|
paigan0
|
 |
« on: July 19, 2016, 12:01:20 AM » |
|
I make 4-minute fractal animation music videos and need around 7-10K frames per video. I love a lot of the new features of currently version 2.08.1, such as the ability to take every parameter and turn it into a keyframe parameter to be morphed between keyframes, but it'd be really, really nice to have more Open Cl support. I use two MSI Geforce 980s to render in 2 days what takes my 240-core render farm around 2-3 weeks to a month to render by CPU. The difference in hardware and especially power/electricity and noise from all of the PCs makes the GPU rendering very much the more attractive way to go. Mclarekin has been helping me export some v2.08 formulas into v 1.21-friendly form.
I'd love to be either to use v2 with my GPU (and then get the new GTX 1080), or else have a folder in my v 1.21 version full of the v2.08 fractals formulas. mclarekin has been great at feeding me more and more .c files, and we'll hopefully have several for others to use. I'd about given up on just doing Mandelboxes and bulbs with the GPU when mclarekin took pity on me.
I'd love to put the 240-core render farm to mothballs and get into the GPU business with Mandelbulber, same as Blender will now allow.
I appreciate the awesome work you all do! Thanks so much!
|
|
|
|
« Last Edit: July 19, 2016, 12:16:48 AM by paigan0 »
|
Logged
|
|
|
|
|
paigan0
|
 |
« Reply #1 on: September 07, 2016, 06:23:52 PM » |
|
I accidentally loaded up 2 instances of Mandelbulber v.1.21 at a time--which I've done before--but I never realized that the two instances could run just fine next to each other. I used to have 2 GeForce NVDIA MSI 980s installed, one in two different computers. I had previously had both 980s in the same PC, and I SLI'd them together, but Mandelbulber would only ever use one at a time. I could switch from one to the other card, but never get them both to work.
I finally started two instances of Mandelbulber Open CL and started one normally, but started the other one using the Device Index dropdown on the OpenCL tab and I changed the 2nd instance to 1 and kept the first at zero.
I loaded up the exact same file and keyframes and everything on the 2nd MDBR program and hit rend animation.
THE TWO VIDEO CARDS ARE RENDERING PERFECTLY NEXT TO EACH OTHER ON THE SAME PC!!!! (I'm pretty excited here, you might notice!)
I have the 1080 rendering away on a pc by itself, and then the PC with the 2 980s has two separate instances of Mandelbulber running as well.
That's a 980 and a 980 and a 1080 all making frames!!!!
Just thought I'd share that multi-video card bit of wisdom!
--Stoked Steve (paigan0)
|
|
|
|
|
Logged
|
|
|
|
|
geomagnet
|
 |
« Reply #2 on: October 14, 2016, 05:51:49 PM » |
|
I hope they bring back OpenCL and fix the bugs so that it works with all the formulas. I love animating and can't find the patients to wait 20+ minutes for a frame to render. I found a handful of formulas that won't crash Mandelbulber and WOW, what a difference. Last time I did that I had a single GTX780Ti and it worked fantastic. Now I have 4x GTX980Ti's and would love to get back into animating beautiful scenes.
Also, I have a 36 core / 72 thread workstation and find that only 18 cores are being used. I've tweaked the registry to "unpark" all cores. Even though your settings recognizes all 36 cores, it doesn't use them. What can I do to wake up the other cores?
Thanks James
|
|
|
|
|
Logged
|
|
|
|
|
Patryk Kizny
|
 |
« Reply #3 on: January 30, 2017, 11:29:44 AM » |
|
Looksl ike you have similar config to what I work with. (2x E52696, so 72 cores total) and 2x980Ti+2x1080Ti. I switched long time ago to using synthclipse and fragmentarium, and although these are fully GPU-based, I can not get more than one card run at a time. Looks like I'll be happy to check out OpenCL build again.
Thanks for sharing your insights.
|
|
|
|
|
Logged
|
Visual Artist, Director & Cinematographer specialized in emerging imaging techniques.
|
|
|
|
mclarekin
|
 |
« Reply #4 on: January 31, 2017, 08:06:15 AM » |
|
@ geomagnet. Unfortunately Mandelbulber openCL V1.21 only had a few openCL formulas. Then Buddhis focus was directed to converting the program to QT for the Mandelbulber V2 releases.
With V1.21 , if you tried to use any formula or function that was not adapted for the openCL part, you would crash. It took a while to figure out what worked and what didn't.
The beauty of openCL is still the speed, and the ability to code custom formulas. I have recently converted other formulas to be used with openCL V1.21, and I often test new formulas in OpenCL V1.21 before I code them for the current development version of V2.#.
|
|
|
|
|
Logged
|
|
|
|
|
mancoast
|
 |
« Reply #5 on: April 01, 2017, 05:59:20 PM » |
|
Greetings!
I critically value OpenCL for the close future of the mandelbulber system. We need to get everyone using the same code-base to increase testing.
We will continue to aggressively expedite release schedule of the binaries. OpenCL is next on our list, and its quickly approaching.
We plan to offer designs for custom FPGA hardware for PCI-Express acceleration.
In the near future, we will record our mandelbubler animations using FRAPS.
thanks, coast
|
|
|
|
|
Logged
|
|
|
|
|