DGX Station
97 articles
-
How to Monitor NVIDIA DGX Station GB300 With DCGM, Prometheus and Grafana
Monitor a DGX Station GB300: DCGM-Exporter on port 9400, Prometheus, Grafana dashboard 12239, key GB300 metrics and the metric renames to plan for.
-
Your GPU Has 20 Petaflops of Power. So Why Is Your AI Model Still Slow?
Why a 20-petaflop GB300 can still run slow: memory bandwidth, CPU-GPU links, storage, software and network bottlenecks, with figures to diagnose each.
-
Inside NVIDIA GB300: Why 748GB of Coherent Memory Changes the AI Workstation Game
How GB300's 748GB coherent memory and NVLink-C2C change daily AI work vs older workstations, with HP ZGX Fury G1n and MSI WS300 notes.
-
HP ZGX Fury vs MSI DGX Station: Same GB300 Superchip, Different AI Workstation Experience?
HP ZGX Fury G1n vs MSI XpertStation WS300: storage, RTX PRO cards, ConnectX-8 networking, rack fit, software and warranty compared for Singapore buyers.
-
A S$150,000 AI Workstation: Are You Paying for the GPU or the Technology Connecting Everything?
What a GB300 workstation's price buys: HBM3e, NVLink-C2C, ConnectX-8, cooling and support, and when the connecting technology is worth paying for.
-
NVIDIA GB300 vs Traditional RTX Workstations: Is NVLink the Missing Ingredient?
GB300 coherent memory vs 2-4 RTX PRO 6000 cards over PCIe: memory per GPU, links, measured throughput, prices and where each wins for large AI models.
-
RTX PRO 6000 vs GB300: Is More Memory Always Better for AI?
Why memory bandwidth and compute matter as much as VRAM: token-speed maths for RTX PRO 6000, RTX 5090, DGX Spark and GB300, with published results.
-
Why Buy a DGX Station When You Can Build an RTX PRO 6000 Workstation?
Build an RTX PRO 6000 workstation or buy a DGX Station GB300? Illustrative costs, throughput per dollar and the signs a GB300 becomes necessary.
-
RTX PRO 6000 Blackwell vs NVIDIA DGX Station GB300: Which AI Monster Deserves Your Money?
RTX PRO 6000 Blackwell vs DGX Station GB300: 96GB vs 748GB, bandwidth, measured throughput, software compatibility, use cases and Singapore pricing.
-
Why NVIDIA's Most Powerful AI Systems Don't Rely on PCIe Alone
NVLink 5, NVLink-C2C and NVSwitch vs PCIe: bandwidth, coherence, 72-GPU switching and power. Where PCIe still fits in GB300 and DGX Station.
-
From DGX Station to GB300 NVL72: How NVIDIA Scales AI From Your Desk to the Data Centre
How to build and debug AI on a DGX Station GB300 and scale to GB300 NVL72: what carries over, what needs testing in the rack and a step-by-step...
-
Can NVIDIA GB300 Replace Your Company's Multi-GPU AI Server?
An 8-step checklist for replacing a 4-8 GPU RTX PRO 6000 or HGX H100 server with GB300: model size, concurrency, parallelism, Arm, uptime, power.
-
From RTX PRO 6000 to GB300 NVL72: When Does Your AI Workstation Need to Become an AI Factory?
The scaling ladder from one RTX PRO 6000 to four Max-Q cards, DGX Station GB300, RTX PRO Server, HGX B300 and GB300 NVL72, with signs it is time...
-
Can RTX PRO 6000 Blackwell Be the Perfect Companion to NVIDIA DGX Station?
Pair an RTX PRO 6000 with a DGX Station GB300: physical AI, digital twins, rendering and CFD workflows, power caps, and inside-the-box vs separate workstation.
-
Two RTX PRO 6000 GPUs vs One DGX Station: Which Makes More Sense for Local AI?
Two 96GB RTX PRO 6000 cards vs one GB300: which models fit, splitting over PCIe, memory bandwidth, measured throughput, cost and who should buy which.
-
RTX 5090 vs DGX Station GB300: Can a Gaming GPU Compete With a Six-Figure AI Workstation?
RTX 5090 vs NVIDIA DGX Station GB300: AI throughput, 32GB vs 748GB memory, concurrency, MIG, management and enterprise requirements.
-
From 10,000 Documents to One Private AI Brain: What Can DGX Station Really Deliver?
What a DGX Station GB300 delivers for private AI search across 10,000 documents: corpus sizing, RAG pipeline, users served, permissions and pitfalls.
-
Can NVIDIA DGX Station Become the Foundation of Your Company's Private AI Cloud?
Turn a DGX Station GB300 into a private AI cloud: OpenAI-compatible API, gateway budgets, SSO, audit logs, MIG quotas and the limits of a single node.
