Game Development Community

FPS of TGE 1.4 is MUCH lower than FPS of Tribes 2?

by Sammy · in General Discussion · 08/19/2006 (7:55 pm) · 12 replies

Hi all,

I am learning TGE 1.4 and reading GPGT.
TGE 1.4 has built-in tools and GPGT are well written.

However, as I ran the sample program, I found that
FPS of TGE 1.4 are VERY low for a few simple objects (some simple numbers and cubes only). FPS is around 30 ~ 40. Sometimes FPS was 20!
I could not imagine if I had a large indoor and outdoor scene and large no. of characters running on same screen (FPS will be VERY VERY low that is not playable)!!!
I also found that the final Maze prg does not run smoothly.

I remembered when I ran Tribes 2. Its refresh was MUCH smoother on SAME PC!
Tribes 2 has large terrain and many characters.

I asked this question on
http://www.garagegames.com/mg/forums/result.thread.php?qt=49380
But, up to now, there is no reply (2006/8/20 10:58am)

While other 3d engines are talking about 100~200 fps, how could we compete with them for doing the same things?
I really don't want to learn things that are finally not usable to make playable games.

What do you think?

Thanks in advance.

#1
08/19/2006 (8:00 pm)
Well for one the human eye can't see fast frames.

Also you didn't list the hardware your running it on?

Tribes 2 was written for an older API. Torque Game Engine is not tribes 2...


Bottom line.. TGE, TSE, TGB all have very good framerate on modern hardware.

Tribes 2 does not put money in your pocket... tribes 2 is an old game... stop worrying about tribes 2.
#2
08/19/2006 (9:19 pm)
Sammy,

There are some other possibilities to consider. If you are running a compiled version of TGE 1.4, it's very possible that it is a Debug build rather than a release build. Debug compiled executables run dramatically slower than release builds, due to extra instructions and performance/problem analysis stuff.

Also, if you are running TGE in Direct3D render mode, you will experience a significant slowdown.

And it's possible that you are running your Torque app at a higher resolution than you did previously with Tribes. Or you might have AntiAliasing and Ansiotropy settings cranked up to 8x and 16x.

You might also find that Tribes 2 has smaller texture maps, lower polygon budgets, and more aggressive LODs than the TGE missions you're comparing to.

And if you've got old OpenGL drivers, that might be an issue as well.

Have you run Tribes 2 on that machine lately? A direct comparison will probably be impossible, but you might be able to get a fairly close approximation.
#3
08/19/2006 (10:46 pm)
TGE is written for older hardware Voodoo 3 and up really. So it doesn't work efficiently with modern GPUS. Modern GPU's like fewer batches of higher polycounts, 200 polys or 2000 polys makes little difference on modern hardware.

Older GPU's prefer TGE's way of doing things where you get hundreds of batches of low polycounts. Like TGE terrains can have 1000's of them split into tristrips and fans of maybe 10 triangles each. All incredibly inefficient on modern hardware. The same with BSP, which allthough great for collision and lighting, on older engines like TGE leave you CPU bound when rendering.

Same with DTS, load up a mesh in showtool pro and see the tristrip view.
From looking at the TSE milestone list, it does look like they are addressing these issues. Most modern engines only use a BSP structure leaning towards a more polysoup solution for the visuals.
#4
08/19/2006 (11:06 pm)
@Randy "Vashner" Anthony: I am using AMD 2200 512MB + Geforce FX-5200 128MB,
window xp sp2, running GPGT samples at 800x600,32 bit color
I am not worrying about Tribes 2, I am just worrying about the FPS of TGE 1.4 of my final game.
http://www.garagegames.com/products/1
Quote:The ground-breaking engine behind Tribes continues to evolve and remains unparalleled.

Geforce FX-5200 was popular two years ago. I think babies born 2 years ago should be considered modern man. Window XP sp2 should be modern OS. I don't think everyone has
Geforce 7/8 series installed.
There will be no more update of TGE 1.4 , so TGE 1.4 is not for next generation hardware.

@Aaron Ellis: I run the GPGT samples directly on CD. I think it should be release version because the author should show the capabilites of TGE 1.4. I will ask the author about this.

@Adrian Tysoe: Do you mean that TGE 1.4 is inefficient for today development?
I am going to invest time on learning it. Why GG are still promoting TGE 1.4 to new customers?
I am confused...
#5
08/19/2006 (11:18 pm)
TGE is good for a broad range of hardware, TSE is for modern hardware. TGE isn't as efficient as it could be because it has to be compatible with really old computers going back to the tribes days. That doesn't mean it will run worse on a new computer, it will run very well, it just doesn't make as good use of modern hardware as it could.
#6
08/20/2006 (2:30 am)
Sammy, I haven't run the GPGT samples yet, but I believe the exe that comes with it is a debug version. Easy way to tell? Is there a black console window that hides behind the game window? If so, it's a debug version.
#7
08/20/2006 (5:44 am)
Sammy,

Do you run the executable directly from the GPGT CD? Meaning, whenever you want to run an example, you pop the CD into your system, browse to the proper folder, find the TGE.EXE file (or whatever it's called) and double click it? If this is what you meant, then that's probably a big part of the problem. I've created TGE "live CDs" like that for demonstrations of my software, but they run a LOT slower than running from hard drive. If this is what you're doing, you shoud try installing/copying the files onto your hard drive first.

However, If you meant that you installed the examples onto your hard drive and then ran the examples and experienced a slowdown, then the issues I described above might explain the slow performance.
#8
08/20/2006 (9:10 am)
@Sammy: I just happen to use a geForce FX5200 for mid to low end system testing. On that test system it is common to have a 20-30 FPS difference between release and debug versions depending on the situation. Debug builds run somewhere between 20-40FPS, and release builds usually run around 40-60FPS. This is on an AMD 3000+ (2.0GHz) CPU with 512MB of ram.

So if you are getting 30-40FPS I would assume you are running a debug build. You should get around 50-60FPS with a release build. It makes sense for the book examples to use a debug build as you have more options for developing; even if you are just scripting. One example is the debug render modes which are handy when developing.
#9
08/20/2006 (5:42 pm)
Thanks a lot, all experts.
This community is really helpful.
I ran the examples on harddisk. I will try to run codes using TGE 1.4.

By the way, compiled TGE 1.4 bought from GG should be release version.
If we want a debug version, do I need to do some settings and recompiling the source??

Thanks in advance.
#10
08/20/2006 (7:59 pm)
You can setup your dev to compile both at the same time if you want to. I did at one time just to find a problem.
#11
08/22/2006 (12:37 pm)
So is there a FPS boost by switching to TSE?

Thanks.
#12
08/22/2006 (12:48 pm)
@Ben,

I think that depends on the hardware youre using, but once it is officially released I think it will have a MARKED improvement on modern hardware.

Also very little to nothing seems to be changeing with Torque Script, so learning it on TGE will be benificial for when TSE comes to full release.

The Trusted One