FX5200 in linux low performance

[Benit]

Hello,

after some tests, i have reached the theory that exists some problem with linux nvidia drivers and fx5200.

The problem that it has is that the fx5200 run's slow in linux (in windows with the last drivers and directx 9c in quake3 i have 214fps and in linux with de same config (agp speed, fastwrites, sba), only have 140). And my surprise was when I was do the same tests with a gf4mx440 that in windows have 200 fps and in linux 214fps.

Anyone have the same problem? Anyone have found the solution?

PD: sorry for my low english skill :/

[energyman76b]

Hi,

in ut2003 my gf fx5200 is not faster than my gf 4 mx 440.

BUT there are a lot of additional graphic effects with the 5200. So, you migh have a look, if the 5200 has some fancier graphics.

Oh, and you can not compare directx+win with opengl+linux. You have to use ogl on both sides.

[LordMorgul]

Windows Q3 runs openGL as well as Linux, so despite having DX9.0c installed it is more or less irrelevant (he's used a valid test game). I haven't seen a similar problem with the cards I've used with Q3, but have not tested a 5200 under linux. That is a huge performance hit however, so I wonder if you have some hidden settings different. You did verify things like FSAA and aniso being the same right? Color depth?

Keep in mind some settings in Q3 won't be active even if you have them set.. and require restarting Q3. Its best to start it up, exec the config to test with, quit, and return to the game for the test.

[energyman76b]

Hi,

I only have the quake3 demo, but that demo was faster with my fx5200 than the mx440.

[Benit]

the way that i make the test's are:

1 - mount the fx5200 & going to linux
2- i've deleted q3config
3- timedemo 1, demo four = 135 fps
4- run again the timedemo = 140 fps
5- mount the gf4mx440 & going to linux
6- delete q3config
7- doing the timedemo = 214 fps
8 - going to windows, delete the q3config
9 - timedemo in windows = 200 fps
10 - mount the fx5200 & going to win & delete q3config
11 - timedemo in windows = 212 fps

the pointrelease of quake in win & linux is 1.32 and the colour depht is 16 bit

[how]

generally speaking the linux driver in all versions I used are alot slower in many cases than windows driver. I compared with viewperf 6 and quake3. Hope Nvidia will share some light on this issue as to why?

[goland55]

I'm also having problems with slow performance with the FX5200. I have the 128mb version. I haven't done any back to back testing but I've had to do some major graphics quality reduction/tweaking to get Enemy Territory to run fairly consistently above 45fps. My windows computer has an old Asus v7700 with 32mb and seems to not run too much differently than my FX5200 speedwise. I'm not quite sure what framerates I should be getting but I would think I'd be able to sustain something higher than 80 fps with this card.

Here's some outputs of my setup

cat /proc/driver/nvidia/cards/*
Model: GeForce FX 5200
IRQ: 10
Video BIOS: 04.34.20.22.00
Card Type: AGP

cat /proc/driver/nvidia/agp/*
Fast Writes: Supported
SBA: Supported
AGP Rates: 8x 4x
Registers: 0x1f000e1b:0x1f004302
Host Bridge: PCI device 10de:01e0 (nVidia Corporation)
Fast Writes: Supported
SBA: Supported
AGP Rates: 8x 4x
Registers: 0x1f00421b:0x00000302
Status: Enabled
Driver: AGPGART
AGP Rate: 8x
Fast Writes: Disabled
SBA: Enabled

XF86Config

Section "Device"
Identifier "geforcefx5200"
Driver "nvidia"
#VideoRam 131072
# Insert Clocks lines here if appropriate
Option "RenderAccel" "true"
vendorname ""
boardname ""
EndSection

The fastwrites disabled thing caught my eye... though I believe I enabled fw in the past and didn't notice much difference.

I'm not sure how my config matches up to the original poster's but I think the issue is pretty much the same.

[Reeve]

Same problem here, FX 5200 performs much worse than it should (I know it's not a great super-fast card or anything, but I mean like "OMG My GeForce 2 MX was almost as fast" type of slow ) I've tried everything including enabling Fast Writes and SBA, nothing helps. My friend however also has an FX 5200 which performs beautifully, which the only difference I can think of is that his was purchased back when the FX series first came out, mine was bought less than a year ago. Anyone have any ideas?

[whig]

He might have the ultra as these benchmarks show (on Windows anyway),

http://www20.graphics.tomshardware.c...charts-06.html

[Spyke]

The shields covering the power cores in UT2004 also look like crap, (like a spotty shell), no effects there at all, nothing like the old geforce I had. Any ideas?

[Wolfman [TWP]]

My suggestion would be to run your 'nvidia-settings' (Nvidia Panel) utility, and check that in the OpenGL Settings section, that the 'Sync to VBlank' is not checked.

As I replaced my GF4 MX440 with a GF FX5700 (Here at the Office) and I ran the glxgears, only to find that it showed only 85-90fps?!. Where my GF4 MX440 showed over 700fps (Which is normal for this card). So I changed the VBlank setting to off, and now the fps's are up around the 2900+ mark. (I know that glxgears is not a benchmarking utility, but it does give you indications if the card is running the way it should or not.)

Wolf

[Benit]

I've looked in nvsettings and I don't have Sync to Vblank checked, and glxgears runs "normal"

4176 frames in 5.0 seconds = 835.200 FPS