Pitty performance on OpenGL benchmark

enorbet · 03-25-2014, 04:03 PM

Greetings
the xscreensaver test seems quite limited to me in that much of the advance in graphics in recent history has been in texture rendering and I don't see much texture in xscreensaver. Like anything else the validity of a test is determined by what one uses theirs' for.

I'm getting 156 FPS on the xscreensaver test with an nVidia GTX760 and nVidia driver, fwiw.

GazL · 03-25-2014, 04:15 PM

I have a Nvidia GT-140, which is essentially a re-released/re-badged 9600GT. Decent, but nothing fancy by today's standards.

__GL_SYNC_TO_VBLANK=0 /usr/libexec/xscreensaver/glknots -fps -delay 1
gives me about ~ 800fps. I guess it's less cpu intensive than glschool

__GL_SYNC_TO_VBLANK=0 /usr/libexec/xscreensaver/glplanet -fps -delay 1
gives me ~1380fps.

GazL · 03-25-2014, 04:21 PM

enorbet is right of course, There are many facets to OpenGL, and these are no replacement for proper benchmarking. As I said above, I only mentioned then as they're already installed and moisespedro said he didn't have anything else. You'd likely notice the difference between hardware and software rendering, but they're probably not much use for anything else.

genss · 03-25-2014, 05:34 PM

this should be good for quick tests
http://www.geeks3d.com/gputest/

all except triangle are shader intensive
triangle is basically a swapbuffer test (not a real benchmark, kinda like glxgears)

ReaperX7 · 03-25-2014, 05:41 PM

If you're using the free driver you should probably use the driconf utility from SlackBUilds to see if it has any extra settings you can tweak on your card.

enorbet · 03-25-2014, 06:02 PM

Holy Framerate, Batman! 0_0

Quote:

Originally Posted by GazL

I have a Nvidia GT-140, which is essentially a re-released/re-badged 9600GT. Decent, but nothing fancy by today's standards.

__GL_SYNC_TO_VBLANK=0 /usr/libexec/xscreensaver/glknots -fps -delay 1
gives me about ~ 800fps. I guess it's less cpu intensive than glschool

__GL_SYNC_TO_VBLANK=0 /usr/libexec/xscreensaver/glplanet -fps -delay 1
gives me ~1380fps.

This may be the first time I've ever seen such framerates. I got 3700 and 4400 fps, respectively on those. So for humility's sake I immediately ran Unigine Heaven . Ahhh! the "real" virtual world.

moisespedro · 03-26-2014, 06:53 AM

Quote:

Originally Posted by whizje

I have a Radeon HD5750 and I reach between 24 and 28 fps.

Code:

bash-4.2$ glxinfo | grep -i render                                                                                          
direct rendering: Yes                                                                                                                           
OpenGL renderer string: Gallium 0.4 on AMD JUNIPER                                                                                              
    GL_MESA_window_pos, GL_NV_blend_square, GL_NV_conditional_render,

You can achieve more performance with the amd catalyst driver.

I really didn't want to have to ise the proprietary driver and, last time I tried, it didn't work anyways.

cascade9 · 03-26-2014, 08:40 AM

Quote:

Originally Posted by moisespedro

EDIT: Oh, nvm, thought it was a more intensive task. Getting 30FPS and 72000 polys on it. Is that any good?

Quote:

Originally Posted by whizje

I have a Radeon HD5750 and I reach between 24 and 28 fps.

So the 6670 (800MHz core) is getting 30FPS and the 5750 (700MHz core) is getting 24-28? Seems right to me.

Quote:

Originally Posted by enorbet

Holy Framerate, Batman! 0_0

This may be the first time I've ever seen such framerates. I got 3700 and 4400 fps, respectively on those.

Yotre' suprised that a mid end 2009 GPU is being beaten that much by a 2013 'gamers' GPU?

whizje · 03-26-2014, 12:32 PM

With the fglrx driver I get between 50 and 60 fps (HD 5750).

enorbet · 03-26-2014, 02:41 PM

Quote:

Originally Posted by cascade9

Yotre' suprised that a mid end 2009 GPU is being beaten that much by a 2013 'gamers' GPU?

It's not that. TBH I just never considered that any benchmark would go that high, especially since it is essentially useless information.

I'm somewhat painfully familiar with how GPU manufacturers play with features to hit price points for a given year. They even disable features to protect their high-end CAD quality market. Some can be hacked back on, but still....

I retired an Asus A8NE machine to secondary use after the 8800 GT burned out. I replaced the GPU with a current GT640. The 8800GT cost me ~$250 US and the GT640 cost $85 US. Just for curiosity sake I ran some of the old benchmarks just to see how the new one measured up. It didn't trash it but it beat it in almost every category and some tests that would not even run on the 8800, ran on the 640.