Difference between revisions of "Talk:Glxgears is not a Benchmark"

From cchtml.com
Jump to: navigation, search
(hmCYxOdYyhTxigc: new section)
m (Reverted edits by 134.2.15.69 (talk) to last revision by Mooninite)
 
(4 intermediate revisions by 3 users not shown)
Line 1: Line 1:
This piece was cogent, well-wrtiten, and pithy.
+
this leads to the obvious questions:
  
== hmCYxOdYyhTxigc ==
+
* what '''is''' a benchmark? please could someone provide links..
 +
* where can I get a list of benchmarks using something better than glxgears?
  
Your answer was just what I neeedd. It’s made my day!
+
==Rather depends on how you define benchmark..==
 +
Glxgears -is- a benchmark if you are comparing the same hardware before and after a change. People have been parroting the 'is not a benchmark' line irrespective of -what- is being measured for so long now that it is getting tedious.
 +
It is only 'not a benchmark' if you are comparing different hardware, or posting 'l33t' framerates on a forum.
 +
Glxgears is an excellent benchmark if you are doing a before and after test of your system when updating something like display drivers or Xorg, or if trying a config change that is supposed to improve 3d performance.
 +
Supposing you got 500fps before a change, and still get 500fps after. Then it is a sure sign that whatever you have done does not improve performance on your hardware (of course it may have additional benefits, stability etc..)
 +
If, however, you get 500fps before you upgrade the display drivers, and then get 700fps afterwards, then you have successfully used glxgears to benchmark your system and demonstrate a 40% speed improvement merely by upgrading some software. Of course that may not translate to a real-world improvement of 40%, but do you think it shows that nothing has changed?
 +
:How about saying 'Glxgears is not a benchmark for comparing dissimilar systems' 
 +
 
 +
 
 +
I found just what I was neeedd, and it was entertaining!
 +
 
 +
 
 +
The ubiqine link is dead, and seems to be a closed-source x86 only commercial demo. We need
 +
1) Open source!
 +
2) Multi-platform
 +
 
 +
I am for instance wanting to benchmark Power VR 3D accelerators found in high-end Omap (CortexV8) CPUs.
 +
:I've updated the links, but come on, it's easy to find them on the site. As for OMAP devices, this wiki is *not* for ARM devices. It is for '''x86 desktop computers'''. If you want an ARM benchmark, compile Quake 3, which is available for [http://maemo.org Maemo 5], if you want an ARM7 (Cortex A8) binary. --[[User:Mooninite|Mooninite]] 11:22, 9 November 2011 (CST)

Latest revision as of 12:52, 5 July 2013

this leads to the obvious questions:

  • what is a benchmark? please could someone provide links..
  • where can I get a list of benchmarks using something better than glxgears?

Rather depends on how you define benchmark..[edit]

Glxgears -is- a benchmark if you are comparing the same hardware before and after a change. People have been parroting the 'is not a benchmark' line irrespective of -what- is being measured for so long now that it is getting tedious. It is only 'not a benchmark' if you are comparing different hardware, or posting 'l33t' framerates on a forum. Glxgears is an excellent benchmark if you are doing a before and after test of your system when updating something like display drivers or Xorg, or if trying a config change that is supposed to improve 3d performance. Supposing you got 500fps before a change, and still get 500fps after. Then it is a sure sign that whatever you have done does not improve performance on your hardware (of course it may have additional benefits, stability etc..) If, however, you get 500fps before you upgrade the display drivers, and then get 700fps afterwards, then you have successfully used glxgears to benchmark your system and demonstrate a 40% speed improvement merely by upgrading some software. Of course that may not translate to a real-world improvement of 40%, but do you think it shows that nothing has changed?

How about saying 'Glxgears is not a benchmark for comparing dissimilar systems'


I found just what I was neeedd, and it was entertaining!


The ubiqine link is dead, and seems to be a closed-source x86 only commercial demo. We need 1) Open source! 2) Multi-platform

I am for instance wanting to benchmark Power VR 3D accelerators found in high-end Omap (CortexV8) CPUs.

I've updated the links, but come on, it's easy to find them on the site. As for OMAP devices, this wiki is *not* for ARM devices. It is for x86 desktop computers. If you want an ARM benchmark, compile Quake 3, which is available for Maemo 5, if you want an ARM7 (Cortex A8) binary. --Mooninite 11:22, 9 November 2011 (CST)