On my 2xE5-2687W v4 it took 34sec on Windows.
Thats on machine, that cost more then my car

KaiStavginski
I'm guessing the fact it took longer without args is just a coincidence? Did you run multiple times?

Did you run that first, maybe? Could be disk caching.

Just to throw in some more numbers - HairSOP2.hiplc took 16.75 seconds on my i7 5960x. It seems like we're having some trouble on threadripper - or maybe on windows. I'll see if I can find out more.

Hello martin, thanks for bringing in your test, it seem that houdini start to behave really bad above a certain amount of core count, could you try launch houdini with the -j 8 and -j 16 parameter, and see how the speed behave ? I have that feeling that you'll get the same performance.

I received my new quad channel memory last week, my old stick being :
3x 2400mhz Single channel
my new memory stick are now :
FlareX 3200mhz-14Q Quadchannel
I did the same test with HairSOP2 again, (less this time)


ps: second cook refer to forcing houdini to recompute the fur by changing the density a little bit

I'll also put the old data :



you can see that the new ram improved the performance by at least 50%, but above 8 core I almost don't get any gain. (I am supposed to have 16 core, I would expect to have at least 50% more performance when enabling them)
could there be a memory bottleneck ?

I noticed there is 2 node that take up for 70% of the hair generation part :
the pointwrangle2 (6sec of the 16sec required to generate fur on my machine)
set_animated_point_attribs (4s of the 16s)

ps: disabling the compile block didn't affected the performance.

with -j2 im getting 47s, -j8 24s, -j16 24s and without limiting 34s.
Funny thing is that i can see that its targeting 8, 16,48 cores. With probably same CPU utilization.

PS: correct about the compile node, while disabled, performance stays the same

Thanks martin! I was sure that would confirm my theory, your test match the one I did, you're getting better performance by using less core and it not amd related
It seem to happen above 8 thread.

Could SideFX also check that and keep us informed?

Is there something the user can do to prevent a hair groom to recook every time the scene is opened?