Release Pandora Crude Bench - Benchmark your Pandora


ptitSeb

Serial Porter
Joined
Aug 15, 2012
Messages
9,306
Age
51
Location
France, near Lyon
Here is a little CPU/GPU Benchmark for the Pandora.
 
 
This software is a Bench software split in 3 parts to mesure performance.
 
The 3 Parts are:

  • The Mandelbrot fall
A fall (zooming) in a crude Mandelbrot fractal. This part is CPU Heavy. It draws a mandelbrot in a 200x120 that is scalled fullscreen using Notaz's optimized SDL.
Score for this part is FPS*40 (expect fps starting at 8).

  • Hypnotic Pandora
A simple hypnotic shader, using GLES2. No calculation here, this tests is GPU only.
Score for this part is FPS*10 (expect fps starting at 35).

  • Irrlicht Demo
The demo that come with Irrlicht 1.8. It's a Flyby of a small Quake3 map. This test use both CPU and GPU.
Score for this part is FPS*10 (expect fps starting at 30).


 
Final PCB Score is the sum of Score from three parts.
 
Here are some result:
 
C.C @ 800MHz (standard SysSpeed), from minimenu : 1280 (11/42/38)
Rebirth @ 1GHz (standard SysSpeed), from Xfce: 1420 (14/41/41)
GigaPandora @ 1GHz (standard SysSpeed), from Xfce: 1951 (14/74/61)
 
The bench is quite reproductible, with only a few percents difference between consecutive runs.
 

History log
=========
 

Build 02

----------

  • Modified Part2 so it works even with beta SGX driver. Result of Part2 may be affected (faster?).

Build 01
-----------

  • Initial build
  • Results are saved in apddata/pcb1/home in text files named with date and time of test.
Ready, Set, Bench, and put your score here :)
 
Last edited by a moderator:
1GHz Pandora: 1965 ( 15 / 74 / 62 ); cpuspeed = 1000 MHz; sysspeed = 400 MHz

The result is very stable, ran it four or five times. Got a 14 for the Mandelbrot once because I moved the mouse with the Nub (just to see if it does anything).

(what exactly is sysspeed?)
 
1GHz Pandora: 1965 ( 15 / 74 / 62 ); cpuspeed = 1000 MHz; sysspeed = 400 MHz

The result is very stable, ran it four or five times. Got a 14 for the Mandelbrot once because I moved the mouse with the Nub (just to see if it does anything).

(what exactly is sysspeed?)
SysSpeed is the System Speed (ok, that's the obvious part) that also control Memory Speed (MEMSpeed = SysSpeed/2) and GPU Speed (for GHz it's GPUSpeed = SysSpeed/2, for CC&ReBirth GPUSpeed = SysSpeed/3).

You got 15fps on Mandlebrot with stock speed on GHz. Strange that I got 14. I launched it from Xfce, Wifi activated (but not Dropbox or Samba yet). You launched it from Minimenu or Xfce?
 
Last edited by a moderator:
Ah, I see. So my GPUspeed is 200 MHz?

I launched from Xfce. Just tried with Wifi activated, but nothing changes there.

Sometimes I get some textrue problems with the Irrlicht-demo, is that normal?
 
Last edited by a moderator:
Ah, I see. So my GPUspeed is 200 MHz?

I launched from Xfce. Just tried with Wifi activated, but nothing changes there.

Sometimes I get some textrue problems with the Irrlicht-demo, is that normal?
I have seen that too. Not reproducible evrytime. But we know the current driver on newer Pandora has some glitch sometimes. Sometimes it runs fine.

And yes, your GPU runs at 200 MHz, where the CC & Rebirth one runs at 110 MHz.
 
rebirth  under xfce  1437 (15/41/42) cpuspeed=1000 sysspeed=332

just bumped up the ram and got:

1501 (15/44/45)cpu=1000, sysspeed=360
 
Last edited by a moderator:
score of 1468 and 13/42/40 on 900/332MHz CC unit

Edit: score is 1370 when launched from MiniMenu with the same values.
 
Last edited by a moderator:
score of 1468 and 13/42/40 on 900/332MHz CC unit

Edit: score is 1370 when launched from MiniMenu with the same values.
Hum, the decrease in performance when launching from Minimenu is strange... But that can be explained if you "just" quit xfce to go to Minimenu without rebooting, as I think Xfce stay loaded that way.
 
Yes switch GUI is all I did. It still reported 13/42/40 though.

 Frist time I've used MM for more than a year now, I never knew I had to start it at boot to see a difference in performance :D
 
Giga Pandora: CPU 1GHz; Sysspeed 400:

1948 (14 / 73 / 61)

Giga Pandora: CPU 1,2 GHZ, Sysspeed 460:

2299 (18 / 85 / 72)
 
Giga Pandora: CPU 1GHz; Sysspeed 400:

1948 (14 / 73 / 61)

Giga Pandora: CPU 1,2 GHZ, Sysspeed 460:

2299 (18 / 85 / 72)
Nice overclock ! I tried 460 on my Ghz, but it's not working. To bad, the result looks impressive ! I'll try some more modest O/C for sysspeed.
 
Giga Pandora:

1954 (14, 74, 61), cpu=1000, sys=400

2113 (18, 74, 65), cpu=1200, sys=400

3D demo looks a bit strange though, the texture on the character in the end looks smeared on, and the specularity (I guess?) causes strange artifact-like effects depending on mipmap distance I think.

... ... ... will it get C4A high-score compatibility? =D
 
Last edited by a moderator:
Giga Pandora:

1954 (14, 74, 61), cpu=1000, sys=400

2113 (18, 74, 65), cpu=1200, sys=400

3D demo looks a bit strange though, the texture on the character in the end looks smeared on, and the specularity (I guess?) causes strange artifact-like effects depending on mipmap distance I think.

... ... ... will it get C4A high-score compatibility? =D
The demo on Part 3 is the full Demo of Irrlicht engine. Using only GLES1, there are many approximation. Plus GLES1 doesn't run too well on GHz Pandora...

For Information, I have, at last, a running version of Irrlicht using GLES2 ! The (obvious, when you think about it) point is that you cannot have both GLES1 & GLES2 Renderer compile in the same Irrlicht, because it use "static" dynamic library, and both GLES_CM and GLESv2 have comonf function... If activate both, you have garbled screen in GLES2. I have messed with the shaders and now have a fairly good rendering. There are still artefect that I don't understand, and the speed is slower than with GLES1, but there are much more effect activated. I will try some SuperTuxKart with that...

C4A Hiscore, yeah, could be funny ;)
 
My first pcb score was a 2117 (18 / 73 / 65), with a cpuspeed=1200 MHz and syspeed=400 MHz

I have a 1GHz unit purchased from Link a month ago.  Before I ran the test, I confirmed that I had a setting of 1.2 GHz and an OPP4 setting (using _wb_'s great system info program).

My second run (a couple of minutes later) of the crude bench netted me a score of 2122 (18 / 74 / 65).

Third run got me (1 minute later) a 2124 score (18 / 74 / 66).  I did not change any settings in between runs.  Seems stable and the scores seem to be getting (progressively) better.  Wonder why?...and I wonder if my Pandora is improving with age?   :)
 
Wait a mounth and you'll have a 2GHz Pandora ;)

More seriously, Linux is a multitask OS, so a few things are always runing in the background, creating small differences between runs.
 
Heaayyy...A 2GHz Pandora, I like the sound of that!

...ooops, my bubble just bursted - my 4th and 5th runs both were a 2123 score - looks like I've hit the threshold (without overclocking the CPU more or the GPU and memory)...Oh well, it was a fun ride while it lasted.   :D
 
Your PCB Score is 1966 (15 / 74 / 62 ), with a cpuspeed=1000 MHz and sysspeed=400 MHz

Your PCB Score is 2123 (18 / 74 / 66 ), with a cpuspeed=1200 MHz and sysspeed=400 MHz

Your PCB Score is 2154 (18 / 74 / 66 ), with a cpuspeed=1250 MHz and sysspeed=400 MHz

Your PCB Score is 2286 (18 / 81 / 71 ), with a cpuspeed=1250 MHz and sysspeed=440 MHz

Your PCB Score is 2242 (18 / 81 / 67 ), with a cpuspeed=1250 MHz and sysspeed=460 MHz

Using fresh boots and minimenu.

Seems like my pandora doesn't like the too high sysspeed!

I noticed glitches in the irrlicht demo at sysspeed=460 MHz, and some specularity ones like said before appears randomly from test to test.

Seems like the cpuspeed affects test 1 and 3, and sysspeed test 2 and 3 mainly, as expected.

Nice app!
 
Last edited by a moderator:
Back
Top