allow you to explicitly share resources such as textures between devices (in this case monitors), so your doing a double resource copy at that point anyway.

All that said, I think the real crux of the issue is that the app tends to be CPU bound. We did make some upgrades to the existing renderer, but it would likely take a major overhaul and an upgrade to DirectX10/11 before we could hope to take full advantage of todays hardware.

Beau

Simon Lyon · Post by **Simon Lyon** » Fri Dec 17, 2010 5:19 pm

Thanks Beau, I'll probably take the chance and see what I can get out of it.

Your reply does beg the question though <grin>: Anyone in your team looking at turning "DirectX10 Preview mode" into the real deal?

I've never worked out if it's something largely finished, in terms of being a valid code base that just needs lots of bug-fixing, or if it's nothing more than sub-Alpha experimentation thrown in as a "coming soon" bit of marketing.

Being a natural cynic where MS is concerned - in other words a long-time IT professional

- I of course suspect the latter!

Post by **Beau Hollis** » Mon Dec 20, 2010 9:27 am

I probably shouldn't comment on our roademap or the state of the code-base itself, but form an end-user perspective there wasn't a huge value-add from the preview mode in terms of Dx10 specific features. You gained some self shadowing in the cockpit and slightly nicer water, but lost aircraft and object shadows.

N4GIX · Post by **N4GIX** » Mon Dec 20, 2010 2:32 pm

I truly believe that the only reason for the DX10 Preview was to shortcircuit any attempt at litigation based on the pre-release claims made by MS's PR folks...

...I could be wrong though!

FSMP · Post by **FSMP** » Mon Dec 20, 2010 4:08 pm

Right or wrong, if the time wasted on developing a less than usefull DX10 Preview, had been spent in other areas of FSX, the final release of FSX may well have been significantly more bug free in the DX9 mode, Also, a lot of 3rd Party developers would not have wasted time, chasing after DX10 compatability, that very few people currently use.

All the DX10 Preview developement seemed to show, was that developing for DX10 was not as straight forward, as we were all lead to believe, when DX10 was announced

FSMP · Post by **FSMP** » Mon Dec 20, 2010 4:27 pm

Quote:
Quote from bhollis on December 20, 2010, 09:27

You gained some self shadowing in the cockpit and slightly nicer water, but lost aircraft and object shadows.

The self shadowing in the Cockpit was a nice idea in theory, but in practice just lead to less than desirable effects. While the self shadowing might have been reasonably accurate as far as absolute lighting was concerned, the human eye is not a fixed appature camera. In reality, those lighting changes in the cockpit would have be largely compensated for by the human eye, and thus the effect "seen", would have been far less dramatic than they appear in the simulation.

Like many of the other "effects", (like Blume), having figure out that they could be achieved, and spending the time to develop them, it seemed their use was way "Over the top", and exagerated, just to make sure we all saw they were there.

I am looking forward to the day when P3D starts simulating the perceived lighting that the Human eye would see in real life, compensates for the restricted dynamic rage of the display systems, to produce a display that more accuratly simulates Bright (Blinding) lighting, and especially Dark lighting, with a heavy emphasis on the Human eye's NIGHT VISION characteristics.

Currently, flying at night in the simulator, as far as lighting in concerned, is very poorly simulated, and I look forward to that improving over the next few years.

I am hoping that the demands for realistic nightime training will accelerate that development.

Geoff

Post by **Beau Hollis** » Mon Dec 20, 2010 4:39 pm

To be fair, DirectX10 and DirectX11 are vastly improved APIs and, for new development, the API's are cleaner, more powerful, and easier to use. The simplicity and power comes at the cost of deprecating the old-school fixed function support. It had to be done, but it makes migration of an existing rendering system very complex. XP's dependency on DirectX9 forced many developers to support both, which nullified the simplicity of architecture. With XP reaching end-of-life, Vista supporting DirectX11, and a new generation of consoles on the horizon, DirectX11 is likely to pick up steam much faster than 10 did.