http://www.3dspeedmachine.com/News/compare/compare3.htm
Please do not post links to those Tests anywhere, they are heavily flawed !!
Those are VERY staged to make XSI look best and Max look worst
I am arguing via email about this with Testauthor Jon Peddie about the Tests
but got only one responce till now…
You can DL the test-scenes yourself to see the faulty setups, and ALL of them
are favoring XSI via various settings in the scenes. Well Softimage commissioned
the Benchmark, but i don’t believe they really wanted risk their good name for such tricks.
Here’s a list of flaws i found till now:
* General Flaw for all Tests:
Max’s Viewport poly/vertex counter causes a performance hit if geometry
reevaluation is required somewhere in the modifier-stack (skin in this case)
The hit in “jogger” is at least 20% on my System. Just toggling poly/vtx
display to OFF, leaving only the FPS counter will show that immidiatly…
* “a 1000 cubes” - Flaw 1:
XSI performance is measured in BOUNDINGBOX mode, whereas Maya+Max are
in Wireframe. Result: XSI is 20% faster. If you turn on Wireframe in
XSI too, its actually 20% slower than Max (on my machine) with Maya
being fastest. Boundingbox in Max is nearly double the Speed of XSI
* “a 1000 cubes” - Flaw 2:
XSI + Maya uses optimized rotation-keyframes ( 1 key per 40 frames)
Max has set a key per frame, which causes a viewport hit because
of additional interpolation overhead. Additionally Max+Maya’s keys are
set to smooth interpolation, whereas XSI’s are set to linear.
* “Particles_with_4_goals”
In XSI’s + Maya’s scene, particles are NOT moving towards their goals as intended.
They simply move up vertically in one solid stream and converge to a point.
In the Max scene they stream is actually splitting and moving towards the 4 goals.
* “Particles_Along_Curve”
It’s statet that Max can not perform a simple “Particle along Path” anim with PFlow,
but that’s totally untrue. I’m rather unfamiliar with PFlow, but in 10 minutes i got the
solution for particles following the path (same particlerate as XSI)…
Result: 5 secs in Max, 8 secs in XSI on my System
* Massive_Urching_Twisting"
Of course the XSI-ICE Twist method is faster than the good old
“Twist-Modifier” in Max. But looking at the Maxscene, you will find several keys hidden
in Maxs “Edit Poly”-Object and the “Twist-Modifier” besides the animated twist and
rotation keys. Of course performance is slower in Max, if you set some keyframes
on Softselection-Falloffs in Editpoly, animate the Twist-Modifiers Center and several
other keyframed animation going on in the Max-Scene. Again all Keys in Max are set up
on a “one key-per-frame” base, whereas the XSI Scene only has two animated parameters
(twist + rotation) and this on a “one key per ~40 frames” base
Come to your own conclusion about this Benchmark. It’s really hard to defend Max in these days,
but in this case i feel the strong urge to do so.
I will open a thread on cgtalk in the Max section to discuss this matter, because i find this degree
of bias absolutly unacceptable…
