Home › Blog › AI & Workstations
AI & Workstations
Research on DGX Spark, DGX Station, ASUS Ascent GX10, local LLM deployment and AI PCs: capability, cost and fit.
327 articles
-
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.
Wei Fei
·
-
NVIDIA GB300 NVL72 vs 72 Individual GPUs: Why Connecting Them Changes Everything
NVL72 NVLink at 900GB/s per direction vs 8-GPU servers on 800G networks and PCIe GPUs: bandwidth, all-reduce time, power, reliability and when each fits.
Wei Fei
·
-
800Gbps InfiniBand vs 130TB/s NVLink: Why These Numbers Don't Tell the Whole Story
Convert 800Gb/s and 130TB/s to GB/s per GPU per direction: bits vs bytes, aggregate vs per GPU, bidirectional, bisection, NCCL busbw and latency.
Wei Fei
·
-
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.
Wei Fei
·
-
NVLink vs InfiniBand vs PCIe: Is Your AI Supercomputer Only as Fast as Its Weakest Connection?
NVLink 900GB/s, InfiniBand 100GB/s and PCIe Gen6 121GB/s per direction compared. When the slowest link really limits AI training and inference.
Wei Fei
·
-
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.
Wei Fei
·
-
96GB VRAM Changes Everything. But Who Actually Needs That Much?
Which AI models, KV caches, 3D scenes and CFD meshes need 96GB of GPU memory? A sizing guide for the NVIDIA RTX PRO 6000 Blackwell.
Wei Fei
·
-
Four RTX PRO 6000 GPUs, 384GB VRAM: Is This the Ultimate AI Workstation?
Four RTX PRO 6000 Max-Q cards, 384GB: why Max-Q, power and cooling, why VRAM doesn't pool, which models fit, and how it compares with a DGX Station.
Wei Fei
·
-
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.
Wei Fei
·
-
RTX PRO 6000 Blackwell vs RTX 5090: Are You Paying Thousands More Just for VRAM?
RTX PRO 6000 Blackwell vs RTX 5090: published benchmarks, 96GB vs 32GB VRAM, ECC, drivers, MIG, licensing and price per GB compared.
Wei Fei
·
-
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.
Wei Fei
·
-
RTX PRO 6000 Blackwell Uses PCIe 5.0. So Why Does NVIDIA GB300 Need NVLink?
Why the RTX PRO 6000 is fine on PCIe 5.0, when PCIe becomes the bottleneck, and why GB300 uses NVLink-C2C and an NVL72 NVLink fabric. Figures compared.
Wei Fei
·
-
NVIDIA RTX PRO 6000 Blackwell: A S$20,000 Graphics Card — Brilliant or Completely Overkill?
Is NVIDIA's 96GB RTX PRO 6000 Blackwell worth it? What you pay for, when it beats an RTX 5090, when it's overkill and Singapore pricing.
Wei Fei
·
-
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.
Wei Fei
·
-
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...
Wei Fei
·
-
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.
Wei Fei
·
-
Could Your S$1 Million AI Cluster Be Wasting Money Because of Poor Networking?
An illustrative S$1M AI cluster budget: what it buys, how network stalls waste GPU time, what idle time costs and the networking mistakes to avoid.
Wei Fei
·
-
Can RTX 5090 Run Your Own AI Video Generation Studio?
Can an RTX 5090 run a local AI video studio? Wan 2.2, HunyuanVideo and LTX-2 VRAM needs, measured render times and ComfyUI workflows.
Wei Fei
·
Wei Fei
