[ec2m] Physics engines: MathEngine/Karma/Vortex v. Havok

[Taylor, Tim]

Dear all,

Having just returned from last week's ECAL and
talking with various people there, it certainly
seems that many groups are currently deciding
on what physics engine to use for future projects.

In particular, most of the people I spoke to were
wondering what the differences were between Havok
(http://www.havok.com) and MathEngine (Karma)
(http://www.mathengine.com) or Critical Mass (Vortex)
(http://www.cm-labs.com). (For those who don't know,
Critical Mass is a recently independent offshoot of
MathEngine. Critical Mass will concentrate on the
academic and engineering market, while MathEngine
will concentrate on developing for computer games
companies. Their respective products, Vortex (Critical
Mass) and Karma (MathEngine) are more or less
identical at present, although development will
diverge in the future).

I have used previous versions of MathEngine and
Havok, but haven't had a chance to test the latest
releases of either yet, so I'm not really in a position
to give a definitive comparison.

Havok have just brought out a fairly major new release,
which I am in the process of incorporating
into my code (lots of API changes...), so I
haven't had a chance to evaluate that yet. Also,
I (still) haven't found the time to look at Vortex/
MathEngine 2.x, so I really can't give you a fair
comparison of the latest offerings from these companies.

I know that my ex-MathEngine colleague, Colm Massey,
is now using the latest MathEngine/Vortex software, and
has been finding it very stable/a vast improvement on
earlier versions (e.g. SDK 1.x).

On the other hand, Havok also continues to bring out
new and improved releases.

On the business side, although Havok certainly seems to
have more customers at the moment, MathEngine has
recruited some high-profile staff recently (including
Mike Gamble, ex-Microsoft, as a director), so I can't
see them going down in the near future... although what
this means for Critical Mass, who knows?

I took the liberty of emailing Critical Mass a few days
ago to ask them why I should choose Vortex over Havok.
The response I received didn't go into too many details
from a technical point of view, although they had this
to say about their academic licence arrangements v. Havok's:

"Our model allows for up to 5 users on a single license; even though it may
cost more than Havok's, we give more flexibility in that regard. We have
found that most labs have multiple people using it, which is why we chose
this model. In addition, as a user you qualify for all point upgrades and
maintenance releases, and some e-mail support (depending on support
availability). For full access to support, we charge the same support fees
as for the commercial, since it is largely based on how much time we have to
spend, and can't discount that too much :)"

Amongst other information, they also provided the following
comparison between Vortex (Critical Mass) and Karma (MathEngine):

"Karma is a real-time physics toolkit sold by MathEngine plc in the computer
games market. Based on the same technology as Vortex, there are many
differences between the two toolkits. Karma was developed to be as fast as
possible and to run on lower end console machines. Vortex is targeted to
engineering applications. The differences in the code include:

1. Karma doesn't have mesh collisions or polygon soup.
2. Karma doesn't allow rotational (inertial) effects
3. Karma has Playstation and XBox versions, but Vortex does not.
4. Vortex is available for SGI IRIX, SGI Linux, Linux and Windows platforms.
5. Karma is available for Windows and Console platforms.
6. Vortex has improved stability in the dynamics and better contact
generation.
7. Vortex has an object management layer enabling quick development of
projects.
8. There is a 3DS Max plugin available for Vortex.
9. Joint limits work better in Vortex, particularly for the actuated all and
socket joint."


If anyone has anything to add to this, or has had experiences
with other physics engines, please post to the list, as I'm
sure that many other list members would be interested!

One last thing before I sign off - here at IC CAVE we are
going to seek funding to develop a robust physics engine based
upon Featherstone's method plus collision detection between
multiple bodies (i.e. pretty similar to what Sims used, and
more robust - but a bit more restrictive - than MathEngine,
Havok etc.) It would be designed with evolutionary algorithms
in mind - in particular, it would provide a very easy way to
farm out evaluations to networked computers to parallelize the
process. The idea is that the final system would be open source
and free. I would be very interested to hear anyone's thoughts
on this.

Okay - that's enough from me for now!

Tim


Tim Taylor_________________________________________________________
Artificial Life Research Associate, IC-CAVE (http://www.iccave.com)
University of Abertay Dundee, Bell Street, Dundee DD1 1HG, Scotland
mailto:tim.taylor at abertay.ac.uk             tel. +44-(0)1382-308959
http://computing.tay.ac.uk/timtaylor/       fax. +44-(0)1382-308627