Game Development Community

T3D performance on low-end machines ?

by Orion Elenzil · in Torque 3D Beginner · 08/13/2009 (3:46 pm) · 12 replies

howdy all you dashing explorers in the T3D beta program !
how's the weather out there ?

i've got a Q which i'm hoping to get answered through yr charity -

has anyone tried T3D on a low-end machine ?
for example a contemporary $200 eMachine

or maybe a machine that's five years old or so.

OR, if it's possible to load up identical or very very similar missions with identical or very very similar rendering quality in both TGE 1.5 and T3D on the same machine and see how they compare, that would be swell too.

many tia,
ooo


#1
08/13/2009 (4:11 pm)
I was able to run T3D with basic lighting on my Eee PC netbook. I only tried the blank room mission though.
#2
08/13/2009 (4:14 pm)
hey James, thanks for the reply.
that's cool that it ran; how was the performance ?
#3
08/13/2009 (4:20 pm)
Not too great, but it was usable, even with the WorldEditor open (though the trackpad wasn't very efficient). I was also running a debug build though.

Maybe later I'll post some fps with a debug / release build.
#4
08/13/2009 (4:22 pm)
if you've got the time & energy that would be awesome.
i'm mostly concerned w/ user experience, so release build is the most interesting.
#5
08/13/2009 (4:42 pm)
I got Warrior Camp to run on an old Intel 945 ... at about 2fps, but it did run.

My 9700GS only pulls WC at 15fps in advanced lighting and 30fps basic (windowed). WC isn't very performance friendly.
#6
08/13/2009 (8:16 pm)
If you design with Basic Lighting in mind as a fallback for the low end you should get performance that at least equals TGEA.

There is alot of back pocket optimizations going in between now and the 1.0 release... so expect more performance in AL and general improvements.

Quote:My 9700GS only pulls WC at 15fps in advanced lighting and 30fps basic (windowed). WC isn't very performance friendly.
Gah... that 9700GS should be faster than that in BL mode.
#7
08/14/2009 (7:01 am)
My 7900GS is faster than that in BL mode ... except in WarriorCamp, it's just not optimized.

I've got a test forest (600 trees) which gets 18fps AL, 40fps AL noshadows and 65fps BL.

In my old TGEA town model test I got:
TGEA DRL --------------- 114fps
T3D AL ----------------- 26fps
T3D AL (no sun shadow)-- 57fps
T3D BL ----------------- 94fps

One thing I have noticed about T3D is that it is hugely more playable at lower fps than TGE/A. In TGE/A I'd notice lag if the fps dropped anywhere near 60. T3D gets the same noticeable amount of effect at below 30.
#8
08/14/2009 (8:21 am)
In some moments I've noticed some suspicious behaviour on that sense, which made me doubt of the metrics() command accuracy.
#9
08/14/2009 (9:49 am)
thanks for the numbers, guys.

Tom, i'm glad to hear there's optimizations going on.

Steve, sounds like currently it's slightly less performant than TGEA (which is itself slightly less performant than TGE ?) which is a little bit of a bummer for using it w/ a more casual audience. That's very odd what you say about it lagging less at low FPS than TGEA tho. I'm not sure i comprehend that. Are you getting the FPS numbers from FRAPS or from the game itself ? Is this a single-player or networked situation ? I guess a change to how input & messages are handled in the game loop might explain less sense of lag at the same FPS.

thanks again,
ooo
#10
08/14/2009 (10:09 am)
Quote:
I'm not sure i comprehend that.

I always put it down to how the tech worked .. dodgedly. I found that TGEA is actually HUGELY more performant than TGE, even with all the extra stuff - and T3D has a lot more extra stuff on that.

SinglePlayer, and I tested with both Fraps and Metrics, same result.

My "huge town model tests" don't quite follow the same format in T3D as they did in TGEA. New drivers mean I use a (slightly) bigger resolution, and T3D has multiple layers of larger terrain textures (multi normal maps, details, etc) that kinda stuff all adds up on a lower system. Even having the same things in game, there's a lot more "stuff" over TGEA. But it's about as representative a comparision as I could get.

I wouldn't use x1024 textures anymore as it kills fps on lower end machines. It's the dynamic shadows of AL which really munch though.
#11
08/17/2009 (10:12 pm)
The problem with comparing TGE to TGEA or T3D is that it is very difficult to get a scene with exactly the same art resources and effects enabled.

For example, even though Stronghold in TGE and TGEA are very similar in the artwork and effects, the one in TGEA uses the TGEA equivalent of WaterBlock which does far more intensive/costly effects (reflection and refraction) and there is not really an easy way to force the TGEA WaterBlock back to the same level of effects as the TGE one.

The same is also true of the dynamic lighting systems (which can have a huge effect).

In general, TGEA and T3D will out perform TGE if you manage to tune the artwork and effects to be exactly the same (no cool water, no specular, no dynamic shadows, no normal maps) since it has a far more efficient rendering pipeline and it uses the graphics APIs that are tuned appropriately to their respective OS'es.
#12
12/13/2009 (11:19 am)
hi everyone:)
i bought last week the alpha version.. and i made a large level for testing all the new nice things of T3D. Also for test the power of T3D.

This level contends:
-15 ai , 1 player ( all Torque models)
-around 1500 trees or more
-a lot streets and a realy nice lava effect, made with the path editor
-running Day night system
-wind , water block and water plane
-light rays and post eff on
-around 150 models in level
-a lot lights ( for all the night street lamps)
-a lot particle on running

I use a atlon phenom 4 core cpu (2.2Ghz) and ati 4870 GC, 4 Mb ram
Windows Xp 32 bit system.

I play the level with 30 fps :), and i be sure i can put in more models and things, without get lag. (just have to make my self made Models better).

Just hope this info helps the Community a bit, and sry about my small English:)