[General] Rendering Speedup (3 Minutes per Step)

Tobi · July 15, 2014, 14:05

Dear all,

I have just a very simple question to you.
While rendering my actual project (multiphase ~ 2.000.000 cells, hexadominant from snappyHexMesh) one rendering step takes round about 3 minutes. I have 1200 steps to render. Therefor I should run my computer for 3600 minutes --> 60h.

This time is "okay" if I will render it during weekend but I think this problem could be speed up in an very good way, isn't it?

First some information

My personal desktop pc:

CPU: FX-8150 (8 xore)

GPU: NVIDIA GTX 560

Mainboard: Crosshair Formula IV

My cluster:

CPU: 2x Opteron 6200 (12 Core)

Mainboard: Supermicro h8dgi-f

GPU - Onboard 16 MB (so none)

At the moment I use paraview on my desktop PC and the rendering step take ~ 3 minutes.

So is there a way to make it faster, any link, literature, hints are recommended.

I think there are users that doing high rendering with PV.
There should be a way to speed it up!

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

Additionally:

setting up a PV server on my cluster with OSMesa and off-screen-rendering does not work --- I can not see volume rendering on my client.

setting up a pv server on my local machine is the same as use pv as single client - render time ~ 3 minutes.

Regards Tobi

Quefrency · November 12, 2019, 02:56

Hi,

Did you manage to speed up your renders? I am sitting in a similar situation that volume renders take very long to load.

flotus1 · November 19, 2019, 10:43

With a "traditional" CPU rendering approach, there are usually two bottlenecks that can be reduced:
1) CPU processing time. Start Paraview in parallel on more than one core
2) I/O bottleneck. Paraview has to read the data for each new step from disk. Use faster storage, ideally NVMe SSDs or something with similar read speed. And of course, make sure there are no slow interconnects between the machine that runs Paraview, and the storage device.

Quefrency · November 20, 2019, 01:33

Thanks for the reply,
I have only just recently learnt about running paraview in parallel which is helping but it seems it will only run in parallel automaticity with certain data while a recomposed case will not, and a decomposed case has volume rendering problems on my Mac. It noticed the storage speed problems, unfortunately my mother board can’t take NVMe but I bought another SSD for my computer and more ram. So I hope it helps.

July 15, 2014, 14:05	Rendering Speedup (3 Minutes per Step)	#1
Tobi Super Moderator Tobias Holzmann Join Date: Oct 2010 Location: Tussenhausen Posts: 2,708 Blog Entries: 6 Rep Power: 51	Dear all, I have just a very simple question to you. While rendering my actual project (multiphase ~ 2.000.000 cells, hexadominant from snappyHexMesh) one rendering step takes round about 3 minutes. I have 1200 steps to render. Therefor I should run my computer for 3600 minutes --> 60h. This time is "okay" if I will render it during weekend but I think this problem could be speed up in an very good way, isn't it? First some information My personal desktop pc: CPU: FX-8150 (8 xore) GPU: NVIDIA GTX 560 Mainboard: Crosshair Formula IV My cluster: CPU: 2x Opteron 6200 (12 Core) Mainboard: Supermicro h8dgi-f GPU - Onboard 16 MB (so none) At the moment I use paraview on my desktop PC and the rendering step take ~ 3 minutes. So is there a way to make it faster, any link, literature, hints are recommended. I think there are users that doing high rendering with PV. There should be a way to speed it up! ------------------------------------------------------------------------------------ Additionally: setting up a PV server on my cluster with OSMesa and off-screen-rendering does not work --- I can not see volume rendering on my client. setting up a pv server on my local machine is the same as use pv as single client - render time ~ 3 minutes. Regards Tobi __________________ Keep foaming, Tobias Holzmann

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
AMI speed performance	danny123	OpenFOAM	21	October 24, 2020 04:13
Stuck in a Rut- interDyMFoam!	xoitx	OpenFOAM Running, Solving & CFD	14	March 25, 2016 07:09
Superlinear speedup in OpenFOAM 13	msrinath80	OpenFOAM Running, Solving & CFD	18	March 3, 2015 05:36
Help for the small implementation in turbulence model	shipman	OpenFOAM Programming & Development	25	March 19, 2014 10:08
How to write k and epsilon before the abnormal end	xiuying	OpenFOAM Running, Solving & CFD	8	August 27, 2013 15:33

November 12, 2019, 02:56		#2
Quefrency New Member Kyle Moon Join Date: Apr 2019 Posts: 5 Rep Power: 7	Hi, Did you manage to speed up your renders? I am sitting in a similar situation that volume renders take very long to load.

November 19, 2019, 10:43		#3
flotus1 Super Moderator Alex Join Date: Jun 2012 Location: Germany Posts: 3,400 Rep Power: 47	With a "traditional" CPU rendering approach, there are usually two bottlenecks that can be reduced: 1) CPU processing time. Start Paraview in parallel on more than one core 2) I/O bottleneck. Paraview has to read the data for each new step from disk. Use faster storage, ideally NVMe SSDs or something with similar read speed. And of course, make sure there are no slow interconnects between the machine that runs Paraview, and the storage device.

November 20, 2019, 01:33		#4
Quefrency New Member Kyle Moon Join Date: Apr 2019 Posts: 5 Rep Power: 7	Thanks for the reply, I have only just recently learnt about running paraview in parallel which is helping but it seems it will only run in parallel automaticity with certain data while a recomposed case will not, and a decomposed case has volume rendering problems on my Mac. It noticed the storage speed problems, unfortunately my mother board can’t take NVMe but I bought another SSD for my computer and more ram. So I hope it helps.