2xmark '06, The Alternative Gamer's Benchmark


Epicenter

Well-Known Member
Joined
Oct 9, 2005
Messages
2,068
Age
40
Location
USA
Website
www.epicgaming.us or http
I'd like to announce the first release of a benchmark utility I've been working on for the GP2X, dubbed '2Xmark 2006' (har har). A little project to tide you over while I work on bigger and better things (I'm keeping most of it to myself for now, but I will let on one project is an ultra-high intensity arcade-style top-scrolling shooter in the 'Danmaku / Curtain Fire' shooter genre with VERY advanced graphics, music and rapid gameplay, and the other is a first-person shooter based on the Quake engine.) :) ) Anyway!

2Xmark serves to gauge the performance of your '2x system by measuring its arithmetic speeds and will show you an accurate reading of speed at Fixed-point arithmetic (Addition, Subtraction, Multiplication, Division-- the basics), and the same using Floating Point math.

There are more multimedia tests to come in the next release, but I have also included an OGG Vorbis decoding performance benchmark. Expect the next build (or one after that) to have SD card read/write speed tests, RAM speed tests, more complex math tests and some software sprite manipulation, scaling and rotation performance tests.

This is, of course, the first public release so there may be bugs, please let me know about them. Expect Division to report as much slower than the other operations since it's not natively supported on the ARM9 processor series. (and yes, I know addition and subtraction are essentially the same thing from an actual arithmetic standpoint, I'm just being thorough. There aren't THAT many tests thus far.)

Probably the most fun things to do with this program aside from find out your performance level at various overclocked speeds is to show it off. You'll note I've added my highest achieved record to my signature below. :) You can also use it as a stability testing tool (the next release should include a Torture-Test feature to see if your shiny new clockrate is stable.)

You can get the first release here:
http://archive.gp2x.de/cgi-bin/cfiles.cgi?0,0,0,0,8,1435

.. or from my personal server if the file archive is experiencing difficulty:
http://www.epicgaming.us/gp2x/2xmark_20060407.zip

Enjoy and please let me know what you think!

EDIT: Fixed a bug that kept the program from working on some people's systems. Sorry 'bout that.
 
How are you planning to weight the various benchmarks when it comes to calculating the 2XMark score? :p Just curious, very nice idea.
 
hackgrid posted on Apr 9 2006 at 08:52 AM said:
21118 2xmarks @ 319MHz

;)

Hahaha, nice. I got to 320 MHz but couldn't get through a full battery of tests successfully.

The tests are weighed mostly evenly right now, but I will be changing that when I add more multimedia tests. The first version you might call 'proof of concept'.
 
Last edited by a moderator:
That's never happened to me, but it doesn't sound like anything the program did. It's not doing anything all that complicated in that stage, it's just multiplying some numbers together and shoving them in a variable.
 
cowai posted on Apr 14 2006 at 09:02 AM said:
i get 21144@317mhz.. why does epicenter get 1k more than me at the same mhz? I have open2x..

There will be some variation from unit to unit. This is mostly due to the fact that the arithmetic operations 2Xmark uses for its testing vary from iteration to iteration, otherwise, the ARM920T would cache them and they would run unnaturally and inaccurately fast compared to a real game situation, thereby being a very poor performance simulation of how a real game would benefit from your overclocking efforts.
 
Last edited by a moderator:
Just thought I'd mention your file on the archive just points to the file on your server lol! Ill fix that for you :p Otherwise if something happens to your server then the archives one stops working.
 
it would be a good addition if you could overclock your cpu directly in 2xmark, instead of relying of an external application.

On topic: 14850 @ 199Mhz
 
I probably will-- I also intend to add a bigger battery of tests, a stability-testing burn-in test mode, and a battery-life benchmark for real, standardized comparison of battery types. Also probably some pretty graphics courtesy of HydraLancer-- not just ASCII text. ;)
 
Epicenter posted on May 21 2006 at 10:58 PM said:
and a battery-life benchmark for real, standardized comparison of battery types.
That's what i want the most to see real change between firmeware and between different batterie seller
 
Last edited by a moderator:
Sorry to revive this older thread, but how about adding RAM read write tests? It surely would be interesting to see how fast RAM gets by playing around with CPU/RAM Tuner.
 
Back
Top