Pidoku
7The NVIDIA Hardware Line

The NVIDIA Hardware Line

Module 77 lessons~5h 25m

Lessons, in order

7.a The NVIDIA Line-upNVIDIA sells one architecture each generation in many shapes: a flagship data-center GPU, PCIe cards for smaller servers, workstation and gaming cards, a desktop "AI computer", a robotics … 35 min 7.b Hopper: H100 and H200The H100 is the best-documented AI chip ever made and still the largest installed base, so it is the one to know by heart. It is one large die (GH100) in the middle of a package, surrounded … 50 min 7.c Blackwell: B200, B300 and the SuperchipHopper's die was already as large as a lithography machine can print. To grow, Blackwell put two reticle-sized dies in one package and joined them with a link so fast (10 TB/s) that software … 55 min 7.d The NVL72 RackOn an 8-GPU server, GPUs talk over NVLink at 900 GB/s each way; the moment a job needs a ninth GPU, traffic drops to a network card at 50-100 GB/s. That cliff shaped how every large model … 50 min 7.e Rubin and the RoadmapRubin is the first NVIDIA GPU built around HBM4, and its headline change is the one LLM serving cares about most: about 22 TB/s of memory bandwidth per GPU, 2.75x Blackwell. It also doubles … 45 min 7.f Networking: the Other HalfA large GPU cluster has four separate networks, each built for a different job: 50 min 7.g Workstations, PCs and the EdgeNot every AI workload belongs in a data center. Developers fine-tune on a desk, companies keep private data on premises, and robots must think without a network. NVIDIA sells three kinds of … 40 min

About this module

Module 2 showed what is inside a GPU. This module walks through the real NVIDIA products, one drawing at a time: the H100 die and its SMs, the two-die Blackwell package, the superchip and compute tray, the 72-GPU rack, Rubin, the networks that join racks together, and the parts that live on a desk or inside a robot.

Every drawing is an engineering diagram built from the vendor’s published layout: numbered pins point at the parts and a key explains each one. Press Expand on any of them to pan and zoom.

#LessonThe question it answers
7.aThe NVIDIA Line-upWhat does NVIDIA sell, and how do I decode the names?
7.bHopper: H100 and H200What is physically on an H100 package, down to one SM?
7.cBlackwell: B200, B300 and the SuperchipHow do two dies act as one GPU, and what are FP4 and the superchip for?
7.dThe NVL72 RackWhy is the rack, not the GPU, now the unit of computing?
7.eRubin and the RoadmapWhat changes with Rubin, and how do I read announcements still to come?
7.fNetworking: the Other HalfWhat are the four networks of a GPU cluster, and which one limits me?
7.gWorkstations, PCs and the EdgeWhat can I run on a card, a desktop box or a robot module?

Product facts were checked on 16 October 2026. Rubin had been shipping since August 2026; Blackwell and Blackwell Ultra were most of the new capacity; Hopper was still the largest installed base. Figures are vendor numbers unless a lesson says otherwise.

esc

↑↓ navigate↵ openesc close

100% drag to pan · scroll or pinch to zoom · double-click to zoom in