What is up with the optimization of this software

Also it's worth adding to what Domi said (now this fact isn't 100% free of speculation), rF2 is programmed in 2 threads, one for physics and one for graphics etc. If physics would truly require as much power as some here claim, then surely ISI would consider dividing physics computations into further threads, as almost everyone owns a quad core CPU these days.

Truth is that it doesn't make much difference whether you run rF2 with an i5 2K something or i7 4K, it's not where the difference comes from, neither in FPS nor laptime. It's all about rendering graphics, which is pretty much the same complexity of task as in AC and pCars, thus one would expect similar performance.

 

That slowmotion is what happens todo me although the fps counter isla displaying 100 frames

 

A little bird told me that, without tables, the required CPU power would be about 1.000.000 more .... in approximation, say a strip long about 400 km of PC linked together.

This info so that you can understand the relativity of the thing ... <<<so a lot of the stuff is not really done in realtime>>> ... and the physical complexity of the TGM (thermodynamc brush model) which is actually an elastic body.

The physical engine cant be an issue because a 100% load results in a fall of the real time computations (slow motion effect), but, more the CPU load, more the"fatigue" executing the dedicated tasks.

 

Device drivers need CPU time, too.
Some steering wheel drivers may not be optimized enough to handle high FFB rates produced by rF2 and end up taking a lot of CPU. Then there's previously mentioned graphics engine, which is using a lot of CPU. Graphics drivers often spread their load across multiple CPU cores so it wouldn't matter if physics run on 2 cores, sharing both of them with graphics driver, or on one core, with graphics driver running on the other one.

Also - if physics would run on 2 cores, both of them would have to finish their task at hand, before simulation could move to next frame (you can't continue with partial results), so in fact, you would get blocked if either core gets blocked. Another risk of having stutters.

So it's not as straightforward as you think.


Take some words of wisdom from me and leave engineering stuff to engineers

 

There is no dark side of the moon really...matter of fact it's all dark.

 

And running on an iMac has no effect!?!?!

 

If in 10-20 years consumer PCs can run simulators with realtime CFD normally I will keep saying the same then Obviously level of implementation increases with time, so does our CPU power. But you are somehow agreeing with what I initially said in your reply to stonec; if the CPU handles the physics engine, devices, and even part of the graphics load (and according to what you said, maybe not even very efficiently, due to lack of expertise in modern graphics from simulation companies), then we can conclude that the complexity of physics engines in modern sims isn't a problem at all. If we raise the bar and we put many more variables to do in realtime, with more calculations, faster rates etc, then yes, I would say they can be hard for our CPUs. But I repeat, my comment was strictly about current sims, in general.

I don't disagree with this, and that's why I textually said that a lot of the stuff is not really done in realtime. A lot ≠ everything; as you have explained there are many things that cant be defined in lookup values.

 

I run the same resolution you do but I have a 4Gb 770 GTX. I recommend dropping the anti-alias to 4 on the game, we already play at higher resolution... don't need the effects at such high values. HDR will cost you on average 15 frames, but it is a bit of a necessity if you worry about appearance. But it's nice you got it running well, enjoy. And ignore fanboyism here.

 

"If A+B+C is a problem then we can conclude A is not a problem at all" ?

I don't think logic works that way

 

Just for information. the CPU is certainly involved in creating the graphics. It creates a 3D wireframe world which you race in. The graphics card textures, lights and transforms the 3D world to the 2D image required by your screen.

My own experience with rF2 is the change from full player physics to simplified AI physics had no effect on framerates, indicating the advanced physics are not the cause of frame-rate issues, at least on my PC.

 

Doing some tests s build or two back, I lost around 30-40% of framerates when enabling FXAA together with AA. As soon as I disabled FXAA (kept regular AA enabled) my fps shot back up.

 

I see that OP uses Sync GPU, i never had smooth frames with that, only with Sync Video in the time I've used ingame option. Now I'm forcing V-Sync through Inspector anyway..

 

Seems that ISI devs simply don`t mind or maybe they don`t know how to solve rF2 SLI performance, however it is possible to ask SLI profile in nvidia forums like pCars community did, maybe you could make thread and ask if it is possible to get some help from nvidia

 

Naturally few of us here if none knows how gMotor internally work, so let's for sake of argument consider it all a black box. Let's then consider what we can measure: PC components (input) and performance (output).

It is well known based on experience from users here that rF2 runs fine on a couple of generations old CPU's. It is also known that rF2 physics can run in realtime even on an ancient 2006 Core 2 E6600 CPU (would never work with Assetto Corsa), I confirmed this last weekend on my old PC. Finally it is known that rF2 FPS, on today's hardware, scales quite directly with GPU horsepower (see for example this chart).

Considering the fact that, to my knowledge, physics are not calculated on GPU (ignoring PhysX), everything points to the fact that the bottleneck here is gMotor 2.5 graphics engine.

 

So, regarding that graph, how in the hell is possible that right now my FX6300 + Radeon 7950 could thrash that same benchmark at more than 70 fps?

Do ISI installed a 3.0 version of GMotor in my computer without telling?

 

Well, that's probably because AMD improved their drivers a lot at some point last year, AMD performance on rF2 used to be abysmal before that. Or because Loch Drummond was updated, if I remember correct it brought big FPS gains too. Too many factors has changed to know for sure... But doesn't change the fact that FPS in rF2 is very much GPU dependant only.

 

If it was true, why at Loch Drummond I have about 70% more fps than in the chart posted above (55 fps at race start, starting last with 29 AI, then 65 min 85 max in the first race lap) with a i5 4670 and a 760, maxed in-game graphic details, AA level 4, but not FXAA, single monitor 1920x1080?

In other words, when the hardware is not well balanced, sometime the bottleneck is the GPU, but sometime it is the CPU ... it is matter of the weaker hardware component, and the hardware balancement is the key indeed for the best possible performance.

Another example ... my old machine was a e6600, in XP with 2 GB RAM, and a 4870 ... in such machine the bottleneck was the CPU, as the graphics engine programmer (Joe) told me.

 

A pc is basically a calculator on steroids.

To see how good a 'calculator' you have,calculate pi to 1 million places.........

http://www.superpi.net/Download/

My server pc with 2x dualcore zeon with 8gb of DDR2 does it in 30 seconds,quite slow.

 

I bet on the second hypothesis. And maybe it's because there's no solution. I drop hours and hours to the garbage can trying to extract some performance through Nvidia Inspector and no success at all. The staff statement about how good rF2 SLI performance sounds like a joke to me.

 

I'm surprised to read this especially when this comes from you.

60hz LCD = 60fps = 16,7 ms
100hz LCD = 100fps = 10 ms
120hz LCD = 120fps = 8,3 ms
144hz LCD = 144fps = 6,9 ms

LightBoost 100hz @ 100% = 100fps = 2,8 ms
LightBoost 120hz @ 100% = 120fps = 2,4 ms
LightBoost 100hz @ 50% = 100fps = 2,3 ms
LightBoost 120hz @ 50% = 120fps = 1,9 ms
LightBoost 100hz @ 10% = 60fps = 1,7 ms
LightBoost 120hz @ 10% = 60fps = 1,4 ms

It is 2015 and If 60 hz = 60fps= 16,7 ms is a goal to the ISI please let me know and I find other platform as my hobby.