Local AI
169 articles
-
748GB Coherent Memory Sounds Incredible. But What Does Coherent Actually Mean?
What coherent memory means on the DGX Station GB300: NVLink-C2C, one address space, real bandwidth limits and what changes for AI teams.
Wei Fei
·
-
RTX Spark vs DGX Spark: Which One Runs Your Favourite AI Models Better?
Published RTX Spark and DGX Spark LLM benchmarks compared: GPT-OSS, Qwen and Llama tokens per second, prompt speed, concurrency and how to test fairly.
Wei Fei
·
-
NVIDIA RTX Spark vs GB10: Is Windows Finally Ready to Challenge Linux for Local AI?
RTX Spark brings CUDA and PyTorch to Windows on Arm. Which AI frameworks run natively, where Linux GB10 systems still lead, and how teams can use both.
Wei Fei
·
-
NVIDIA DGX Spark vs RTX Spark: Same AI Ambitions, Two Very Different Computers
NVIDIA DGX Spark vs RTX Spark compared: Linux vs Windows, 128 GB unified memory, model support, software and who each suits, with Singapore prices.
Wei Fei
·
-
Two NVIDIA GB10 Systems, One AI Model: Does Performance Actually Double?
Does linking two NVIDIA GB10 systems double AI performance? Published dual-unit tests, tensor vs pipeline parallelism, and when separate units serve more users.
Wei Fei
·
-
Can a GB10 Fine-Tune Your Own AI Model, or Is It Mostly for Inference?
Can NVIDIA GB10 fine-tune AI models? NVIDIA's published LoRA and QLoRA results, job-time estimates, supported tools, memory limits and when RAG is better.
Wei Fei
·
-
Can One GB10 Support 20 Employees Using AI at the Same Time?
How many employees can share one NVIDIA GB10 system? Published concurrency benchmarks, per-user speeds, what slows it down and when to add a second unit.
Wei Fei
·
-
Can a GB10 Transcribe 100 Hours of Meetings? Let's Put It to the Test.
How long would an NVIDIA GB10 take to transcribe 100 hours of meetings? Whisper and Parakeet speeds, multilingual tips, storage and running your own test.
Wei Fei
·
-
Is NVIDIA GB10 Powerful Enough to Build Your Own Corporate ChatGPT?
What a private corporate ChatGPT needs beyond NVIDIA GB10 hardware: chat UI, sign-in, permission-aware RAG, guardrails, logging, PDPA and an SME rollout plan.
Wei Fei
·
-
Can NVIDIA GB10 Analyse Thousands of Company Documents Without Sending Them to the Cloud?
How an NVIDIA GB10 system indexes and answers questions on thousands of documents locally: RAG pipeline, NVIDIA playbooks, indexing time, privacy and PDPA.
Wei Fei
·
-
NVIDIA GB10 and DeepSeek: How Fast Can It Really Think?
Which DeepSeek models run on NVIDIA GB10 and how fast: R1 distills 14B to 70B, why reasoning models feel slow, and DeepSeek-V4-Flash on GB10 clusters.
Wei Fei
·
-
Can a GB10 Run AI Agents 24/7 Without Anyone Watching?
How to run always-on AI agents on an NVIDIA GB10 system: OpenClaw and NemoClaw, sandboxing, approvals, scheduling, logging, UPS and laptop vs desk unit.
Wei Fei
·
-
Can NVIDIA GB10 Run an Entire Company's Private AI Assistant?
Can one NVIDIA GB10 system run a private AI assistant for your company? Company-size guide, memory, permissions, availability and deployment limits explained.
Wei Fei
·
-
RTX Spark vs DGX Spark: What NVIDIA's 120B and 200B AI Model Claims Really Mean
What NVIDIA's 120B (RTX Spark) and 200B (DGX Spark) model claims mean: quantisation, memory, context length, MoE vs dense models and real inference speed.
Wei Fei
·
-
Can RTX Spark Replace Your NVIDIA GB10 AI Supercomputer?
Where a Windows RTX Spark laptop or desktop can replace a Linux GB10 system such as DGX Spark, and where it cannot: sharing, agents, fine-tuning, clustering.
Wei Fei
·
-
NVIDIA GB10 vs RTX PRO 6000 Blackwell: When Does a Professional GPU Make More Sense?
NVIDIA GB10 vs RTX PRO 6000 Blackwell: 128 GB vs 96 GB, bandwidth, published 70B and 120B benchmarks, workstation flexibility and total cost compared.
Wei Fei
·
-
NVIDIA GB10 vs RTX 5090: More Memory or More GPU Power — Which Matters?
NVIDIA GB10 vs GeForce RTX 5090 for local AI: 128 GB vs 32 GB, bandwidth, published LLM benchmarks, power and which to buy, with Singapore prices.
Wei Fei
·
-
NVIDIA GB10 vs Mac Studio M5 Ultra: Who Really Wins at Local AI?
NVIDIA GB10 (DGX Spark, ASUS GX10) vs Apple Mac Studio M5 Ultra and M5 Max for local AI: unified memory, bandwidth, benchmarks, software and Singapore prices.
Wei Fei
·
Wei Fei
