OpenGL benchmark

Is there a true benchmark on OPENGL apps under PW ?
Could be good to have that for all CAD application users.
no ?

 

looking at the number of views of my posts about OPENGL speed, I think there is some interest on it...
Would somebody, from parallels, or not, try to make a serious benchmark on openGL apps, and CAD applications ?
I would if I had a macbook pro...I promise !
Giving real numbers, or make a video on Unigraphics or Catia running in PW may decide many of us to buy a MBP and PW..!

 

Personally, i believe, that any product should be benchmarked by independent experts. I really hope that someone will be interested in such a research and will make it.

 

I agree on that, even so I feel that parallels may provide some information about the performance we can expect on openGL apps.

But I guess you are overbooked, and can understand that ! and anyway, I did not try PW yet, because of a lack of a MBP but it seems it is a great product, big bravo !

Maybe if the readers of theses posts can add some posts to request a benchmark, then people that have everything needed (a MBP, PW, NX4 and Catia V5 or any OpenGL app...) would make it !

 

I have run Google Earth on OpenGL mode and the speed is poor. I imagine it is possible in some way for Parallels to provide OpenGL speed improvements by hooking into CoreImage or Quartz Extreme in some way .. I'm not an expert, so take that with a grain of salt. If you're intending to get work done in an OpenGL intensive app (one that, for instance, needs good OpenGL hardware on the PC to begin with) I wouldn't count on it being useable in PW. If the app works OK on a PC that doesn't have any OpenGL or DirectX acceleration, then it may be OK under PW.

Not a benchmark, but maybe the info you needed?

Cheers,
Brice

 

Good idea to check how the app run when disabling OpenGL acceleration on the PC...
I try tomorow morning...

Anyone else has some input to give ? or has t request a true benchmark ?

 

Cinebench 95 Benchmark Results

OK, I'm not an "expert", but here are some Cinebench 95 (Maxon) results:


Computer MacBook Pro dual 2.0GHz with 2GB RAM, setup for BootCamp dual boot and also PW beta 5 on OS X.

Software Mac OS X 10.4.6 with latest patches / updates. Windows XP Professional with SP2 and latest patches / updates.

Description The benchmark runs the same tests 3 times: once using a pure software shader, then using OpenGL with software lighting (SW-L), then with hardware lighting (HW-L). In all cases, higher numbers mean better results.


(A) Direct boot Windows (Boot Camp) Defragged disk, no other apps running. OpenGL 3D acceleration set to "Balanced" in Control Panel.

C4D Shading: 274
OpenGL SW-L: 1020
OpenGL HW-L: 2088
OpenGL Speedup: 7.62x

(B) Parallels WS beta 5 with clean XP install Parallels VM Compactor run and disk defrag also. No other apps running (Windows or Mac OS X). PW was running full-screen, native (1440 x 900) resolution during benchmark. Intel VT-x enabled. VM assigned 1GB RAM.

C4D Shading: 676
OpenGL SW-L: 402
OpenGL HW-L: 367
OpenGL Speedup: 0.5x


Analysis
These numbers were so odd I ran the all the tests over again; the results were the same.

Not surprisingly, the BootCamp (BC) results show significant speed-up when moving from software shading to hardware, the hardware running 7.5x faster.

Also not surprisingly, the PW VM showed a decrease in performance when moving from the cinebench shader to the OpenGL shader. This really shows that the software shader in the Cinebench engine (derived from Cinema 4D) is better than the software emulated OpenGL. These results really show why there is a place for BootCamp even after PW ships; it will just never be able to get the same graphics performance as raw hardware. (Note: These numbers don't indicate a bug in PW; you get similar ratios with VMware.)

What is odd are the absolute values for the software shader: 274 for BC and 676 for PW. Visually watching the benchmark it was clear that something was wrong, since BC was subjectively much faster than PW. My suspicion is that whatever mechanism Cinebench is using to time its results is broken when running in a PW VM, which means that the absolute numbers don't really have any meaning. It's probable, however, that the relative values within a single run do have meaning and so the slow-down when running OpenGL HW-L in the VM is probably real.

--Tim

 

Thanks alot Tim for these investigations !
I guess that next step, for CAD engineer, would be to make videos of CAD application running inside PW...

 

I think you can get a pretty, albeit subjective, feel for OpenGL performance, in Winodws, under PW just by running any of the OpenGL screen savers.
Just to be fair I used Hyperspace or helios OGL screen savers, honestly they run so poorly I was pretty surprised.
This is one area where I have to give VM technology like Parallels a pretty poor grade. I love the concept and the product, I really do, however I think OGL is going to be a sore spot for people relative to a pure x86 experience like Bootcamp.

cheers,
Peterwor

 

seems that somebody tried Solidworks, and it worked great...good news for CAD engineers...
anyone tried another cad app ? autocad, ug, catia, vectorworks, microstation....

 

Just installed UG NX4 and it works great. It's clearly doing software rendering as I can see my CPU spike during model rotates, but the frame rate is quite acceptable at around 10-20fps. I'm hoping they continue to add OpenGL optimizations over time, passing more and more of the OpenGL calls directly through to the video card if that's possible, but it's completely useable now.

Has anyone had any luck getting a 3DConnexion SpaceMouse or SpaceTraveler to work yet?

 

whaou ! I am buying my MBP this WE !
would you do a film ? ?
what is your config ?
what kind of model did you try ? many pieces or only one or two ?
thanks for the informations !

 

To be fair, I don't think any virtualization product for the Mac was able to properly get OpenGL accelerated by the Mac host video.

I'd love to be corrected, though. =)