[Rga_Noris]

I got bored today and did a quick test to see just how much actual performance is gained with usage of LODs. I've always wondered if it really made that big of difference.

Here are my flags:  -spec -glow -env -mipmap -nomotiondebris -missile_lighting -normal -3dshockwave -no_vsync -dualscanlines -targetinfo -orbradar -rearm_timer -ship_choice_3d -weapon_choice_3d -3dwarp -warp_flash -post_process -fps -show_mem_usage  -32bit -ambient_factor 35 -ogl_spec 20 -spec_exp 15 -spec_point 1.2 -spec_static 1.5 -spec_tube 1.5

Using hte MediaVP's and a modded Uly

The test went as follows:
LODTest1 was comprised of 30 Hostile Ulys(with their LODs) vs. 30 friendly Ulys(also with LODs)
About 80-90fps for about 10 minutes of play.

The second test was the same except with Ulys with no LODs.
About 60-70fps for about 10 minutes of play.

So it seems to be obvious that it does have a positive effect, but I think the test may be tainted. All I did to remove the LODs was duplicate the Uly and set its $Detail Distance to (0, 80000, 80000, 80000). The player is placed only 3000 meters away so no LODs should be active in the NoLOD Uly. But my question is did I just disable mipmapping as well? I dont want to test the difference between mipmapping on and off, I want to isolate the LODs.

Any help would be appreciated.

[headdie]

To get more reliable results you could just open the pof in pcs, delete the lods, save as a new file and modify the duplicate tbl entry refering to the new pof and rerun the test

[FUBAR-BDHR]

Also remember that the LOD values you place in the table != the LOD distances in the game.  If you have a good graphics card and are set on very high you are looking at up to 16x times the lod1 distance before it kicks in.  (not positive on the exact number and too damn buzzed to check)

[Water]

Yup, x8 at the highest. Not much you can do about mip maps though as the game uses them automatically.

[Nuke]

or you could just play nukemod

[Rga_Noris]

Just to make sure I'm understanding this correctly, mipmaps wll kick in regardless of LOD level? If that's the case, excellent. That means the only difference between the two test would be the LODs, which is what I want.

I'm aware of the detail modifiers, which is why I did a quick table edit of 80000. Regardless of the modifier, it should not kick in. I'm still going to do the PCS edit, though. This will make it easier for the next series of test, aimed at determining the usefulness of each LOD level.

[The E]

Well.......I think the outcome will be a foregone conclusion there as well. Under normal conditions, with every setting maxed out and assuming average mission distances, LOD3 and LOD4 will not see much use.

[Rga_Noris]

Oookay, so I just did the following test:

Same mission with the 60 Ulys running around. They are invincible so the ship count never varies. Each test was run three times and the average FPS is estimated by glancing at the FPS counter periodically (I know, I know...not really scientific)

Test 1 (normal Ulys, detail0, detail1, detail2 present) Player is 3000 meters away.
Avg FPS ~96

Test 2 (Modded Ulys, detail0, detail1 present, detail 1 active throughout play)
Avg FPS ~92

Test 3 (Modded Ulys, detail0, detail2(renamed to detail1) present, detail 2(1) active throughout play)
Avg FPS ~96

Test 4 (Detail0 only, no LODS)
Avg FPS ~94

Now I fully understand that these are result on MY machine, and they do not represent everyones possible outcome. However, it appears that the performance gain between detail0 and detail1 is hardly even existant, and detail2 only provides a moderate increase. As mentioned before by The E, the lowest detail meshes would hardly ever be used given a normal mission.

So is it even worth taking the time to make LODs? At least for fighters? Cap ships could be a whole different story, given their vastly different poly counts. ( and no, I will not make a 60 ship Orion battle )

[The E]

As someone stuck with a rather low-end machine, I would most emphatically say "yes, LODs, even 3 and 4, are necessary".

During BP development, we're playing a lot with Steve-O's models. While they are beautiful, they are also horrifically unoptimized (no LODs, for starters). Getting them LODed has a noticeable effect on performance, so making these things should never be considered a waste of time.

[Rga_Noris]

Soooo im an idiot.

I kept putting the modified fighter01.pof in the wrong directory, so the test was ran all four times using a normal Uly.

Redid it properly and found a noted difference. Without any lods, FPS was around the low 80's. Upper 90's to 100 with the LODs.  That's a pretty significant difference.

It is worth noting however that this difference was absent when comparing detail0 to detail1. Only with detail2 did the performance boost become noted. Now this could be due to some effeciency issues with the model itself.

If anything, detail2 and up should take a high priority with modeler, as the largest performance boost was noted at that point. 

[Mahak]

Remembering of course you are using a model that would be considered 'low res' by today's standards... (unless you're not... whats the poly count of your modified uly?)

I would be interested in seeing some comprehensive tests of varying mesh densities... just if you're bored one day

[Rga_Noris]

Actually, I was thinking of making three to four 'ships' that would consist of just a single sphere. Each ship would be a different detail level, so that way I could completely avoid any issues with how/when the game draws LODs.

But the E makes a good point, especially considering that large parts of our little community here have older machines. While may argue that these slower machines are hampering the progress of the SCP, you could just as easily argue that by alienating a large portion of the community by leaving out LODs and other optimizations would put the SCP in its coffin.

[The E]

This is not really an SCP thing. The executable is (in theory) capable of running on the same hardware that was around when FS2 was released.

However, if you're making a shiny new mod, with models of great beauty and polycounts, taking into account the audience that you are making the mod for would be a good idea. If noone is able to play your mod with good framerates, it's your mod that will be blamed, not the SCP.

Also, Rga, please check your PMs.

[Iss Mneur]

This is not really an SCP thing. The executable is (in theory) capable of running on the same hardware that was around when FS2 was released.

Yes, for the record. If the hardware from that time has OpenAL 1.1, and OpenGL (1.5 I believe is FSO minimum) support, then FSO will run on the same hardware.  The problem is most hardware of that time either was Glide or DirectX, both of which are not supported anymore by FSO.

Potentially, FSO could run faster as a large number of bugs bug fixes and other optimizations have been added to the code base.

[MatthTheGeek]

as a large number of bugs and other optimizations have been added to the code base.

When we read you, sounds like you added bugs I guess it happens, but I'm sure that not what you wanted to say...

@ Rga : Try with something more resource-consuming than Ulys.

[Iss Mneur]

as a large number of bugs and other optimizations have been added to the code base.

When we read you, sounds like you added bugs I guess it happens, but I'm sure that not what you wanted to say...

Ya, that was not really what I wanted to say, but at the same time, the new features can add bugs or make other bugs happen or more apparent.