James
@jams3223.bsky.social
Developer & Linux Addict.
Steam Deck Owner.
Hardware is already there, software needs to keep up.
Digital media should be half the price of physical ones.
There's 2 kinds of people in this world. Those who are humble and those who are about to be humbled.
Steam Deck Owner.
Hardware is already there, software needs to keep up.
Digital media should be half the price of physical ones.
There's 2 kinds of people in this world. Those who are humble and those who are about to be humbled.
Sparse 3D Radiance Cascades
www.youtube.com/watch?v=TGLA...
www.youtube.com/watch?v=TGLA...
https://youtu.be/TGLAxW0xVDU?si=ay--SbhDTzm3Ry0q
t.co
November 3, 2025 at 7:45 PM
Sparse 3D Radiance Cascades
www.youtube.com/watch?v=TGLA...
www.youtube.com/watch?v=TGLA...
www.cemyuksel.com
November 3, 2025 at 6:52 PM
Different limitations of global illumination techniques and the SDFGI solution tend to cause another issue (bluriness).
ar5iv.labs.arxiv.org/html/2007.14...
ar5iv.labs.arxiv.org/html/2007.14...
Signed Distance Fields Dynamic Diffuse Global Illumination
Global Illumination (GI) is of utmost importance in the field of photo-realistic rendering. However, its computation has always been very complex, especially diffuse GI. State of the art real-time GI ...
ar5iv.labs.arxiv.org
November 2, 2025 at 7:08 PM
Different limitations of global illumination techniques and the SDFGI solution tend to cause another issue (bluriness).
ar5iv.labs.arxiv.org/html/2007.14...
ar5iv.labs.arxiv.org/html/2007.14...
Most Efficient, Least Precise, and Most Performant to Least Efficient, Most Precise, and Least Performant:
INT4
FP4
INT8
FP8
INT16
FP16
INT4
FP4
INT8
FP8
INT16
FP16
October 22, 2025 at 5:40 PM
Most Efficient, Least Precise, and Most Performant to Least Efficient, Most Precise, and Least Performant:
INT4
FP4
INT8
FP8
INT16
FP16
INT4
FP4
INT8
FP8
INT16
FP16
Reposted by James
Get up to speed with our latest advanced graphics research:
📝 GATE (Geometry-Aware Trained Encoding) for neural rendering.
📝 Real-time procedural GPU tree generation.
📝 Neural Visibility Cache for real-time light sampling.
📥 Download the papers: gpuopen.com/learn/public...
📝 GATE (Geometry-Aware Trained Encoding) for neural rendering.
📝 Real-time procedural GPU tree generation.
📝 Neural Visibility Cache for real-time light sampling.
📥 Download the papers: gpuopen.com/learn/public...
October 16, 2025 at 7:56 PM
Get up to speed with our latest advanced graphics research:
📝 GATE (Geometry-Aware Trained Encoding) for neural rendering.
📝 Real-time procedural GPU tree generation.
📝 Neural Visibility Cache for real-time light sampling.
📥 Download the papers: gpuopen.com/learn/public...
📝 GATE (Geometry-Aware Trained Encoding) for neural rendering.
📝 Real-time procedural GPU tree generation.
📝 Neural Visibility Cache for real-time light sampling.
📥 Download the papers: gpuopen.com/learn/public...
Reposted by James
Exciting news for developers working with our AMD DGF geometry compression technology!
🗜️ The DGF SDK has been updated with a new animation-aware encoding pipeline, bug fixes, and improvements.
👉 Learn more and download: gpuopen.com/learn/animat...
🗜️ The DGF SDK has been updated with a new animation-aware encoding pipeline, bug fixes, and improvements.
👉 Learn more and download: gpuopen.com/learn/animat...
Animating geometry with AMD DGF - AMD GPUOpen
The AMD DGF SDK has been updated with improvements and new features, including the addition of an animation-aware encoding pipeline.
gpuopen.com
September 24, 2025 at 9:08 PM
Exciting news for developers working with our AMD DGF geometry compression technology!
🗜️ The DGF SDK has been updated with a new animation-aware encoding pipeline, bug fixes, and improvements.
👉 Learn more and download: gpuopen.com/learn/animat...
🗜️ The DGF SDK has been updated with a new animation-aware encoding pipeline, bug fixes, and improvements.
👉 Learn more and download: gpuopen.com/learn/animat...
AMD doesn't add tensor cores to its GPU because it has a good memory bandwidth with the Infinity Cache, whereas NVIDIA needs to increase the number of tensor cores to address its poor memory bandwidth.
September 19, 2025 at 9:19 PM
AMD doesn't add tensor cores to its GPU because it has a good memory bandwidth with the Infinity Cache, whereas NVIDIA needs to increase the number of tensor cores to address its poor memory bandwidth.
The AMD WMMA accelerator does not increase inference performance in any form; it's the number of operations that these cores can process per clock that has caused a ~50% performance increase, primarily due to improvements in node and clock speed. (...)
September 19, 2025 at 9:10 PM
The AMD WMMA accelerator does not increase inference performance in any form; it's the number of operations that these cores can process per clock that has caused a ~50% performance increase, primarily due to improvements in node and clock speed. (...)
Reposted by James
in-depth thread: optimizing GDScript performance from ~83ms to ~9ms in a scene our tower defense game Rift Riff, made in Godot, with 1000 enemies. 🧵👇🏼
#gamedev #godotengine
#gamedev #godotengine
July 14, 2025 at 9:17 PM
in-depth thread: optimizing GDScript performance from ~83ms to ~9ms in a scene our tower defense game Rift Riff, made in Godot, with 1000 enemies. 🧵👇🏼
#gamedev #godotengine
#gamedev #godotengine
The only thing AI is good at is being unbiased.
August 3, 2025 at 10:57 PM
The only thing AI is good at is being unbiased.
As it turns out, people don't hate 'authoritarianism'. They resent not being part of the group that would wield that authority. People seek to benefit from 'authoritarianism' at the expense of others.
(....)
(....)
July 26, 2025 at 7:11 PM
As it turns out, people don't hate 'authoritarianism'. They resent not being part of the group that would wield that authority. People seek to benefit from 'authoritarianism' at the expense of others.
(....)
(....)
Top 5 African Resources Being Stolen by Foreigners:
Gold
Uranium
Human
Ants Farms
Meteorites
Gold
Uranium
Human
Ants Farms
Meteorites
July 23, 2025 at 1:00 AM
Top 5 African Resources Being Stolen by Foreigners:
Gold
Uranium
Human
Ants Farms
Meteorites
Gold
Uranium
Human
Ants Farms
Meteorites
Based on my analysis, the Nintendo Switch 2 uses a custom version of DLSS that leverages INT8 instead of FP16.
June 21, 2025 at 5:55 PM
Based on my analysis, the Nintendo Switch 2 uses a custom version of DLSS that leverages INT8 instead of FP16.
Reposted by James
November 24, 2024 at 4:53 PM
GPU Driver Infrastructure:
-> GLSL/HLSL (Shading Language)
--> glslang/DXC (Shader Compiler)
---> SPIR-V/DXIL (Intermediate Language)
----> RADV/AMD Software: Adrenalin Edition (Driver)
-----> NIR (Analyze and validate the generated SPIR-V Intermediate Language to the NIR Intermediate Language (...)
-> GLSL/HLSL (Shading Language)
--> glslang/DXC (Shader Compiler)
---> SPIR-V/DXIL (Intermediate Language)
----> RADV/AMD Software: Adrenalin Edition (Driver)
-----> NIR (Analyze and validate the generated SPIR-V Intermediate Language to the NIR Intermediate Language (...)
November 21, 2024 at 7:09 PM
GPU Driver Infrastructure:
-> GLSL/HLSL (Shading Language)
--> glslang/DXC (Shader Compiler)
---> SPIR-V/DXIL (Intermediate Language)
----> RADV/AMD Software: Adrenalin Edition (Driver)
-----> NIR (Analyze and validate the generated SPIR-V Intermediate Language to the NIR Intermediate Language (...)
-> GLSL/HLSL (Shading Language)
--> glslang/DXC (Shader Compiler)
---> SPIR-V/DXIL (Intermediate Language)
----> RADV/AMD Software: Adrenalin Edition (Driver)
-----> NIR (Analyze and validate the generated SPIR-V Intermediate Language to the NIR Intermediate Language (...)
This simple gif shows the effect of the Mercator projection on the real size of continents and countries on a planisphere.
November 13, 2024 at 4:24 PM
This simple gif shows the effect of the Mercator projection on the real size of continents and countries on a planisphere.
Reposted by James
I'm excited to announce that Spark⚡ 1.0 is finally out!
ludicon.com/spark/
Spark is the culmination of over a year of dedicated research and development, built upon a solid foundation of years of experience in GPU compression technology.
What sets Spark apart?
ludicon.com/spark/
Spark is the culmination of over a year of dedicated research and development, built upon a solid foundation of years of experience in GPU compression technology.
What sets Spark apart?
February 6, 2024 at 12:34 PM
I'm excited to announce that Spark⚡ 1.0 is finally out!
ludicon.com/spark/
Spark is the culmination of over a year of dedicated research and development, built upon a solid foundation of years of experience in GPU compression technology.
What sets Spark apart?
ludicon.com/spark/
Spark is the culmination of over a year of dedicated research and development, built upon a solid foundation of years of experience in GPU compression technology.
What sets Spark apart?
These are some great optimization tips.
www.youtube.com/watch?v=40Jz...
www.youtube.com/watch?v=40Jz...
I Optimised My Game Engine Up To 12000 FPS
YouTube video by Vercidium
www.youtube.com
November 13, 2024 at 4:17 PM
These are some great optimization tips.
www.youtube.com/watch?v=40Jz...
www.youtube.com/watch?v=40Jz...
I examined different GPU architectures to assess their AI capabilities. I discovered that NVIDIA allocates a larger portion of the die space to the tensor cores. The tensor cores are primarily responsible for executing FP16, INT8, and INT4 instructions, rather than the CUDA cores (...)
November 13, 2024 at 4:12 PM
I examined different GPU architectures to assess their AI capabilities. I discovered that NVIDIA allocates a larger portion of the die space to the tensor cores. The tensor cores are primarily responsible for executing FP16, INT8, and INT4 instructions, rather than the CUDA cores (...)
Unreal Engine 5 Rendering Progression
www.reddit.com/r/UnrealEngi...
www.reddit.com/r/UnrealEngi...
Reddit - Dive into anything
Unreal Engine 5, by Epic Games.
www.reddit.com
November 13, 2024 at 4:11 PM
Unreal Engine 5 Rendering Progression
www.reddit.com/r/UnrealEngi...
www.reddit.com/r/UnrealEngi...
Some confusion arises from the fact that if the AMD RDNA 2 architecture has twice the number of ray boxes compared to the NVIDIA 30 series, it should be faster, while NVIDIA claims to have twice the ray triangle count. (...)
November 13, 2024 at 4:03 PM
Some confusion arises from the fact that if the AMD RDNA 2 architecture has twice the number of ray boxes compared to the NVIDIA 30 series, it should be faster, while NVIDIA claims to have twice the ray triangle count. (...)
Here's a great paper on what benefits we can reap with quantized inference for more efficient AI on mobile.
ieeexplore.ieee.org/stamp/stamp....
ieeexplore.ieee.org/stamp/stamp....
IEEE Xplore Full-Text PDF:
ieeexplore.ieee.org
November 13, 2024 at 3:50 PM
Here's a great paper on what benefits we can reap with quantized inference for more efficient AI on mobile.
ieeexplore.ieee.org/stamp/stamp....
ieeexplore.ieee.org/stamp/stamp....
Revolutionary 3D Rendering Techs:
Work Graph
Neural Texture Block Compression
Visibility Buffer Rendering with Software VRS
Radiance Cascades with Hierarchical Digital Differential Analyzer
NeRF (AR/VR)
Work Graph
Neural Texture Block Compression
Visibility Buffer Rendering with Software VRS
Radiance Cascades with Hierarchical Digital Differential Analyzer
NeRF (AR/VR)
November 13, 2024 at 3:46 PM
Revolutionary 3D Rendering Techs:
Work Graph
Neural Texture Block Compression
Visibility Buffer Rendering with Software VRS
Radiance Cascades with Hierarchical Digital Differential Analyzer
NeRF (AR/VR)
Work Graph
Neural Texture Block Compression
Visibility Buffer Rendering with Software VRS
Radiance Cascades with Hierarchical Digital Differential Analyzer
NeRF (AR/VR)
Fast Matrix Multiplication on CPU
siboehm.com/articles/22/...
siboehm.com/articles/22/...
Fast Multidimensional Matrix Multiplication on CPU from Scratch
Numpy can multiply two 1024x1024 matrices on a 4-core Intel CPU in ~8ms.This is incredibly fast, considering this boils down to 18 FLOPs / core / cycle, with...
siboehm.com
November 13, 2024 at 3:40 PM
Fast Matrix Multiplication on CPU
siboehm.com/articles/22/...
siboehm.com/articles/22/...