Difference between revisions of "Talk:3Dfps"

From ThinkWiki
Jump to: navigation, search
(Broken redirect)
 
(8 intermediate revisions by 5 users not shown)
Line 1: Line 1:
I [[User:Nomeata|Nomeata]] suggest
+
''(for discussion of formatting, see history -- [[User:Nomeata|Nomeata]])''
  
*Thinkpad Model (i.e. Thinkpad T40)
+
Shoudn't the kernel be after the driver, since I'd rather compare different kernels and same driver than different driver and same kernel? --[[User:Nomeata|Nomeata]] 13:25, 29 Sep 2004 (CEST)
**test application (i.e. glxgears / 3Dmark2000 / ... )
 
***Driver (i.e. fglrx / xfree / xorg / ibm / dna / omega / ... )
 
  
Since you most likely want to compare results of same hardware and benchmark than yours.
 
 
--[[User:Nomeata|Nomeata]]
 
----
 
I think so, too. Obviously the goal of fps comparison is to find out the best configuration for a specific model.
 
 
--[[User:Akw|akw]] 00:22, 25 Sep 2004 (CEST)
 
 
----
 
----
 +
I think one more often compares different drivers at same kernels. Also i think that driver variation will be more than kernel variation.
 +
However, i'd suggest the following order: Model, Benchmark, Driver, fps, XServer, Kernel, Notes, User. This would represent the following search pattern: "Ok, i have a T41p, i'm using the fglrx driver and glxgears seems a bit slow to me, lets have a look... ahh...100 fps more...why...ahh...XServer is more recent...kernel is less recent (does that matter?)...ohh...and i should probably check my mtrr."
  
done it
+
[[User:Wyrfel|Wyrfel]] 13:57, 29 Sep 2004 (CEST) 13:56, 29 Sep 2004 (CEST)
--[[User:Nomeata|Nomeata]] 00:31, 25 Sep 2004 (CEST)
 
 
 
 
----
 
----
I had it like this in the beginning.
+
I coded a new MediaWiki Specialpage, to get the table dynamic (to be database-driven, enabling sorting and so on).
 +
It is not yet finished. Hope I will get this finished soon. The MediaWiki code is strange and rather poorly documented.  
 +
Suggestions are appreciated! :-)
  
But i thought about splitting it into single pages and then this order would mean a lot of single pages. So i turned it around, because:
+
Layout:
 +
* First part as a template (to be editable by the public): [[Template:Benchmarks_Header]]
 +
* Dynamic database-driven table with sorting
 +
* Lats part as a template again (for notes): [[Template:Benchmarks_Footer]]
  
Testapplication -> Driver ---link---> TestAppDriverPage -> ThinkpadModel is a hirarchy with a low number of single pages (All models for one TestApp-Driver-combination on one page).
+
Further discussion on the new dynamic benchmark page on [[Template:Benchmarks_Header]], please!
 
 
But you are right, it makes more sense from the comparison point of view to put the drivers in the end.
 
 
 
Shell we split into separate pages anyway? And if so, on the Model level or on the testapp level?
 
 
 
[[User:Wyrfel|Wyrfel]]
 
  
 +
--[[User:Akw|akw]] 21:57, 8 Oct 2004 (CEST)
 
----
 
----
Not before it is not too full
 
  
Or just use a tabluar format, sorted this way, for example:
+
== how to run instructions needed ==
  
{| border="1"
+
I think we need instructions on how to run the benchmark.
|+ Benchmark results
 
|-
 
! Model !! Benchmark !! Driver !! fps !! User and Notes
 
|-
 
| T41p [[2373-GHG]] || fgl_glxgears || radeon || 259fps || [[User:Nomeata|Nomeata]] ''vesa framebuffer active, therefore no mtrr for radeon''
 
|-
 
| T41p [[2373-GHG]] || glxgears || radeon || 1209fps || [[User:Nomeata|Nomeata]] ''vesa framebuffer active, therefore no mtrr for radeon''
 
|}
 
  
--[[User:Nomeata|Nomeata]] 00:44, 25 Sep 2004 (CEST)
+
For instance it is not clear if the benchmark should just be started from the cmdline and have it run with default window size, or if it should be resized to fullscreen.
----
 
  
I prefer the tabular way, it makes a lot of sense here.
+
For instance I am running Fedora Core 3T2 with x.org 6.8.1 and am getting 1067 FPS with the default window size, but only 92 FPS at fullscreen.
 +
This is on a T30 (2366-52U) which has a 1.8GHz P4M, running the opensource radeon driver.
  
[[User:Wyrfel|Wyrfel]] 00:55, 25 Sep 2004 (CEST)
+
== Broken redirect ==
 
 
----
 
Done
 
  
--[[User:Nomeata|Nomeata]] 00:57, 25 Sep 2004 (CEST)
+
The redirect is currently broken.
----
 
Maybe it would be also interesting which version of X and fglrx was running while benchmarking. I noticed some discussion about the new fglrx ver. 12 being somewhat slower than older versions and so on.
 
  
--[[User:Akw|akw]] 01:10, 25 Sep 2004 (CEST)
+
Which is just as well -- glxgears is '''not a valid benchmark'''. It was not intended to be one, and the workload it generates on modern cards is totally different from any realistic X app (heck, I get better glxgears FPS when I ''disable'' hardware acceleration). Please stop using it.
  
 +
--[[User:Thinker|Thinker]] 15:57, 28 November 2006 (CET)
 
----
 
----
Sure, but then we need a way to shorten the notes. For example, Just put "''no mtrr''" in there and explain it below the table
 
--[[User:Nomeata|Nomeata]] 01:19, 25 Sep 2004 (CEST)
 
----
 
Good idea!
 
 
--[[User:Akw|akw]] 01:27, 25 Sep 2004 (CEST)
 

Latest revision as of 15:57, 28 November 2006

(for discussion of formatting, see history -- Nomeata)

Shoudn't the kernel be after the driver, since I'd rather compare different kernels and same driver than different driver and same kernel? --Nomeata 13:25, 29 Sep 2004 (CEST)


I think one more often compares different drivers at same kernels. Also i think that driver variation will be more than kernel variation. However, i'd suggest the following order: Model, Benchmark, Driver, fps, XServer, Kernel, Notes, User. This would represent the following search pattern: "Ok, i have a T41p, i'm using the fglrx driver and glxgears seems a bit slow to me, lets have a look... ahh...100 fps more...why...ahh...XServer is more recent...kernel is less recent (does that matter?)...ohh...and i should probably check my mtrr."

Wyrfel 13:57, 29 Sep 2004 (CEST) 13:56, 29 Sep 2004 (CEST)


I coded a new MediaWiki Specialpage, to get the table dynamic (to be database-driven, enabling sorting and so on). It is not yet finished. Hope I will get this finished soon. The MediaWiki code is strange and rather poorly documented. Suggestions are appreciated! :-)

Layout:

Further discussion on the new dynamic benchmark page on Template:Benchmarks_Header, please!

--akw 21:57, 8 Oct 2004 (CEST)


how to run instructions needed

I think we need instructions on how to run the benchmark.

For instance it is not clear if the benchmark should just be started from the cmdline and have it run with default window size, or if it should be resized to fullscreen.

For instance I am running Fedora Core 3T2 with x.org 6.8.1 and am getting 1067 FPS with the default window size, but only 92 FPS at fullscreen. This is on a T30 (2366-52U) which has a 1.8GHz P4M, running the opensource radeon driver.

Broken redirect

The redirect is currently broken.

Which is just as well -- glxgears is not a valid benchmark. It was not intended to be one, and the workload it generates on modern cards is totally different from any realistic X app (heck, I get better glxgears FPS when I disable hardware acceleration). Please stop using it.

--Thinker 15:57, 28 November 2006 (CET)