https://flozi.net/en Deep dives into servers, GPUs, AI infrastructure and modern IT systems. en https://flozi.net/en/tools/llm-inference-calculator https://flozi.net/en/tools/llm-inference-calculator Tue, 22 Sep 2026 00:00:00 GMT Interactive planning tools for LLM serving: VRAM capacity, decode throughput, prefill latency, and the batch size where decode stops being bandwidth-limited — plus the maximum sequence length at that operating point. ai gpu inference vllm deep-learning tools calculator https://flozi.net/en/tools/llm-inference-economics https://flozi.net/en/tools/llm-inference-economics Tue, 22 Sep 2026 00:00:00 GMT From physics to euros: memory and compute limits of LLM inference, which models fit which hardware, token yield per GPU, own hardware vs cloud break-even, and open-weight models vs proprietary API token costs. ai gpu inference economics cloud deep-learning tools calculator https://flozi.net/en/guides/ai/asr/why-parakeet-fastconformer-tdt-beats-encoder-decoder https://flozi.net/en/guides/ai/asr/why-parakeet-fastconformer-tdt-beats-encoder-decoder Thu, 08 Jan 2026 00:00:00 GMT A technical breakdown of why Parakeet’s FastConformer + Token-and-Duration Transducer design is a strong fit for production ASR: streaming latency, throughput, long-form audio, and fast domain adaptation via external N-gram LMs. ai guide asr speech datacenter machine-learning https://flozi.net/en/my-journey https://flozi.net/en/my-journey Tue, 25 Nov 2025 00:00:00 GMT A personal story about autism, entrepreneurship, mental health, and how AI became my anchor. about career ai autism personal https://flozi.net/en/guides/ai/scaling/expert_parallel https://flozi.net/en/guides/ai/scaling/expert_parallel Tue, 25 Nov 2025 00:00:00 GMT A deep dive into the hardware mechanics of Expert Parallelism (EP), explaining how Mixture-of-Experts (MoE) models route tokens across GPUs and the challenges of All-to-All communication. ai gpu hardware guide scaling deep-learning moe https://flozi.net/en/guides/ai/scaling/tensor_parallel https://flozi.net/en/guides/ai/scaling/tensor_parallel Tue, 25 Nov 2025 00:00:00 GMT A deep dive into the hardware mechanics of Tensor Parallelism, explaining how matrix multiplication is split across GPUs and why communication bandwidth is the critical bottleneck. ai gpu hardware guide scaling deep-learning https://flozi.net/en/guides/ai/scaling/pipeline_parallel https://flozi.net/en/guides/ai/scaling/pipeline_parallel Tue, 25 Nov 2025 00:00:00 GMT A deep dive into the hardware mechanics of Pipeline Parallelism, explaining how layers are partitioned across GPUs and how scheduling algorithms like 1F1B manage the pipeline bubble. ai gpu hardware guide scaling deep-learning https://flozi.net/en/guides/ai/gpu-model-efficiency-playbook https://flozi.net/en/guides/ai/gpu-model-efficiency-playbook Mon, 17 Nov 2025 00:00:00 GMT A practical overview that matches NVIDIA RTX PRO 6000, H200, and DGX Station platforms to Qwen3 model sizes using the ops:byte method from the LLM inference math guide. ai datacenter gpu deep-learning hardware guide https://flozi.net/en/guides/ai/llm-vram-calculator-explained https://flozi.net/en/guides/ai/llm-vram-calculator-explained Fri, 31 Oct 2025 00:00:00 GMT Comprehensive guide to calculating GPU memory requirements for Large Language Model inference and training with practical examples using Qwen3-VL-32B. ai machine-learning gpu datacenter deep-learning https://flozi.net/en/guides/ai/llm-inference-math https://flozi.net/en/guides/ai/llm-inference-math Fri, 31 Oct 2025 00:00:00 GMT Essential mathematical formulas and calculations for profiling LLM inference performance, determining compute vs memory bottlenecks, and selecting optimal GPU hardware. ai machine-learning gpu datacenter deep-learning hardware https://flozi.net/en/about https://flozi.net/en/about Thu, 30 Oct 2025 00:00:00 GMT Learn about flozi00 - passionate about speech recognition, accessibility, and AI, with expertise in server infrastructure and systems. about cv ai datacenter https://flozi.net/en/guides/linux/solving-nvidia-driver-issues-on-ubuntu-24-04-with-nokaslr https://flozi.net/en/guides/linux/solving-nvidia-driver-issues-on-ubuntu-24-04-with-nokaslr Wed, 29 Oct 2025 00:00:00 GMT A guide to fixing NVIDIA driver problems on Ubuntu 24.04 by using the 'nokaslr' kernel parameter to resolve conflicts with Kernel Address Space Layout Randomization (KASLR). linux systems nvidia drivers ubuntu guide https://flozi.net/en/guides/linux/installing-nvidia-drivers-docker-gpu-support https://flozi.net/en/guides/linux/installing-nvidia-drivers-docker-gpu-support Wed, 29 Oct 2025 00:00:00 GMT A beginner-friendly guide to installing NVIDIA drivers, Docker, and the NVIDIA Container Toolkit for GPU-accelerated containers on Ubuntu. linux ubuntu nvidia docker gpu guide https://flozi.net/en/ai/models/qwen/qwen3-next https://flozi.net/en/ai/models/qwen/qwen3-next Wed, 29 Oct 2025 00:00:00 GMT A technical breakdown of the Qwen3-Next-80B-A3B-Instruct model, exploring its Hybrid MoE architecture, FP8 precision, and how it achieves efficiency with 80B total parameters but only 3B active. ai machine-learning llm qwen moe fp8 nvidia alibaba https://flozi.net/en/hardware/nvidia/benchmarks/b200-vs-gb200-efficiency-comparison https://flozi.net/en/hardware/nvidia/benchmarks/b200-vs-gb200-efficiency-comparison Tue, 28 Oct 2025 00:00:00 GMT Detailed performance analysis comparing NVIDIA Blackwell B200 and GB200 accelerators across various system configurations, revealing efficiency differences of up to 42% nvidia blackwell b200 gb200 benchmark performance efficiency gpu accelerator https://flozi.net/en/hardware/nvidia/communication/nvidia-nvlink https://flozi.net/en/hardware/nvidia/communication/nvidia-nvlink Tue, 28 Oct 2025 00:00:00 GMT Comprehensive guide to NVIDIA NVLink technology for multi-GPU systems, including configuration, testing, and optimization for AI and HPC workloads nvidia nvlink gpu interconnect hpc hardware