GeForce FX5600 glxgears benchmark

[foulmetal]

Do you all test glxgears at fullscreen ?

[tudorv]

Hi All,

Thanks for the replies! So I suppose there's nothing wrong with my setup (at least nothing fundamentally wrong). I use all the default settings, with AGP 4x.

Looking at graphics cards benchmark (like the one on Tom's Hardware), it seems like the performance of a Geforce FX5600 is inferior to what a GeForce4 Ti series card delivers. Don't know about GeForce3, but I wouldn't be too surprised if that's faster too.

Marcus, I couldn't agree more with the idea of a nice and fair database (how nice and fair it can be, that's another issue). Something similar to what 3D Mark provides would eliminate a lot of posts to this forum, and a lot of the existent confusion.

Would it be feasible to use SPECviewperf and this forum for such database? I don't think this forum is the ideal solution, but it's better than nothing.

Thanks,
Tudor

P.S. I didn't try SPECviewperf yet, but I'll do it tonight and let you know how it goes.

[pzgren]

Well, I guess I'll just keep trying to tweak things, this can't be the best my card can do.

This is really not HW-accelerated. This is near my old Riva128/TNT times in the past... Or you have a faked TNT card with a FX 5600 BIOS and Label. ;-)
There must be something in your driver setup wrong.

Look here --> http://comrad.linuxsys.de/?content=/treiber/benchmarks/
for approximately scores.

Look in /var/log/XFree86.0.log for errors/warnings. Search the lines for (or similar):
######################################
....
(II) Loading /usr/X11R6/lib/modules/extensions/libglx.so
....
(II) NVIDIA(0): AGP 8X successfully initialized
(II) NVIDIA(0): Setting mode "1280x1024"
(II) Loading extension NV-GLX
(II) NVIDIA(0): NVIDIA 3D Acceleration Architecture Initialized
(II) NVIDIA(0): Using the NVIDIA 2D acceleration architecture
...

And?

Marcus

[james]

Marcus,
Relevent sections from Xfree86.0.log looks like:

(II) LoadModule: "glx"
(II) Loading /usr/X11R6/lib/modules/extensions/libglx.so
(II) Module glx: vendor="NVIDIA Corporation"
compiled for 4.0.2, module version = 1.0.5336
Module class: XFree86 Server Extension
ABI class: XFree86 Server Extension, version 0.1
(II) Loading extension GLX
...
more stuff .....
...
(II) LoadModule: "nvidia"
(II) Loading /usr/X11R6/lib/modules/drivers/nvidia_drv.o
(II) Module nvidia: vendor="NVIDIA Corporation"
compiled for 4.0.2, module version = 1.0.5336
Module class: XFree86 Video Driver
(II) LoadModule: "mouse"
(II) Loading /usr/X11R6/lib/modules/input/mouse_drv.o
(II) Module mouse: vendor="The XFree86 Project"
compiled for 4.3.0.1, module version = 1.0.0
Module class: XFree86 XInput Driver
ABI class: XFree86 XInput driver, version 0.4
(II) v4l driver for Video4Linux
(II) NVIDIA XFree86 Driver 1.0-5336 Wed Jan 14 18:30:46 PST 2004
(II) NVIDIA Unified Driver for all NVIDIA GPUs
(II) Primary Device is: PCI 01:05:0
(--) Chipset NVIDIA GPU found
...
more stuff......
...
(**) NVIDIA(0): Depth 24, (--) framebuffer bpp 32
(==) NVIDIA(0): RGB weight 888
(==) NVIDIA(0): Default visual is TrueColor
(==) NVIDIA(0): Using gamma correction (1.0, 1.0, 1.0)
(**) NVIDIA(0): Option "NvAGP" "0"
(**) NVIDIA(0): Use of AGP disabled per request
(--) NVIDIA(0): Linear framebuffer at 0xE0000000
(--) NVIDIA(0): MMIO registers at 0xD0000000
(II) NVIDIA(0): NVIDIA GPU detected as: GeForce FX 5600XT
(--) NVIDIA(0): VideoBIOS: 04.31.20.66.00
(--) NVIDIA(0): VideoRAM: 131072 kBytes
(II) NVIDIA(0): Connected display device(s): CRT-0
(--) NVIDIA(0): Detected TV Encoder: NVIDIA
...
...
...
(II) NVIDIA(0): Setting mode "1280x1024"
(II) Loading extension NV-GLX
(II) NVIDIA(0): NVIDIA 3D Acceleration Architecture Initialized
(II) NVIDIA(0): Using the NVIDIA 2D acceleration architecture
(==) NVIDIA(0): Backing store disabled
(==) NVIDIA(0): Silken mouse enabled
(II) Loading extension NV-CONTROL
(WW) NVIDIA(0): Option "Rotate" is not used
(WW) NVIDIA(0): Option "CalcAlgorithm" is not used
...

Yes, I know that I have disabled AGP by using the Option "NvAGP" "0", but I still got much better results from my Ti4200 without AGP.

Last night, following sybourg's suggestion, I attempted to unload the module and then re-install the 5336 driver again. Really strange, but now I get the ' Unable to determine kernel module version' error when compiling, but I know that it worked at least once, because that is how I got the module that I have loaded right now. I thought people used to have problems with this type of error, but that ( if I remember correctly ) was before official Nvidia support for the 2.6.x kernels.

I'll keep plugging away and repost when I find something I did wrong.

[james]

To add to my last post, I reinstalled kernel sources and now the installer found version.h, so that worked. But still no better framerates, without or with AGP enabled. Does that confirm that the card is not using its hardware for GL rendering? Output from /proc/driver/nvidia/agp/status shows it enabled with 4x,2x and 1x available ( when I used the NvAGP=1 option ).

[pzgren]

Originally posted by james
To add to my last post, I reinstalled kernel sources and now the installer found version.h, so that worked. But still no better framerates, without or with AGP enabled. Does that confirm that the card is not using its hardware for GL rendering? Output from /proc/driver/nvidia/agp/status shows it enabled with 4x,2x and 1x available ( when I used the NvAGP=1 option ).

I see, first you have an slower, sorry , 5600 XT in your logs and you run it in 24 depth. Maybe try for a test 16bit, to see if this baby brakes heavy in at 24bit.

Hmm, but you wrote a Ti4200, now?!?!

-------------------------------------------------------------------------------------------------------
If Andy Mecham read this, why is the nvidia_drv module for a long time compiled against an old buggy 4.0.2 XFree86. Just a question... ;-) And is it possible to get
an simple driver nvidia (beta whatever, for testing, unsupported, on a secret site... :-) ) for the the future of Linux desktops

---> http://freedesktop.org/Software/xserver

to get this nice and eyecandy desktop ?

and the new XServer

Marcus

[tudorv]

So, I got to run SPECViewperf, and here is the summary of results:

===========================================
---------- sum_results/3dsmax/summary.txt
3dsmax-02 Weighted Geometric Mean = 5.118

---------- sum_results/drv/summary.txt
drv-09 Weighted Geometric Mean = 29.037

---------- sum_results/dx/summary.txt
dx-08 Weighted Geometric Mean = 34.064

---------- sum_results/light/summary.txt
light-06 Weighted Geometric Mean = 7.816

---------- sum_results/proe/summary.txt
proe-02 Weighted Geometric Mean = 7.460

---------- sum_results/ugs/summary.txt
ugs-03 Weighted Geometric Mean = 5.242
===========================================

The machine was an AthlonXP 1800+, 256MB RAM, GeForce FX5600 256MB, AGP 4X, running Mandrake 9.2 (kernel 2.6.2), XFree86 4.3.99, 1152x864x24bit, NVIDIA drivers 4620.

Normally I run GNOME, but for this I ran a barebone X session (only xterm and twm), and the usual services (postfix etc.).

I wonder what results other people got from this benchmark? I'm curious how it compares against similar setups.

I haven't much of a clue regarding graphics benchmarking, but judging on what Google gave me back, it seems like this one is highly regarded, though targeted at professional applications. And the only one to run on Linux/UNIX?

Tudor

[pzgren]

Ok, Ive found my old viewperf results for my last "MSI Ti4200 TD8x 64MB" on CD. :-)
I had testet it with drivers 4496 - 4620beta - 5328, The 4620beta were the fastest
and here shown. My ASUS FX5700 is in bench-work, cause reinstalled Fedora, software...

FC rawhide, 2.6.x, Barton 2800+, 512 MB DDR 333, SATA Maxtor, Abit NF7-S V2, GNOME 2.4.2 desktop 16bit. I think.

---------- sum_results/3dsmax/summary.txt
3dsmax-02 Weighted Geometric Mean = 9.823

---------- sum_results/drv/summary.txt
drv-09 Weighted Geometric Mean = 53.216

---------- sum_results/dx/summary.txt
dx-08 Weighted Geometric Mean = 50.636

---------- sum_results/light/summary.txt
light-06 Weighted Geometric Mean = 12.104

---------- sum_results/proe/summary.txt
proe-02 Weighted Geometric Mean = 11.670

---------- sum_results/ugs/summary.txt
ugs-03 Weighted Geometric Mean = 5.183


Marcus

[briansladecek]

Originally Posted by james

To all who have posted their benchmarks, thanks, it confirms my suspicions that I am getting really bad framerates for glxgears from my FX5600. I had to replace a Ti4200 that went bad...

There's nothing wrong with your card. It's just that in all the benchmarks I've seen at hardware reviews, the 4200 beats the FX5600 in most cases. In fact, before I bought my Ti-4600, I compared the newer FX boards with the Ti-4600, and only the FX5800 and above would beat it in all the benchmarks.

GLXgears is not very useful, but a rough estimate. With 32-bet depth, I get ~4500 fps, and at 16-bit depth I get ~7000fps. The glxgears measurements you report for your FX card are accurate. In real world benchmarks, hardware review sites put your FX5600 card right at or below the 4200. It's too bad you can't pull the 'ole 4200 outta the trash can and whip out a soldering iron.

[Kamel]

glxgears isn't made for a real benchmark. it can be affected by too many things to be considered "constant".

the main reason for glxgears is to tell if you have DRI working or not. it's easy to tell the difference between 700 and 2,800FPS.

with my radeon 9800 pro aiw i got about 2,600 on my 2.4 kernel, but then again, the ati drivers are ass, so it was probably performing worse, lol