Newton 3 performance vs. Newton 2

A place to discuss everything related to Newton Dynamics.

Moderators: Sascha Willems, walaber

Newton 3 performance vs. Newton 2

Postby pHySiQuE » Fri Aug 09, 2013 1:11 pm

Is there any reason Newton 3 performance for a large number of falling boxes would be slower than Newton 2? I have 1000 boxes falling, and they are spread out so they don't stack much. In Newton 2 the framerate stays near the max, and in Newton 3 the simulation grinds to a halt after a few seconds.
pHySiQuE
 
Posts: 608
Joined: Fri Sep 02, 2011 9:54 pm

Re: Newton 3 performance vs. Newton 2

Postby Julio Jerez » Fri Aug 09, 2013 1:31 pm

newton 3.10 is much faster than any proviuos version of newton.

It is also faster that Physx 3.3, according to their own version of the Nvidiea Tool PEEL that they facilitated to me under NDA.
Accourding to that tool newton is equally fast or faster than Physx in all of thier rigid bodies and collision queries test.
They, true a senior employee promised to release the result before the integration o fnewton to teh tool and I did,
but after the integration they run the test and they retrated saying they where not going to be able to publish the result.
Aparrently teh tool is not endorsed by Nvidea even thought it is written by some of thier most senior people there.
Aparenetly the decided not to relase the tool altogether.
The bough int the Propaganda started by the self appointed and dishones experts of these "Open source engines" spined out from Havok alter version fo Havok
and they thought the comparison to Newion was going to be a blow blowout, but they were ones they were sorry were them.
and now the Cat is out of the Bag.

According to that Tool all versions of Physx are from 10 to 100 time faster than all of other open source libraries integrated to that tool
but the had the suprice of thier life when they found that newton is faster than thier faster version, Phsyx 3.3 while at the same time being more acurate
It end up they having to tweak Phsyx so that they can match Newton combined Speed, accuracy, stability and feature set.

what are you dowing that you find it slower? I have not found a single case where it run slower than a previus version.

1000 or 2000 falling baxe should take less that a 0.1 ms, and it would be less if is was no for the thread switch overehead.
the terrain hiegh field demos in teh sanbox demos do just that.
Julio Jerez
Moderator
Moderator
 
Posts: 12426
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: Newton 3 performance vs. Newton 2

Postby pHySiQuE » Fri Aug 09, 2013 2:36 pm

It was user error. Sorry. :P

Thanks for the explanation.
pHySiQuE
 
Posts: 608
Joined: Fri Sep 02, 2011 9:54 pm


Return to General Discussion

Who is online

Users browsing this forum: No registered users and 2 guests

cron