#Particle FX
15 messages in this thread
– Reply to message #2038 (already scrolled)
Pawel,
Yes, if there are no other apps running, the rendering will be faster. BUT,
the Mac, and Windows, are cooperative multi-tasking environments. Most programs
have a main loop running, every x times through the loop, Mac programs should
call the "GetNextEvent" (if I remember correctly). This Mac Toolbox function is
what lets the Mac operating system and other programs in the background have
some CPU cycles. The program/programmer can be Uncooperative and never call
that function, which would mean the app would take over the machine.
Ken was eluding to reducing the number of times per second that this is
called, or quiting the finder (The macs program/file manager) which is really
just another program running in the background. So, even if there are no other
apps running, the Finder is considered an app and uses CPUs cycles. My vote
would be to not quit the Finder, BUT put a menu/check option to NOT call
"GetNextEvent" and effectivly have total control of the machine (if that truly
is possible!).
Anyway, your new purchase sounds GREAT! Please let me know the render time
improvement. Also, have you experimented with the 9-pass motion blur, and if
so, how much longer did it take to render? And finally, have you used the
Camera Haze much, if so, do you have some tips as far as settings go? That
would save me hours of "test" rendering!!
Thanks alot and enjoy the new machine <G>!!!
Jim Jonah
p.s. Happy Thanksgiving!!!
Jim,
I never used motion blur, but did play with the haze a bit… I would sy that
if you want just a little fog you should set the haze value pretty high 12000
— 20000. If it is too low the "air depth" will be look fake. This is
especially effective if you are going to use A:M with VistaPro (Or Bryce)
images that were rendered with some haze…
Pawel
Pawel,
Thanks for the numbers on the Camera Haze. Just curious, do you know what the
numbers stand for? Percentage, pixels, etc…?
Also, just curious, do you think that the Pentium "division" bug will affect
A:M renderings?
Even though I love modeling and animating on the Mac, I am seriously
considering a Pentium or two for rendering. I'd be curious to know what the
time percentage difference is between a frame being rendered at 640×480 with
anti-alias and photo quality in Render24 vs. 640×480 photo and 9-pass blur.
While the PowerMac and the Compaq dx2/66 were fairly close in the
photo/anti-alias settings for my latest animation (30 minute to 40 minutes per
frame — lots of spine motion) when I changed to 9-pass motion blur, the
PowerMac jumped to 1 hour 54 minutes while the DX2/66 jumped to 5 hours 55
minutes!! Yikes!! The same file on my Compudyne 486/25sx recently upgraded to a
Dx2/50 was still at about %30 of the first frame after about 3 hour 30 minutes,
so I cancelled it! 🙁
I wonder if the motion blurs are using floating point math, if so this may
explain the PowerMacs performance. BUT for me, these times are just not
PRODUCTION times! So, if you could find the time when you get your new toy (I'm
jealous!) to test the render time differences, I would appreciate it! <G> That
9-pass motion blur just looks SOOOO Good!!!!
Thanks alot!
Jim Jonah
Jim,
I have a P-90 with 32megs. If you would like it, I would be more than
happy to render your "test scenes". Pawel looks like he has enough to do
right now<g>. Just send me some Email and tell me if you want me too.
-Mike Johnson
P.S. The 9-pass motion blur is _VERY NICE_!<BG>
Michael,
>>>I have a P-90 with 32megs.
Do you run Windows with 32 bit disk access (Workgroups or 3.11)?
Pawel
Pawel,
>>Do you run Windows with 32 bit disk access (Workgroups or 3.11)?<<
Unfortunately, I cannot run 32-bit access on this computer due to IRQ
conflicts (And I've been too
laz…er…busy to go hunt them down.<G>)
I was planning to get WFW next week after all the discussion on how much
it helps A:M. But after I found out about the 32-bit access prob., I doubt
I'll be getting it for a while.
Sorry, I didn't realize this was the criteria for the test, I thought
it was just the test of 90Mhz pentium compared to the 8100. Oh well, maybe
next time<g>.
Regards,
-Mike Johnson
Michael,
>>>Sorry, I didn't realize this <32-bit access> was the criteria for the test,
I thought it was just the test of 90Mhz pentium compared to the 8100. Oh well,
maybe next time<g>.
I would be very interested to see your numbers without the Win/Worgroup 3.11
just as much… I hope to be able to install the 32 bit version on my new P90
machine and see the comparison to my current DX2-66 with the standard Win
3.1… When it comes to raytracing _ALL_ computers are so terribly slow that
any information helps in picking the right combination…
Pawel
>>>Thanks for the numbers on the Camera Haze. Just curious, do you know what
the numbers stand for? Percentage, pixels, etc…?
I think the number is the distance (from camera's target) where the background
color will blend with the scene… If the object is 1000 units away from the
camera (the Direction's grid is set for 200) and your haze is 10,000, you
should get about 10% haze. (If I understand it right <BG>)
Thanks for the Pentium bug info. I've never heard about it, si it comes in good
time… I will check it right away when I get my box. I will be happy to do the
testing, since I am just as interested in finding out how little progress has
been made in the last year and a half… (that's the last time I got the "top
of the line").
Pawel
I haven't done a great deal of rendering with motion blur and have not
carefully benchmarked it but it is subtantially slower than
photo-real/anti-alias. I attempted to use it on a moving starfield/star-tunnel
sequence and got some pretty peculiar results taking a whole lot of time to
produce. If you want some precise numbers,
I'll run a quick test. Maybe have something for you tomorrow.
Sincerely, Breanna Anderson
(P.S.) Machine: Gateway 2000 P5-90 (90Mhz, 16MegRam)
Bre,
I would definately be interested in your 9-pass motion blur tests, and even
640 x480 Photo/antialias in render24!
I just finished a small project that I used the motion blur on to simulate
slow-motion. The results were GREAT! I had the action happen quickly in
direction so there was a large change from frame to frame, and them slowed the
sequence down to 25% in Premiere.
While I am a big Mac fan, I also am budget concious. So, if the pentium
results are good, I would use them for render and my PowerMac for Modeling.
Thanks in advance!
Jim Jonah
Well here are my quick results:
My scene contains a single complex figure, a "hood ornament", that is a
metallic
female figure with wings, a heavy-metal angel so to speak. She is beating her
wings down in a single stroke, 30 frames, the camera angle is out near the tip
of one wing so as it passes, the movement is substantial.
Machine: Gateway 2000 P5-90, 16Meg Ram, rendering to a clean local drive.
Test1:
640×480, 24bit, Photo, Anti-alias
Started out around 4:05 per frame. Render times were up a little in the middle
due to large number of ray-bounces off the complex reflective surfaces. Total
time about 2.5 hours.
Test2:
640×480, 24bit, Photo, Motion-Blur
Stared out around 8:10 per frame with upwards over 10:00 per. Total time about
5 hours but didn't see the last frames.
P.S. Machine was otherwise unoccupied.
Conclude:
Motion-blur seems to result in roughly double the rendering time. I could see
the motion-blur effect in Painter when I inspected the frames but the effect
was very subtle at least for this scene. I watched both renderings at length
and darned if the effect was noticible – certainly not worth twice the time.
As far as I can see the motion blur effect only occurs when the displacement is
so great that persistence of vision serves much the same function. On a
per-frame
basis, I though the result was not attractive. If I wanted a single rendered
frame
with a motion-bluring effect, I would probably render multiple frames and
composite
them as an after-effect.
Other impressions of motion-blur?
Bre,
Thanks for the P90 rendering numbers. Your hood ornament "Godess" sounds
interesting!
Was the motion-blur the default (16-pass) or was it the 9-pass (-a3 switch in
render24)? Also, were these DOS times or Windows times?
When I tried the 16-pass, the individual frames didn't look great, but the
9-pass frames did look pretty cool.
I tend to learn a product better when I have a project to work on, so I
decided to recreate parts of the Tide/Arm-and-Hammer boxing commercial. In
addition to what is in the commercial, I wanted to zoom in and do a "slow-mo"
on the box being knocked out of the ring. Over a period of nine frames, the
boxing glove/arm does a complete 360. Since there is so much movement from
frame to frame, the blur was pretty strong. Then is premiere I slowed it down
to 25%, the effect seems to work well for me. I may upload a flic of part the
animation showing the slow-mo part, if I can get the file size down low enough.
Thanks again for your help <G>, if your blur was the 9-pass, those times are
good enough to justify a P90, for me.
Jim
Your "hood ornament" animation sounds cool. Could you maybe upload it for the
rest of us to admire?
I tried the trusty clam.cho provided with the program for more tests on my NEC
Ready 433es (souped up to a 486DX2-66 with 20 meg and Windows 3.11 with
32-bit access enabled for disk and files, a 4meg cache setting in Windows,
and Virtual Memory set to none and Drive Rocket software running, prefs.hsh set
at 16000KB).
In the Windows Render, set at 640×480, photo, anti-alias and 24-bit selected,
the
first frame (usually the quickest render along with last frame) came in at
3:43.
Frame 15 (the biggest clam image..wide open with tongue out) took 5:35. All 30
frames were finished in 2:16:01.
Same .cho with 640×480, photo, 24-bit and motion-blur selected, the first frame
took 9:08 and frame 15 took 15:11. All 30 frames were rendered in 6:02:59.
Motion blur is not good looking for still frames IMHO but works when used
sparingly with fast motion.
Kevin
Pawel,
I just found this in another forum, thought you might find it interesting, or
depressing 🙁
>>Here's another way to check you numeric processor for those of you who don't
own SPSS:
If you are running Windows, you can test your Pentium chip using the
Calculator in the Windows "Accessories" group. Just bring up the Calculator and
enter the following keystrokes exactly as shown here:
4195835/3145727=*3145727=
If you get 4195835, you're OK. If you get 4195579, you have a Pentium with a
bad numeric processor. Any program that uses the numeric processor's
double-precision floating-point division will give the wrong answer for this
particular problem with the Pentium chip, but not with a 486 or 386. It seems
that it is possible that Intel corrected the problem for the more recently
released Pentium machines, but I haven't heard of a report yet of a Pentium
that doesn't do this.<<
Don't you just love computers? <G>
Jim Jonah
p.s. you can just copy the whole equation and paste it right into the
calculator!