Uploaded July 2022 | Updated November 2026, 3 hours ago
This is a quick FluidX3D CFD simulation of the Boeing 757 at Re=100k. 912×1824×456 grid resolution, 100k time steps, 2h54min compute+render time (~80min compute, ~94min render) on an A100 40GB GPU.
Timestamps:
0:00 grid view
1:00 follow view
2:00 bottom view
2:59 side view
3:59 top view
Since the last video I have added Smagorinsky subgrid turbulence model to get turbulence down to the resolution limit while still having a stable simulation.
Rendering: velocity-colored Q-criterion at Q=0.00005.
I wrote the FluidX3D software from scratch during the last 4 years. Here it's standard lattice Boltzmann method (LBM), D3Q19, BGK collision, mid-grid bounce-back boundaries. For the box walls I use equilibriium boundaries. My implementation is ultra (memory) efficient with Esoteric-Pull streaming and FP16 memory compression, reducing memory demand to 55 Bytes/node, 1/3 of other FP32 implementations and almost doubling performance. This way I can get extreme resolution even on a single GPU with limited memory capacity.
Compute efficiency on the Nvidia A100 40GB is ~79% at 16000 MLUPS/s.
Esoteric-Pull: doi.org/10.3390/computation10060092
FP32/FP16 mixed precision: researchgate.net/publication/362275548_Accuracy_and_performance_of_the_lattice_Boltzmann_method_with_64-bit_32-bit_and_customized_16-bit_number_formats
FluidX3D: doi.org/10.15495/EPub_UBT_00005400
#FluidX3D #OpenCL #CFD
This is a quick FluidX3D CFD simulation of the Boeing 757 at Re=100k. 912×1824×456 grid resolution, 100k time steps, 2h54min compute+render time (~80min compute, ~94min render) on an A100 40GB GPU.
Timestamps:
0:00 grid view
1:00 follow view
2:00 bottom view
2:59 side view
3:59 top view
Since the last video I have added Smagorinsky subgrid turbulence model to get turbulence down to the resolution limit while still having a stable simulation.
Rendering: velocity-colored Q-criterion at Q=0.00005.
I wrote the FluidX3D software from scratch during the last 4 years. Here it's standard lattice Boltzmann method (LBM), D3Q19, BGK collision, mid-grid bounce-back boundaries. For the box walls I use equilibriium boundaries. My implementation is ultra (memory) efficient with Esoteric-Pull streaming and FP16 memory compression, reducing memory demand to 55 Bytes/node, 1/3 of other FP32 implementations and almost doubling performance. This way I can get extreme resolution even on a single GPU with limited memory capacity.
Compute efficiency on the Nvidia A100 40GB is ~79% at 16000 MLUPS/s.
Esoteric-Pull: doi.org/10.3390/computation10060092
FP32/FP16 mixed precision: researchgate.net/publication/362275548_Accuracy_and_performance_of_the_lattice_Boltzmann_method_with_64-bit_32-bit_and_customized_16-bit_number_formats
FluidX3D: doi.org/10.15495/EPub_UBT_00005400
#FluidX3D #OpenCL #CFD


![FluidX3D [FLUID SIMULATION] - Real Time Free Surface LBM with Volume-of-Fluid
FluidX3D is a real time 3D fluid simulation based on the lattice Boltzmann method. It is written in OpenCL C (GPU code) and optimized to the physical limit (video memory bandwith, ~520GB/s).
All demonstrations in the video are simulated and visualized in real time on a Nvidia Titan Xp. Graphics are done with the OpenCL C version of Line3D, the fastest graphics engine ever written for primitive shapes like lines, dots and circles. It can handle up to 2 billion lines per second.
The free surface is done with the volume-of-fluid (VoF) extension to LBM. It allows simulating free surfaces with a sharp interface layer. Having to avoid race conditions makes an efficient GPU implementation of VoF is quite challenging. Currently there is no surface tension yet; the curvature estimation algorithm is still under construction.
The volume force (gavity) is implemented with the Gou forcing scheme.
Hardware Setup:
CPU: Intel Core i7-8700K @ 4,5GHz all core
GPU: Nvidia Titan Xp
RAM: 2x8GB Corsair Vengeance LPX DDR4 3200MHz CL16
Mainboard: ASUS ROG STRIX Z370-I GAMING
SSD 1: Samsung 970 PRO 512GB
SSD 2: Samsung 860 EVO 1TB
CPU Cooler: Cooltek LP53
PSU: Corsair SF600
Housing: Fractal Design Node 202
Timestamps:
0:00 - Intro
0:02 - honey coiling
0:34 - large drop impact
0:45 - drop splashing in flowing water
0:55 - throwing drop sideways in resting water
1:00 - eternal periodic water tap
1:07 - dam break without obstacles
1:18 - asymmetric dam break with two connected chambers
1:29 - droplet impact
1:39 - eternal periodic fountain
1:50 - dam break with cuboid obstacle
2:06 - more honey coiling (just because it is so cool)
More information at http://www.projectphysx.de
Music: Notaker - Fatal System Error FluidX3D [FLUID SIMULATION] - Real Time Free Surface LBM with Volume-of-Fluid](https://i.ytimg.com/vi/a1u2g9ahIDk/mqdefault.jpg)
![FluidX3D [FLUID SIMULATION] - Smashing the Memory Wall with FP16 | 100 Subscriber Special
FluidX3D is a real time 3D fluid simulation based on the lattice Boltzmann method. It is written in OpenCL C (GPU code) and optimized to the physical limit (video memory bandwith, ~520GB/s).
All demonstrations in the video are simulated and visualized in real time on a Nvidia Titan Xp. Graphics are done with the OpenCL C version of Line3D, the fastest graphics engine ever written for primitive shapes like lines, dots and circles. It can handle up to 2 billion lines per second.
LBM performance is only limited by video memory bandwidth - the so-called memory wall. Storing the LBM distribution functions in memory as FP16 instead of FP32 cuts the memory bandwidth requirements in half, doubling the simulation speed at the cost of a bit of accuracy. My implementation has a hardware efficiency of up to 95%, meaning that 520 GB/s of the total 548 GB/s of video memory bandwidth are used. This results in a peak performance of 5.2 GLUPs (giga lattice updates per second, how many LBM lattice points are processed in 1 second) for the D3Q19 velocity set.
More information at http://www.projectphysx.de
Music: Notaker - Fatal System Error FluidX3D [FLUID SIMULATION] - Smashing the Memory Wall with FP16 | 100 Subscriber Special](https://i.ytimg.com/vi/aWFi__zMUkk/mqdefault.jpg)






