Support and Utilize are two different concepts when it comes to SLi/Crossfire. When I say Utilize, I mean the GPUs are involved earlier in the pipeline sharing processing duties ... this doesn't happen in FSX nor P3D. That's where we see 50-80% fps improvements.

 

But I'm curious what specific situations does SLi work with FSX? I've seen some posts regarding it's usage, but most turned out to be false positives and there was really only a slight benefit in the processing of certain type of AA.

Till 2 days ago I had 2 GTX680 in SLI. I only use the 2d cockpit without instruments and sometimes spotplane view ( 737 home cockpit ).

Using SLI instead of a single card gave me 40-45% extra fps...

Process Explorer

 

Great, thanks

As I am going to upgrade wat would you guys advice ?

Most of the time I will still be flying FSX.

 

Setup : 3 monitor in 5892x1080.

Vram usage in FSX : 900-950 mb.

 

EVGA GTX780 Ti Superclocked 3 Gb --- perhaps no vram enough, but the fastest single gpu card ( simular perf as 690 )

SLI GTX780 reference card 3 gb. --- perhaps no vram enough

GTX Titan 6 Gb --- after a few months a second one for SLI

 

I favor the Ti, but I would like to invest for the next 2-3 years.

AMD R290

I think there is a misunderstanding here in a sense of what people are seeing with VRAM usage and the conclusions they are coming up with.

 

The question should be not what is P3Dv2 using, but what does it NEED for your setup..   As the old saying goes.. The worst kind of RAM is the kind of RAM not used. Using a lot doesn't necessarily mean it needed it but hey, it's there might as well take advantage of it. Just because the sim is taking liberties with 3-4GB of VRAM.. Does it need it to run decently in most cases? From my experience, and reading. Nope..  

 

Exclusions apply obviously. Surround setups, higher resolutions etc. 

The question should be not what is P3Dv2 using, but what does it NEED for your setup.. As the old saying goes.. The worst kind of RAM is the kind of RAM not used. Using a lot doesn't necessarily mean it needed it but hey, it's there might as well take advantage of it. Just because the sim is taking liberties with 3-4GB of VRAM.. Does it need it to run decently in most cases? From my experience, and reading. Nope..

 

Exactly.

 

Using RAM is good.  Requiring RAM, not so much.

 

All people are looking at is the "using".

Does it need it to run decently in most cases?

 

 

 

The question should be not what is P3Dv2 using, but what does it NEED for your setup..

 

This doesn't make sense?  VRAM usage is based on the scene it's trying to render, that scene will change as you fly thru the world.  Your P3DV2 setting will influence the VRAM usage, but it's a dynamic usage.  If the scene needed to be rendered exceeds the amount of VRAM you have available, then the scene has to use system RAM (VAS) -- at which point you'll start moving to the limits of 32bit address space (OOM).

 

DDR5 VRAM is considerably faster than RAM ... so to process a frame you want that data in the fastest location possible (VRAM).  DX11 has helped reduce VAS usage considerably, the more data you put in VRAM the less VAS footprint you have ... but it'll depend much on how LM coded resource usage (so far I think they've done a fantastic job on resource usage).  

 

As I understand it (I'm not pretending to be a DX11 expert, I've only just recently started working with it), there are some legacy elements that will not get placed in VRAM and use more VAS (CPU based) resulting in OOMs in P3DV2 ... what is causing that I don't know, but it's most likely some legacy element in a 3rd party product (probably a FS9 aircraft or something FS9 based).

 

Here is a good read on CPU vs GPU resources and limitations of usage (direct from Microsoft's MSDN):

 

http://msdn.microsoft.com/en-us/library/windows/desktop/ff476259(v=vs.85).aspx

 

What it boils down to, the more VRAM you have for P3DV2, the better it's going to perform and the less likely you'll run into an OOM.

 

Rob

This doesn't make sense?

 

Well, it depends on, as you say farther down, how LM coded the resource management.

 

I haven't really kept up with all the under-the-hood details, but aren't terrain mesh and some other traditional CPU tasks being handled by the GPU?  Is LM pre-loading textures, vertexes and other data to VRAM based on anticipated near-future requirements to improve performance?  I dunno.  Maybe.  If so, VRAM actually used doesn't necessarily indicate VRAM required.

Ditto.

 

On my system, i7@4800ghz with an ATI 6950 (2gRam) feeding 3 displays @5700x1200 combined rez, I can easily see the scalability.  With moderate sim settings @ full rez I typically get about 15fps.  If I reduce the sim window to the equivalent of 1024x768...I can get 30fps.  In both cases when I look at the load on my GPU it is reported at 98%....which is basically maxed.  For me, thats good news....as there are much faster GPU's with even more memory to be had which will only provide better FPS at higher resolution.   And I must say 30fps at this detail level is pretty awesome.

 

 

 

 

The point is, P3D is scalable.

 

 

how LM coded the resource management.

 

True, but LM did mention the heavy usage of Instancing benefits in DX11.

 

 

 

In both cases when I look at the load on my GPU it is reported at 98%

 

Interesting, I rarely go above 30% GPU load, but I lock frames at 30 fps.  Big scaled improvements will come when they introduce SLi/Crossfire utilization in V2.1.

I only use the 2d cockpit without instruments and sometimes spotplane view ( 737 home cockpit ).
Using SLI instead of a single card gave me 40-45% extra fps...

 

Did you confirm this with Fraps or DXTORY?  I recall a few years ago I did similar testing and notice the improved FPS with SLi as reported by FSX, but when I loaded up Fraps it displayed a value closer to my original (but this was single monitor).

Did you confirm this with Fraps or DXTORY? I recall a few years ago I did similar testing and notice the improved FPS with SLi as reported by FSX, but when I loaded up Fraps it displayed a value closer to my original (but this was single monitor).

My tests were done with the FSX framerate counter.

Fps went up by 40% .

The difference could be seen without the framerate counter : Aerosoft Amsterdam Schiphol with 35% MT traffic and an overcast , the ai traffic remained smooth and not without small stutrers and a framerate of 17-19.

Framerate was 25-33.

 

Also flying at 2500 ft at 250 kias with a zoom of 0.90 over 3 screens remained sharp without stutters,

Defenitely a massive improvement.

 

That kind of smoothness I also want in P3D.

 

I prefer the 780 Ti , but the 3 gb vram is holding me back.

The Titan is still a good card with it's 6 gb vram, but about 10-15% slower than a Ti.

Using SGSSAA will give false looks of SLI working with both GPU's way over 90% GPU load

 

 

The Tiran is still a good card with it's 6 gb vram, but about 10-15% slower than a Ti.

 

Reports I've seen show a 780Ti 5% gain over stock clock Titan, however an overclocked Titan is about 10-13% ahead of a 780Ti which bares out with the BenchMark results I'm getting on my OC Titan.  But I haven't seen any 780Ti OC benchmark results yet.

I know that the moment I commit to either a 780 or a Titan, the prices will drop the very next day, or else a better card will be introduced.... :rolleyes:

Reports I've seen show a 780Ti 5% gain over stock clock Titan, however an overclocked Titan is about 10-13% ahead of a 780Ti which bares out with the BenchMark results I'm getting on my OC Titan. But I haven't seen any 780Ti OC benchmark results yet.

The test I was referring to was with an EVGA 780 Ti SuperClocked overclocked vs Titan overclocked

Reports I've seen show a 780Ti 5% gain over stock clock Titan, however an overclocked Titan is about 10-13% ahead of a 780Ti which bares out with the BenchMark results I'm getting on my OC Titan. But I haven't seen any 780Ti OC benchmark results yet.

The test I was referring to was with an EVGA 780 Ti SuperClocked overclocked vs Titan

 

http://www.techpowerup.com/mobile/reviews/EVGA/GTX_780_Ti_SC_ACX_Cooler/1.html