AMD ROCm 7.0.2 just overhauled Radeon & Ryzen compatibility. Adds native support for: Strix Halo RX 9000 RX 7000 in the following frameworks: Pytorch and Tensorflow LM Studio Megatron-LM CumfyUI Triton vLLM JAX Llama.cpp FlashAttention-2 ONNX Runtime

https://t.co/TWFVhyCyLh

AMD datacenter revenue up +22% (not including the OpenAI deal)

AMD gaming sales soar! 73% gain y/y Radeon 9000 the MVP of GPUs

AMD ROCm 7.1 Released: Many Instinct MI350 Series Improvements, Better Performance The new @AMD @AIatAMD ROCm 7.1 brings more MI350X/MI355X enhancements, various perf optimizations, HIP additions, and much more. Nice heavy update. https://t.co/cMpePLZnZ0

https://t.co/jrF4yGQ14e

John Carmack says NVIDIA DGX Spark runs at half of the rated power and delivers half the quoted performance - https://t.co/EkFpcBB3yN

DGX Spark: the $4,000 ai fail - Runs at 60 Tflops in most AI inference frameworks (not 1Pflops) - Terrible scaling performance in a clustered config. - Overheating and restarting under sustained workloads. (Sparks out!) - limited to half the rated power.

Credit: Level1Techs https://t.co/uA4Ck8Ds1y

AMD W9700 Pro supremacy AMD GPUs continue to have superior video processing acceleration and now at less than half the price of 5090.

JUST IN: 🇺🇸 US Department of Energy announces $1 billion partnership with $AMD to build two supercomputers.

Talking about whatever the fuck this is:

Terrible scaling on DGX Spark for $8,000.

🚨 BREAKING: $IBM says key quantum computing algorithm can run on conventional $AMD chips IBM announced it has developed an algorithm to run alongside quantum chips that can address such errors on a field-programmable gate array (FPGA) manufactured by $AMD The Xilinx

Desktop AI compute clustering at half the price of DGX Spark 2x Minisforum MS-S1(Strix Halo) cluster LLM inference 256GB memory 128MB cache 2x 120GbE over USB4NET Qwen3-235B Q4 - 10.87tkn/s

ROG Xbox Ally vs Steamdeck - 2x recharge speed - 2x screen res & refresh rate - +50% battery life (20W) - +10-25% perf ( 20W GPU limited) - +16% memory bandwidth - 3mnths gamepass free - Many more compatible games

AI predictions are blind guesses. They don't know what they are talking about.

How Nvidia rig their AI revenue faking loop.

https://t.co/GfjC9VUK4V

Nvidia's biggest con yet: a giant self-dealing corporate revenue loop fraud. Nvidia is recycling the same money back and forth and reporting it as revenue to inflate stock valuation. Details below ⬇️