Ogre Forums

Hello again.

When I run simple demos under OGRE 1.7.x on my Windows 7 PC, I get ~ 1,300 FPS. Now I know that this frame-rate doesn't mean anything, because the scene is essentially empty.

And, when I run the OGRE 1.8.1 SDK Samples on my new Laptop, I get ~ 130 FPS.

One subject which was mentioned somewhere, was that OGRE Devs might want to put an upper cap on the frame-rate, because to be running an empty scene at 1,000 FPS (+) is understandably not regarded as useful.

In OGRE 1.8.1 , has such a cap in frame-rates been applied? If not, I must assume that my laptop is pretty weak.

And yet my new laptop could be weak, if the shared memory its graphics chipset relies on, is not dual-channel RAM... ?

It's possible that if you have an optimus capable laptop, it hasn't recognised ogre as intensive 3d so it's running using the onboard gpu (crappy intel thing) to save power.
(Optimus lets the computer choose between onboard intel gpu and nvidia geforce gpu based on what's running)

I'm pretty sure I've heard of people having trouble with it on here, but I don't own an optimus capable laptop so I've largely ignored it.

There's no framerate limiter, besides vsync.

An interesting suggestion.

According to the left-click of the desktop, I only have one GPU. And according to my Device Manager, I only have one GPU.

But a fact which I was not yet aware of, is that my AMD Catalyst Control Center, has an Advanced View. Previously, I had only been able to call up its Basic View.

In the Advanced View, I was able to change certain settings, specifically on a per-application basis, for the SampleBrowser.exe program.

In so doing, I've boosted my typical frame-rates to ~ 350FPS, which is certainly better than what I got before.

I should also add, that the frame-rates of 1000FPS (+) which I got on my tower PC, were due to a GeForce GTX 460 card , with its own, physical VRAM.

On my new laptop, it's an AMD HD 8610G , which has shared memory , and I have read that the performance varies, depending on whether my system RAM is dual-channel or not, which I have yet to determine with certainty.

Yet, changing these settings specifically for the app, did not make Render To Vertex-Buffer available, which was my complaint in the other thread. It only improved my frame-rate, to beyond what I'd expect from a crappy Intel chip.

Also, when I run the 'ParticleGS' demo, the capabilities test in the source-code shows that I do have Geometry Shader available, which I also wouldn't expect from a crappy Intel chip. It only tells me, that I don't have R2VB available.

If I switch OGRE to DirectX 9, then it tells me, I don't have Geometry Shader available...

Thx,
Dirk

(Edit:) Also, the Advanced View of the AMD Catalyst Control Center should show me, if I have more than one display adapter, which it doesn't.
What I do recall now however, is that it can cause major performance hits specifically with OGRE, if we stick to the system default settings for (multi-sampled) anti-aliasing and/or tessellation . By telling it to use application-determined settings for tessellation , I may be turning that feature off...

And within OGRE, now at 350 FPS, I did have FSAA turned on, to 2x2, because I'd feel a little silly if I simply had FSAA turned off completely...

But what I am noticing now on this hardware, is that for anti-aliasing, I have settings in two places, 'Use Application Settings', 'Override Application Settings', and 'Enhance Application Settings'. I have it once for AA Mode, then again for AA Samples. And then for AA Method I can select Super- or Multisampling. And for AA Filter I have Standard, Performance or Quality.

What I suspect is that if I played with the 'Enhance Application Settings' just right, I might be able to trick OGRE into doing MSAA, which it could not normally do. But if I then had OGRE set to 2x2 FSAA as well, it would just compound the total number of samples computed...

(Edit2:) Haha! If I set 'AA Sampling Mode' to 'Enhance Application Settings', then '...Samples' is automatically also 'Enhance Application Settings', at which point I can set '...Method' to 'Adaptive Multisampling'. Next, I can set OGRE's internal FSAA to 0, and run it.

And then the Cube-Mapping Demo runs at 400FPS!
And the Facial Animation Demo now runs at ~ 1000FPS!