Discover and read the best of Twitter Threads about #MI200
Most recents (2)
I did @officialBinotto's @ScuderiaFerrari SF71H in @FluidX3D #CFD on a supercomputer.
- 1s in real life @ 100km/h
- 20s 4K60 video (3x)
- 14h compute on 8x @AMDInstinct #MI200 64GB #GPU
- 144TB data visualized
What I found is absolutely wild. A #SimulationFriday #F1 thread: 🧵1/5
- 1s in real life @ 100km/h
- 20s 4K60 video (3x)
- 14h compute on 8x @AMDInstinct #MI200 64GB #GPU
- 144TB data visualized
What I found is absolutely wild. A #SimulationFriday #F1 thread: 🧵1/5
The @FluidX3D simulation was done at 10 billion voxel grid resolution (2152×4304×1076), over 217k time steps (1 second), at Re=3.75M (100km/h).
The fins on the front spoiler create a turbulent boundary layer and kick up it up onto the front wheels to reduce drag. 🧵2/5
The fins on the front spoiler create a turbulent boundary layer and kick up it up onto the front wheels to reduce drag. 🧵2/5
The streamlined chassis guides airflow under the spoiler to create down force. The halo - one of the best additions to the sport in terms of safety - is rather aerodynamic.
Each frame of the video is 120GB, 144TB for 1201 frames. @FluidX3D renders the data directly in VRAM. 🧵3/5
Each frame of the video is 120GB, 144TB for 1201 frames. @FluidX3D renders the data directly in VRAM. 🧵3/5
I have benchmarked the new @AMDInstinct MI250 #GPU at @fzj_jsc, and it is disappointing but also impressive. Let me explain.
🧵1/6
cc @HPC_Guru @hpcprogrammer @IanCutress @ProfMatsuoka @sunitachandra29 @aschilling @wkmyrhang @VideoCardz @AMDGPU_
#HPC #Top500 #MI200 #Exascale
🧵1/6
cc @HPC_Guru @hpcprogrammer @IanCutress @ProfMatsuoka @sunitachandra29 @aschilling @wkmyrhang @VideoCardz @AMDGPU_
#HPC #Top500 #MI200 #Exascale
The #MI250 is misleadingly marketed as "one chiplet GPU" with 13312C, 90TFLOPs & 128GB @ 3.2TB/s.
But it is not. The 2 GCDs are 2 separate GPUs with 64GB each, like a K80 dual-GPU but in a socket. One #GPU can't directly access the other's memory.
🧵2/6
amd.com/en/products/se…
But it is not. The 2 GCDs are 2 separate GPUs with 64GB each, like a K80 dual-GPU but in a socket. One #GPU can't directly access the other's memory.
🧵2/6
amd.com/en/products/se…
To use both GCDs, the software needs to be multi-GPU capable. For many algorithms this is very difficult and for some it is entirely infeasible. The desire for large unified memory is huge.
The #MI250 promises exactly that with "128GB", but delivers only half.
🧵3/6
The #MI250 promises exactly that with "128GB", but delivers only half.
🧵3/6
