Arch Linux

toxygen

Member

Registered: 2008-08-22

Posts: 713

Re: Low FPS with glxgears !!!!!

I just wish that some of these experts would answer our questions and tell us what glxgears actually does rather than just say "it isn't a benchmark", eventhough we already acknowledged that fact.

I honestly dont understand why it's not considered a benchmark, considering the code has "#define benchmark" strewn about, and then you have this:

      /* calc framerate */
      {
         static int t0 = -1;
         static int frames = 0;
         int t = current_time();

         if (t0 < 0)
            t0 = t;

         frames++;

         if (t - t0 >= 5.0) {
            GLfloat seconds = t - t0;
            GLfloat fps = frames / seconds;
            printf("%d frames in %3.1f seconds = %6.3f FPS\n", frames, seconds,
                   fps);
            t0 = t;
            frames = 0;
         }

which is called once per drawing of screen (to get the value of frames), and then "fps" is calculated every 5 seconds.  so to me, that sounds like a benchmark of total fps with the given objects being drawn (the gears).  why/how it throws off such different scores is beyond me.  I know back in the day (97-98) I was recompiling mesa, glxgears et al just to get that extra 10fps (i was getting around 1500fps on a 386dx then)

so is it a benchmark? yes, kind of... is it reliable?  i dont really think so, but it's fun as hell trying to figure out why it gives a low fps one instance, and a higher one another time.  and as far as i remember, the driver has little to do with it (i believe back then i was using the framebuffer as a driver for X).

---

"I know what you're thinking, 'cause right now I'm thinking the same thing. Actually, I've been thinking it ever since I got here:
Why oh why didn't I take the BLUE pill?"

bender02

Member

From: UK

Registered: 2007-02-04

Posts: 1,328

Re: Low FPS with glxgears !!!!!

One thing you should be aware of about glxgears is that it matters how big its window is. So if you'd be using a tiling WM, you'd obviously get much lower results than you should

crouse

Arch Linux f@h Team Member

From: Iowa - USA

Registered: 2006-08-19

Posts: 907

Website

Re: Low FPS with glxgears !!!!!

Can you people please use REAL benchmarks instead of glxgears. Unless you're noticing slow performance elsewhere, there's nothing wrong with your systems.

aaaaawwwwwwwww c'mon, can't I use it as a benchmark...... plllleeeeeeeaaassseeeee  :D:D  Mine looks pretty good

[crouse@VistaCrusher ~]$ glxgears
69279 frames in 5.0 seconds = 13855.750 FPS
107393 frames in 5.0 seconds = 21478.479 FPS
108098 frames in 5.0 seconds = 21619.592 FPS
108512 frames in 5.0 seconds = 21702.387 FPS
108201 frames in 5.0 seconds = 21640.148 FPS
108097 frames in 5.0 seconds = 21619.227 FPS
107418 frames in 5.0 seconds = 21483.443 FPS
107560 frames in 5.0 seconds = 21511.992 FPS
107220 frames in 5.0 seconds = 21443.857 FPS
107928 frames in 5.0 seconds = 21585.424 FPS
105831 frames in 5.0 seconds = 21166.100 FPS

[crouse@VistaCrusher ~]$ lspci | grep -m 1 "VGA" | cut -d: -f3 |  sed 's/^[ \t]*//'
nVidia Corporation NV45GL [Quadro FX 3400/4400] (rev a2)
[crouse@VistaCrusher ~]$

http://wiki.cchtml.com/index.php/Glxgea … _Benchmark

ashyanbhog

Member

From: India

Registered: 2008-08-19

Posts: 49

Re: Low FPS with glxgears !!!!!

Did any of you guys check for parameters appended to xorg.conf by Ubuntu's Nvidia driver.

pointone

Wiki Admin

From: Waterloo, ON

Registered: 2008-02-21

Posts: 379

Re: Low FPS with glxgears !!!!!

First result on a Google search of "glxgears": http://wiki.cchtml.com/index.php/Glxgea … _Benchmark

This is what it actually does: http://cvsweb.xfree86.org/cvsweb/xc/pro … .c?rev=1.3

---

M*cr*s*ft: Who needs quality when you have marketing?

rsambuca

Member

From: Calgary, Canada

Registered: 2008-07-21

Posts: 143

Re: Low FPS with glxgears !!!!!

Thanks pointone, but that still does nothing to help explain why there is such a drastic difference with this NONbenchmark test between Ubuntu and Arch.

saz

Member

From: Lisbon

Registered: 2008-04-19

Posts: 115

Re: Low FPS with glxgears !!!!!

Thanks pointone, but that still does nothing to help explain why there is such a drastic difference with this NONbenchmark test between Ubuntu and Arch.

yes, let's try not to forget the initial intentio of the post:

Why is there such a huge difference between the results in Ubuntu and Archlinux ?

toxygen

Member

Registered: 2008-08-22

Posts: 713

Re: Low FPS with glxgears !!!!!

Why is there such a huge difference between the results in Ubuntu and Archlinux ?

I don't really think it's so much between one distro vs another, since different kernels on the same machine (same distro) and even different graphic drivers will change it. 

I had a situation where I build the nvidia driver, installed, ran glxgears a few times and averaged the result.  Then, starting on a new folder, rebuilt the driver, made sure to do a clean removing of the previous one, checked every folder in the system for any remaining trace of nvidia/GL files, then installed the new built driver package, and got completely different results (differences between 200-700 fps) on glxgears.

so my guess is that somewhere in glxgears code is a call to /dev/random and performing some black magic with whatever value is there, and setting frame rate accordingly.

Last edited by toxygen (2008-10-09 17:28:46)

---

"I know what you're thinking, 'cause right now I'm thinking the same thing. Actually, I've been thinking it ever since I got here:
Why oh why didn't I take the BLUE pill?"

rsambuca

Member

From: Calgary, Canada

Registered: 2008-07-21

Posts: 143

Re: Low FPS with glxgears !!!!!

Actually, three different versions of Ubuntu on my rig give at least 2fold, if not higher rates than on my Arch setup.

saz

Member

From: Lisbon

Registered: 2008-04-19

Posts: 115

Re: Low FPS with glxgears !!!!!

Why is there such a huge difference between the results in Ubuntu and Archlinux ?

...and got completely different results (differences between 200-700 fps) on glxgears.

You must be joking... of course it isn't that much of a difference! you're talking about 200-700fps difference, I'm talking about a 20000 fps gap! From 27600 to 7600...

Sorry but I don't buy it.... exactly same rig, same drivers version and same glxgears... the only thing that changes is the option I choose in grub...

such a huge difference of 20000fps can't be simply by a "radomity" of glxgears... I would believe if it was a 3-4k difference, but 20k  is just too much... is the difference between a super rig and a 486...

Last edited by saz (2008-10-10 12:49:28)

brebs

Member

Registered: 2007-04-03

Posts: 3,742

Re: Low FPS with glxgears !!!!!

CONFIG_X86_PAT in the kernel can sometimes be blamed. I certainly need to disable PAT with nvidia for performance, using NVreg_UsePageAttributeTable=0

My /etc/modprobe.conf contains:

options nvidia NVreg_DeviceFileMode=0660 NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=100 NVreg_ModifyDeviceFiles=1 NVreg_EnableMSI=1 NVreg_UsePageAttributeTable=0 NVreg_UseVBios=0 NVreg_RMEdgeIntrCheck=1

Can't find the thread when I want to, but it's on the nvidia forum.

toxygen

Member

Registered: 2008-08-22

Posts: 713

Re: Low FPS with glxgears !!!!!

CONFIG_X86_PAT in the kernel can sometimes be blamed. I certainly need to disable PAT with nvidia for performance, using NVreg_UsePageAttributeTable=0

My /etc/modprobe.conf contains:

options nvidia NVreg_DeviceFileMode=0660 NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=100 NVreg_ModifyDeviceFiles=1 NVreg_EnableMSI=1 NVreg_UsePageAttributeTable=0 NVreg_UseVBios=0 NVreg_RMEdgeIntrCheck=1

Can't find the thread when I want to, but it's on the nvidia forum.

I will give this a try, my modprobe conf already has SBA and FW enabled:

# Nvidia fastwrites and SBA

alias char-major-195* nvidia
options nvidia NVreg_EnableAGPSBA=1 NVreg_EnableAGPFW=1

What is your video card/cpu/ram/mo-bo setup?

---

"I know what you're thinking, 'cause right now I'm thinking the same thing. Actually, I've been thinking it ever since I got here:
Why oh why didn't I take the BLUE pill?"

brebs

Member

Registered: 2007-04-03

Posts: 3,742

Re: Low FPS with glxgears !!!!!

Mine is Nvidia 8800GTS 512mb, quad Core2 Q6600 CPU, 6gb, Asus P5K Premium mobo.

Also see this thread regarding MTRR. If MTRR is used instead of PAT, then write-combining is important for performance. Which recent kernels are having difficulty setting up. *And* also messing up PAT - or the Nvidia driver is messing up PAT, causing jerkiness.

Of course, just to confuse me, my 512mb nvidia card has a 256mb RAM memory hole!

saz

Member

From: Lisbon

Registered: 2008-04-19

Posts: 115

Re: Low FPS with glxgears !!!!!

Can you people please use REAL benchmarks instead of glxgears. Unless you're noticing slow performance elsewhere, there's nothing wrong with your systems.

aaaaawwwwwwwww c'mon, can't I use it as a benchmark...... plllleeeeeeeaaassseeeee  :D:D  Mine looks pretty good

[crouse@VistaCrusher ~]$ glxgears
69279 frames in 5.0 seconds = 13855.750 FPS
107393 frames in 5.0 seconds = 21478.479 FPS
108098 frames in 5.0 seconds = 21619.592 FPS
108512 frames in 5.0 seconds = 21702.387 FPS
108201 frames in 5.0 seconds = 21640.148 FPS
108097 frames in 5.0 seconds = 21619.227 FPS
107418 frames in 5.0 seconds = 21483.443 FPS
107560 frames in 5.0 seconds = 21511.992 FPS
107220 frames in 5.0 seconds = 21443.857 FPS
107928 frames in 5.0 seconds = 21585.424 FPS
105831 frames in 5.0 seconds = 21166.100 FPS

[crouse@VistaCrusher ~]$ lspci | grep -m 1 "VGA" | cut -d: -f3 |  sed 's/^[ \t]*//'
nVidia Corporation NV45GL [Quadro FX 3400/4400] (rev a2)
[crouse@VistaCrusher ~]$

http://wiki.cchtml.com/index.php/Glxgea … _Benchmark

how did you get those values? could you share with us your xorg.conf ?

EasyTarget

Member

From: Amsterdam

Registered: 2009-02-16

Posts: 17

Website

Re: Low FPS with glxgears !!!!!

Jeez I'm fed up of people who post 'glxgears is not a benchmark'.. It's only 'not a benchmark' if you're posting the results to the internet and waving your willy around shouting 'my FPS is bigger than yours'.

I used glxgears to benchmark my system last week, here is how:
Reboot, login, do not start any apps apart from gterm. Run glxgears and note the fps figure once it has stabilised after a couple of minutes.
Upgrade my system.
Reboot, login, do not start any apps apart from gterm. Run glxgears and note the fps figure once it has stabilised after a couple of minutes.
FPS has gone up by ~20%.
In what way did I fail to do a before and after benchmark of a change I made to my system?

Of course, I could have wasted hours researching what benchmark program I -should- use according to all the 'is not a benchmark' geniuses, installing it, and getting it to run.  Am I likely to have got a radically different result? Would all that extra effort have told me anything useful? Of course not, my chief interest is in knowing whether my system is improving. I don't really care how it stacks up against other peoples systems; if I was in a FPS arms race I'd just go out and buy an even bigger GPU anyway.

---

My anger management class pissed me off. - anon

quarkup

Member

From: Portugal

Registered: 2008-09-07

Posts: 497

Website

Re: Low FPS with glxgears !!!!!

well I actually got 16000 fps on a nvidia6800go...

on ubuntu I got 200 fps (using the 'nv' driver of course) ...

---

If people do not believe that mathematics is simple, it is only because they do not realize how complicated life is.
Simplicity is the ultimate sophistication.