Mac Processor Speed Test

Q

quattro

0
Messages
3
Mac Processor Speed Test

Hi folks, I added a Cheetah file for you, to test the Performance of the actual and older Macs, using Cheetah 3D 2.6. It would be great, if you could post your results with following info: Mac / Processor / System-Version /

If we will get enough feedback, I would like to post a statistic / grafic with the results.

Ram makes no difference for this JAS-file.

I tried it on a G5 Dual 2GHz and a Powerbook 1.25 GHz.
Cheetah Version 2.5 to 2.6 rendered the file with no difference in time.

Here are my results:

G5 Dual 2GHz, OSX 10.4.3 : 171 sec
Powerbook 1.25 GHz, OSX 10.3.9 : 480 sec

Just oben the file an click on render, without changing anything. Even if you change the angle or focus a little bit, it could make a difference of 20% and more in time!
 

Attachments

  • rendertestfile.jas_189.zip
    7.5 KB · Views: 1,021
Tiger 10.4.3 / Cheetah3D 2.6
Dual G5 1.8GHz
512 MB RAM ----> 196 sec.

Don't know if my system was optimal during the test (I have some silent app in the background)

Cyril.
 
This should not be really important for a testing purpose, but oversampling in your scene was not set to a typical configuration, as you set 16x during the first pass and 0x for the 2nd one.

With 1x and 16x configuration I get 76 s with the same output.

Cyril.
 
Stock Dual MDD G4 867, 1 GB Ram
OSX 10.4.2 C3D 2.6
386.44 sec

Thanks for doing this quattro. Individuals that can’t afford new cutting edge machines will now be able to make knowledgeable decision via cost versus benefits when they buy used.

Curious why jr57k and I have such different results. Wonder if it’s the OS version?
 
irev said:
Stock Dual MDD G4 867, 1 GB Ram
Curious why jr57k and I have such different results. Wonder if it’s the OS version?
Click to expand...

Without knowing exactly what your configurations were, I'd have to guess that it's because his system may have been multitasking more than yours during the test. As a rule, the system should be restarted immediately before benchmarking, and nothing else should be running at the same time.
 
Hi,
normally render times are very re-produceable. But it is very important to use the same version of Cheetah3D. v2.6 renders some scenes considerably faster. Some others a little bit slower. So always check the About box if you are really using the appropriate version of Cheetah3D.

By,
Martin
 
My big problem with this benchmark is that it doesn't make full use of Cheetah's features, so doesn't really test the renderer's full range of abilities. As such, any benchmark you run with it is pretty well useless for working out how fast the whole renderer can run - if, for instance, HDRI-handling was massively accelerated between releases, this benchmark simply wouldn't show any difference.

Things that it doesn't appear to test, but really should:

-Shadows
-HDRI
-Transparent materials (both glass-like, and with a semi-transparent texture)
-Refractive materials
-Area lights
-Highly reflective scenes
-High polycount scenes
-Modifiers and Creators (such as 2 or 3 iteration subdivides, or arrays of identical objects)
-Specular scatterings (in Radiosity)
-Fog

In short, I think this could be done much better. Not all of these features could be implemented, of course - fog and specular scatterings, for instance, would probably be pointless when placed in the same scene - but our current benchmark scene is far too simple to accurately benchmark Cheetah3D with.

I'd be happy to put together a much more complex benchmark scene in a few days, when I have the time, so I can demonstrate my idea of a better test. It wouldn't take long to put together. =)
 
Hi,
I just got my Intel iMac. As far as I could say after a 1 hour of testing everything works well within Rosetta.


Specs 2.0 GHz Intel Core Duo.
512 MByte Ram.
OS 10.4.4
Time: 303 sec.


I was actually surprised by it's speed. It looks like that Rosetta is working very well. Great job Apple.

Now it's time to make a Universal Binary. But first I have to enjoy all the new iLife apps. :wink:


By,
Martin
 
Martin, since Rosetta translates then optimises and caches blocks of code as they're run, it's capable of running extremely quickly on code that does the same thing repeatedly for a long time. Which might explain why Cheetah3D renders so quickly on the Intel Mac. :)
 
Hi Randy,
you are right but natively compiled code will always be better. And as my first tests show even much better. The native Intel version is 3 times faster. That is appriximately the ratio Apple tells too.


Specs 2.0 GHz Intel Core Duo.
512 MByte Ram.
OS 10.4.4
Cheetah3D 3.0 a4 (Intel)
Time: 91 sec.


I almost can't believe that numbers. 91 sec. !!!! That's even faster than the Quad G5. These new iMacs really scream. :D :D :D :D :D I'm sure Apple will sell lots of them. :D

But please that these preliminary numbers with care.

By,
Martin
 
That's more than twice the rendering time of my PM G5 Dual 1.8 !!! :shock: :shock: :shock:

Cyril.
 
You must log in or register to reply here.
Back
Top