Page 4 of 4

Re: RealFlow 10

Posted: Wed Dec 14, 2016 12:38 pm
by angel.tena.tena
The improvement is noticeable when you deal with large-scale simulations, which is actually the purpose of the Hybrido solver. Just check the demo project Hybrido Uboat which is a good candidate to get a speed up. Simulate that project with RealFlow 2015 and then with RealFlow 10. In a regular Intel Core i7-2600K @ 3.40 GHz we get exactly a 2x speed up here.

rf_screenshot.jpg
rf_screenshot.jpg (150.27 KiB) Viewed 2471 times

Re: RealFlow 10

Posted: Wed Dec 14, 2016 5:04 pm
by ChristianZ
Ok, that is interesting as a benchmark test. I will have a look at it using one of my Xeon-Macs (probably takes some time until the machine is free, so you can expect my results next week).

Re: RealFlow 10

Posted: Wed Dec 14, 2016 5:19 pm
by Forester
Oldcode, RF10 works with Lightwave.

I am just now installing my upgrade, after playing with the Demo for a week or so. My upgrade is showing an "RFConnect" for Lightwave, and also for Softimage. So, the nice Next Limit folks are not leaving the old-but-good software users out in the cold.

Re: RealFlow 10

Posted: Thu Dec 15, 2016 1:07 am
by Oldcode
Thanks forester! :)

Re: RealFlow 10

Posted: Thu Dec 15, 2016 1:15 pm
by stefan_bartram
:) that sound good if the speed really is 2x as I mainly used the hybrido in RF, Unfortanlly I will have to set up that scene again and test it, as you can't open it with the demo version :( . And the speed of hybrido is very essential for me if I shall used 1.500 euros on upgrades. By the way , this is the answer I got from Next Limit.

* Case ID: 00008755
* Case tittle: CGP render
* Message:

" Hi Stefan,

In version 10 of RealFlow no improvements have been made using GPU and you may get better results using CPU. The x2 speed improvements in version 10 are with Dyverson with GPU.

Thank you for contacting us.

Best Regards.

Carlos G. "

Re: RealFlow 10

Posted: Fri Dec 16, 2016 8:30 am
by tsn
@stefan_bartram

Thanks for posting the answer you got from NL. Unfortunately, the information provided is misleading and actually not correct.

With Dyverso we have the following improvements:
GPU 3-8x (over CPU)
GPU CUDA 3x (over RealFlow 2015)
GPU OpenCL 3x (over RealFlow 2015)

With Hybrido the speed boost is up to 2x for large scenes (=several millions of particles).

You're also asking if NL stopped improving Hybrido? Of course not, and there are many new things in RF10:

- The "Dense" creation mode has been removed and the grid is always sparse, because we completely moved to OpenVDB in RF10.
- This means a huge improvement in memory consumption compared to scenes where the "Dense" mode was used.
- New, more realistic, way of generating Hybrido foam.
- "Time Scale" option now affects all Hybrido secondary fluid emitters.
- Huge speed improvement in calculating the volumes of interacting objects.

I hope this helps to clarify a few things. Thanks for bringing this to our attention!

Cheers,
Thomas

Re: RealFlow 10

Posted: Fri Dec 16, 2016 9:13 am
by stefan_bartram
Thanks Thomas , that sounds good :) :)

Re: RealFlow 10

Posted: Fri Dec 23, 2016 9:03 am
by stefan_bartram
Hi , I finally got a test done . It's basic setup with defalt setting beside one thing , the cell offset is change (in both). They are simulationed on the same machine. CPU E5-1650 v4 3.6 Hz 6 cores (12) with 88 GB ram and GTX 1080

145 frames, 30 mill cels, 40 mill particles , file size 1.3 gb to 1.9 gb pr. frame, which took 12.10 hours in 2015 and 7.14 in 10, so FR 10 was 5 hours quicker :) :)

here are the last frame info for 2015

----------------------------------------------------

DasBoat: [25,566 vertices - 50,708 faces] [Z:/RealFlow/submarine/DasBoat.sd] [vx
:0]
HY_Closed_Domain_Container: [8 vertices - 12 faces] [] [vx:0]
HY_Closed_Domain_Manipulator: [0 vertices - 0 faces] [] [vx:0]
HY_Closed_Domain_Water: [8 vertices - 12 faces] [] [vx:0]
HY_Closed_Domain_Domain: [186x68x686 8,676,528 cells] [41,693,578 particles] [0.
300000 cell size]
Gravity01

Timing information
-------------------
Elapsed time: 12:14:53
Remaining time (estimated): 00:00:00
Average time (frame): 00:05:04

>21:37:56: ================ Frame finished ================


and for Rf 10

DasBoat: [25,566 vertices - 50,708 faces] [Z:/RealFlow/submarine/DasBoat.sd] [vx
:0]
HY_Closed_Domain_Container: [8 vertices - 12 faces] [] [vx:0]
HY_Closed_Domain_Manipulator: [0 vertices - 0 faces] [] [vx:0]
HY_Closed_Domain_Water: [8 vertices - 12 faces] [] [vx:0]
HY_Closed_Domain_Domain: [191x91x686 11,923,366 cells] [41,346,689 particles] [0
.300000 cell size]
Gravity01

Timing information
-------------------
Elapsed time: 07:14:52
Remaining time (estimated): 00:00:00
Average time (frame): 00:02:59

>21:31:59: ================ Frame finished ================


SO IT LOOKS VERY GOOD

and may you have a merry christmas and a happy new year

Re: RealFlow 10

Posted: Fri Dec 23, 2016 9:10 am
by ChristianZ
Its good to hear that your test went well!

Wish you Merry Christmas and nice holidays

Re: RealFlow 10

Posted: Fri Dec 23, 2016 10:52 am
by tsn
Thanks Stefan for checking out and confirming our observations. Great work!

Season's greetings and Happy New Year,
Thomas