Comments on: Mobile GPUs and Fill Rate https:/2012/01/mobile-gpus-and-fill-rate/ Developer resources for the X-Plane flight simulator Tue, 17 Jan 2012 14:27:02 +0000 hourly 1 https://wordpress.org/?v=6.6.1 By: Tony https:/2012/01/mobile-gpus-and-fill-rate/#comment-4038 Tue, 17 Jan 2012 14:27:02 +0000 http://www.x-plane.com/blog/?p=3911#comment-4038 In reply to Ben Supnik.

So, what is the solution if you are GPU bound?

I am running XP on an early 2008 MacPro Dual Quad 2.8 GHZ with an ATI 5770 1GB. My graphic settings are, I would say, set mid to low and it looks as though I am GPU bound going by what was posted here.

Lowering my cloud settings seems to have the greatest effect on FPS, I’m at 20% now. I have the textures set to Very High, and roughly 500MB of my VRAM is being used.

HDR completely destroys my FPS so I don’t even try it anymore.

With the newest update of XP, adding a few AI planes doesn’t hit FPS the way it used to. I can get away with 4-6 AI without a hit.

What should I expect from my system specs? How do I unbind from the GPU?

]]>
By: Chris https:/2012/01/mobile-gpus-and-fill-rate/#comment-3992 Sat, 14 Jan 2012 01:19:41 +0000 http://www.x-plane.com/blog/?p=3911#comment-3992 In reply to Ben Supnik.

Ah, that makes sense. I was wondering why I wasn’t seeing an increase in the other cores as I turned up the “CPU Intensive” rendering options. Thanks.

]]>
By: Ben Supnik https:/2012/01/mobile-gpus-and-fill-rate/#comment-3989 Fri, 13 Jan 2012 17:11:44 +0000 http://www.x-plane.com/blog/?p=3911#comment-3989 In reply to Chris.

CPU load is for the one core that renders. So when you are CPU bound in rendering, one core will be near 100%, CPU load will be near 1.0, and the other 3 cores will have variable work load based on what’s going on.

I think I’ve said this before, but I cannot emphasize it enough:

X-Plane will NOT use all of your cores to render.

]]>
By: Chris https:/2012/01/mobile-gpus-and-fill-rate/#comment-3988 Fri, 13 Jan 2012 15:28:28 +0000 http://www.x-plane.com/blog/?p=3911#comment-3988 In reply to Chris.

I meant to say that it is an increase of 8-10fps when reducing window.

]]>
By: Chris https:/2012/01/mobile-gpus-and-fill-rate/#comment-3987 Fri, 13 Jan 2012 15:26:50 +0000 http://www.x-plane.com/blog/?p=3911#comment-3987 Well, it would appear that my GTX 460 OC is indeed fill rate limited, since I see a drop of about 8-10 fps when reducing the window and using medium settings. One question I have though: when you mentioned the CPU load example of <0.7, is a 1.0 then 100% CPU load? Mine stays around 0.97 to 0.99, but looking at my Quad core shows only two of the cores really getting a workout.
Windows 7 64bit, Intel Core 2 Quad E9550 OC 3.4GHz, 8GB RAM, PNY GeForce GTX 460 OC.

]]>
By: Ben Supnik https:/2012/01/mobile-gpus-and-fill-rate/#comment-3986 Fri, 13 Jan 2012 02:38:59 +0000 http://www.x-plane.com/blog/?p=3911#comment-3986 In reply to Filippo.

Spot on. The optimizations available vary based on the genre. Portal culling isn’t terribly useful for a flight sim, for example. There’s also a performance cost to having a platform for third party content, particularly if that content is going to be stable across multiple rendering engine versions.

I believe a fair comparison would be _within_ the genre, and between products of the same generation. FSX and X-Plane 9 both have lighter hardware requirements than X-Plane 10; they had to be designed with limitations in what they did based on the lesser hardware.

None of this means that X-Plane 10 is running as well as it could or as well as we want it to; we will continue to optimize.

But there’s also apples and oranges. With X-Plane 10, you can input full OSM road data and you get extruded 3-d bezier curve junctions with merge ramps; I don’t think either FS X or X-Plane 9 will attempt to model the road grid with that kind of detail. (Note that the expense here is NOT just the detail, but the detail based on _arbitrary_ input data. You can model an overpass with an OBJ easily in any flight sim if you just want one.)

(And yes, I know, the road engine often goes crazy – this is a high priority bug for me to fix!!)

]]>
By: Ben Supnik https:/2012/01/mobile-gpus-and-fill-rate/#comment-3985 Fri, 13 Jan 2012 02:33:29 +0000 http://www.x-plane.com/blog/?p=3911#comment-3985 In reply to Andrea.

The only deleted comments are ones with profanity in them right now. Comments are moderated to fight spam, so they take a while to post.

]]>
By: Andrea https:/2012/01/mobile-gpus-and-fill-rate/#comment-3984 Thu, 12 Jan 2012 18:41:20 +0000 http://www.x-plane.com/blog/?p=3911#comment-3984 Ok, then how am I getting 35+fps in FSX with details turned up and yet X-plane struggles with clear conditions and the window as small as it will go on a 6950M. Curious how a few posts have been deleted from the blog, truth hurts.

]]>
By: Filippo https:/2012/01/mobile-gpus-and-fill-rate/#comment-3983 Thu, 12 Jan 2012 16:05:30 +0000 http://www.x-plane.com/blog/?p=3911#comment-3983 In reply to A Vardi.

Comparing the peformances of a 3D engine of a flight sim with those of a First Person Shooter is generally conceptually wrong, since the requirements for the two engines are rather different.

It would be like comparing a jeep with a sports car and blaming the former for having poor performances.

The 3D engine of a FPS can apply assumptions and optimizations that are not possible within a flightsim. Just one example is the fact that in a flightsim the 3D landscape is not necessarily known a priori (think about the possibility to add custom scenery on top of the default one).

Similar debates arose many times for the MSFS series also, but the principle is always the same. Flexibility and performance are not attainable at the same time.

Regards,
Filippo

]]>
By: Robert Accettura https:/2012/01/mobile-gpus-and-fill-rate/#comment-3981 Thu, 12 Jan 2012 03:20:00 +0000 http://www.x-plane.com/blog/?p=3911#comment-3981 In reply to Ben Supnik.

Really hope to see it improve, or at least degrade better.

]]>