From Sand to Superintelligence · Drill cards · Chapter 06
Drills
Designing the Impossible
10 atomic recall cards. Export to Anki and let spaced repetition do its slow work.
In Anki: File → Import, choose this TSV, set field separator to Tab, deck = Sand to Silicon · Ch 06, note type = Basic.
| Front | Back |
|---|---|
| What process node is Rubin built on? | TSMC N3P — the high-performance refinement of N3, a FinFET node. |
| How many transistors does the Rubin package contain? | About 336 billion, across two dies on a single substrate. |
| What are the three dominant EDA companies? | Synopsys, Cadence, and Siemens EDA. |
| What language do engineers write to describe chip behavior at the RTL level? | Hardware description languages like SystemVerilog. |
| What does 'timing closure' mean in chip design? | Every signal must reach every destination within a single clock period, with margin for manufacturing variation. |
| How many lithography mask layers does a leading-edge chip require? | Perhaps eighty individual mask layers — one per lithography step. |
| What does a mask-making vendor pattern masks on? | Chrome on glass plates, using electron-beam lithography. |
| What is the key structural difference between a FinFET and a GAA nanosheet transistor? | A FinFET's gate wraps three sides of a vertical fin; a GAA gate wraps all four sides of each horizontal nanosheet. |
| Which TSMC node uses GAA nanosheet transistors? | N2 — the successor to N3P, not yet used by Rubin. |
| Why is the Rubin GPU die area close to the 'reticle limit'? | The reticle limit is the largest area a single EUV exposure can pattern at once; at roughly 800 mm², Rubin's die pushes that boundary. |