Need some opinions on GF3 performance with Jedi Knight II

[Rio Rebel]

I'm running Jedi Knight II (Outcast) on a 1.2 T-bird, 640mb of memory (512+128) and a Geforce 3. Currently, I have it clocked at Ti500 speeds, 240 and 500, but it's not a major difference in this game because (as Anand's testing showed) this game is more CPU limited.

Problem is, I don't seem to be getting the performance I expected. I am running at 1280x1024, high detail settings, with Quincux AntiAliasing and 4x Ansiotropic filtering, on detonator 28.32 drivers. I can't break 35fps.

In Anand's roundup of sub-$200 cards, he was getting in the 80fps range at high detail in this game. The only difference is the CPU, which I know makes a difference but I wouldn't think THAT much.

Any opinions?

 

[Rio Rebel]

LOL...

If I want an opinion on finance, politics, or philosophy, I can't keep up with all the responses. But when I ask a question that some of these folks really know about, I can't get an opinion.

Go figure...

 

[jiffylube1024]

You're running a pretty high resolution to be having 4X AA on (even if it is Quncunx). I run it at 1024X768 res with no AA just to keep it at >100 fps always. I'm running it on P4 2.2 Ghz with a GF3 Ti200 @ Ti500 levels ...

Make sure Anisotropic Filtering is off and set lighting to "simple" instead of "complex" for best performance. Also, you might want to try lowering the resolution a notch (to 1024) if you want to keep AA on and a high framerate.

 

[Insidious]

AA knocks the crap outta my GF3 also. (especially if > 1024)

I'd suggest trying 1024 X 728 X 32 with no AA or ansiotropic filtering.

Which ever is the most satisfying....... well, that's gonna be a subjective choice.

I don't have JK2, but I usually experiment with OpenGL vs. DirectX as well.

(I'm not really surprised you are a little slower with those settings. They are pretty demanding)

 

[Rio Rebel]

I knew 4x AA was taxing, but I thought Quincux was not that demanding.

Apparently it makes a difference. Maybe when I get my new Athlon XP 2000+ it'll be playable at 1280x1024 with all my eye candies...

 

[BFG10K]

Problem is, I don't seem to be getting the performance I expected. I am running at 1280x1024, high detail settings, with Quincux AntiAliasing and 4x Ansiotropic filtering

That's a rather ridiculous assortment of settings to have and I'm not at all surprised that your video card is completely hammered by them. Just how fast are you expecting your GF3 to be? :Q

JK2 ran just fine on my old Ti500 at 1152 x 864 x 32 with maximum image quality settings in the game (except for shadows). Of course I had anisotropic and FSAA completely disabled.

 

[Kraid2xd]

Anyone get like really crappy/blocky shadowing on the actual character models themselves in JKII? I have a gf3 ti200 and no matter how many times I play around with the settings, they still keep appearing...running on dets 27.42 @1024x780 with everything turned on high, all graphical extras checked... If I disable all shadows, it goes away, but who wants that!?!?

 

[Rio Rebel]

<< That's a rather ridiculous assortment of settings to have and I'm not at all surprised that your video card is completely hammered by them. Just how fast are you expecting your GF3 to be? :Q >>



Why is this ridiculous? It is my understanding that Quincux is roughly the same as 2x AA in terms of performance hit, and I set the Ansiotropic filtering at the middle setting. I know I'm taxing it somewhat, but I'm not maxing out at 4xAA or full Ansiotropic filtering.

In Anand's benchmarks, he ran this game at 1280x1024, High Quality settings on an XP2100+. , and got 98fps. I am running on a 1.2ghz and turning on Quincux and 4x Ansio, and I'm getting between 30-35fps.

It just surprises me that I'd see THIS much difference. I knew running 4x AA would cripple performance, but not Quincux (as least not as badly).

 

[IndyJaws]

<< Anyone get like really crappy/blocky shadowing on the actual character models themselves in JKII? I have a gf3 ti200 and no matter how many times I play around with the settings, they still keep appearing...running on dets 27.42 @1024x780 with everything turned on high, all graphical extras checked... If I disable all shadows, it goes away, but who wants that!?!? >>



Yep, I'm getting the same thing. Also in caves, the banding gets pretty bad. I thought maybe I had my vid card memory oc'd too high and brought it down, but it didn't really help.

 

[merlocka]

It is my understanding that Quincux is roughly the same as 2x AA in terms of performance hit, and I set the Ansiotropic filtering at the middle setting.

Quincux will cut FPS in about half in BW limited benchmarks, so that 90fps just became 45. Add 4x anisotropic and 35 sounds right.

 

[Rio Rebel]

Just did some further testing:

Turned off all AA, turned off Ansiotropic filtering, turned off shading. 56fps Lowered resolution to 1024, and still can't break 57fps.

Not complaining, just wondering what's up. Maybe Anand tested his "1200mhz" processor as an XP scaled back to 1.2, and my regular old 1.2 lags way behind without the "XP" architecture.

I don't know.

 

[merlocka]

owered resolution to 1024, and still can't break 57fps.

Lemmie take a guess... Your running a 1.2Ghz Tbird on a KT133 platform (100MHz fsb w/ SDram)? This would make a big difference between Anand's test bed (XP running DDR ram).

 

[Rio Rebel]

Good guess, but not in this case. It's at 133fsb with 512mb of DDR (2100).

I'm convinced that the "Quantispeed" architecture of the XP making a difference (such that his 1.2ghz would yield significantly faster fps than my 1.2). It's the only thing left that makes sense.

I'll find out when my Athlon XP 2000+ arrives this week.

Of course, any other suggestions welcome.

 

[BD231]

Dude, your nuts to be using FSAA on anything other than a Ti4600, especialy with that processor. All you have to do is turn off quinox or whatever its called and it'll run smooth.

 

[BFG10K]

JK2 is a very CPU limited game and it's quite possible that the Palomino core is giving it much more power than the TBird core is. Of course as you said, you'll see soon enough when you get your 2000+ XP.

Turned off all AA, turned off Ansiotropic filtering, turned off shading.

Did you also disable vsync, com_maxFPS and shadows?

 

[Rio Rebel]

bdog, thanks for your input. However, you're mistaken about anti-aliasing. There are plenty of games that the Geforce 3 can run anti-aliasing at very playable rates (even 4x in some cases).

bfg10k, why would it help to turn off v-sync (which I've already done)? At 85hz, I'm not going to get higher frame rates than that anyway, I would think.

I'll post when I get the new cpu - it may interest some people to know how much of an effect the Palomino architecture has in this game (assuming I can set the cpu at 1.2)

 

[DefRef]

Even Quincunx AA deals a serious hit on fps when you're dealing with some games, esp. Q3-based ones. The outdoor winter level of MOHAA had me tanking into the 15-20 fps range when I was set at 1024 rez with all options full-up, 16X aniso and Quincunx. (GF3Ti200 & 1900+) When I turned of the AA, rates leapt to the 40-50 fps range.

JK2 has shaders on the textures and those are extra passes that the card has to render - that hurts fps. The volumetric shadows just look lousy, so you're better off going with simple shadows. AA is less important in shooters, so loose that and up rez.

THPS3 looks like ass w/o AA, but enabling it doesn't cause any problems, even on the huge levels, so experiment with settings (use NVMax or an app that allows you to save presets) and see what works best for you.

 

[Leon]

Here is one for you - disable "very high detail" texture slider, instead set it to "high". You should see massive improvement there.

 

[Rio Rebel]

Put in the Athlon XP 2000+ today.

Framerates jumped to 72.4 fps with the goodies turned off. With Quincux and Ansio at the original setting in my first post, it runs at 43.7fps.

Still can't get near the 90's like Anand did, possibly because of DDR 2100 and a slight difference in cpu's, but this is close enough.

I'm off to "test" a little more.

(oh, my board doesn't adjust cpu speeds in bios, so I'll have to wait until I feel like playing with dip switches to see how the Palamino 1.2 compares to the T-bird in this game).

 

[butch84]

I think it may still be some hidden setting for you. You can see what i have in my rig at the bottom of my post, and i get about 90- 105fps running @1024x768. I dont think upping the res shuld make that much difference, but i could be wrong . . . .

 

[BFG10K]

bfg10k, why would it help to turn off v-sync (which I've already done)?

Enabling vsync can sometimes be much more detrimental than simply causing a framerate cap. In certain situations it can reduce your current framerate to half of your monitor's current refresh rate, among other things.

BTW I'm glad you like your new CPU but keep in mind that while JK2 is quite CPU limited, certain levels (especially the swamps) are extremely taxing on the video card.

 

[Roxxor]

how do u display FPS in JK2?

 

[Rio Rebel]

The Swamp level is DEFINATELY taxing. The detail is amazing - down to the little sizzles on the light saber ever time a rain drop hits it.

Roxxor, you can't just display FPS in Jedi Knight II (at least to my knowledge). You can, however, run a timedemo to measure the FPS. Anand's review of sub-$200 video cards explains how to do it.

By the way, I installed the latest detonator drivers (29.11, I think?) and I'm getting in the 80's now. Still experimenting.

Thanks for all the suggestions/explanations.

 

[BFG10K]

how do u display FPS in JK2?

Hold down shift and press ~, then type cg_drawFPS 1.
Also you can use com_maxFPS 0 to disable the framerate cap.

The Swamp level is DEFINATELY taxing.

It sure is - after coming from an average of 80 FPS in the entire game it was a real slap in the face to start getting framerates in the high 20s. It was only after I dropped down to 1152 x 864 x 32 from 1600 x 1200 x 32 that the framerates again increased to an acceptable level. It's amazing how much that level pounds even a GF4 Ti4600. :Q

 

[Rio Rebel]

<< Roxxor, you can't just display FPS in Jedi Knight II (at least to my knowledge). >>





<< Hold down shift and press ~, then type cg_drawFPS 1. >>



Like I said...er, ask BFG10K about it.