CAFE · Communications · ½ × ½ slot
Rendered from the layout at 2.5 µm per pixel. Layers: NwellDiffusionPolysiliconMetal 1Metal 2Metal 3Metal 4Metal 5Top metalPad openings
| Inside the core logic cells per mm² | 5k
(7% of buffer maximum) over the 1.73 mm² core: what the design itself packed |
|---|---|
| Inside the seal ring logic cells per mm² | 2k
over all 4.68 mm² of the slot, pad ring included |
| Densest mm² logic cells per mm² | 8k (10% of buffer maximum) |
| Core covered by logic | 15% |
| Transistor density transistors per mm² | 252k average, 248k in the mm² with the most logic cells |
Every figure here is measured from the public layout, the same way for every design on the run: see data and method, and how its designs used the process.
This chip is an open source Ethernet focused ASIC featuring both an 4 port, 100Mbps capable cut-through, unmanaged, Ethernet switch and an Ethernet connected beacon, broadcasting an ethernet frame every second with an uptime count until the heat death of the universe.
Both designs have been validated on an FPGA.
This is the default 0.5x0.5 slot size for this project and is also the version that has been the most worked on.
| Die | 1.94 × 2.53 mm (4.9 mm²) |
|---|---|
| Core area | 1.73 mm² inside the pad ring
as set in the design's own layout settings |
| Inside the seal ring | 4.68 mm² (the die less its 26 µm seal ring) |
| Placements | 1 on the reticle, in slot G2 |
| Bond pads | 56 |
| Transistors | 436,088 |
| Logic cells | 8,747 + 48,803 filler and tap cells |
| Libraries | 7t-5V (9k) |
| SRAM | No SRAM |
| Licence | CERN-OHL-S-2.0 declared
Repository: CERN-OHL-S-2.0 in LICENSE — agrees |
|---|---|
| Template | wafer-space/gf180mcu-project-template 1.5.7 |
| How it was checked | Commit and PDK pinThe licence and template were read from the repository at commit 80a20dedd124.The template release named above is the one whose Makefile pins the same PDK as that commit (PDK_COMMIT d658698bd8bc). |
| Chip-on-board | Selected at submission |
| Precheck | Passed
the check it was accepted on, 2026-07-15, gf180mcu-precheck 1.7.2 (from its image)
Passed latest re-run, 2026-08-02, gf180mcu-precheck 1.7.3 step results and logs |
Which parts of the GF180MCU process this design uses: its transistor voltages and where they come from, its gate oxides, and any devices drawn by hand outside the standard cells, SRAM and pads.
| 5 V | 7t-5V standard cells |
|---|---|
| Gate oxides | Thick only (5 V/6 V) in the design itself |
| I/O library | gf180mcu_fd_io (424 cells) |
How this design's area splits between logic cells, filler and tap cells, other and pad ring.
The ten standard cells this design uses most.
| Cell | Library | Instances | What it does |
|---|---|---|---|
| clkbuf_1 | 7t-5V | 1,977 | Clock buffer |
| dffq_1 | 7t-5V | 1,831 | D flip-flop (1 bit storage) |
| nand2_1 | 7t-5V | 477 | 2-input NAND gate |
| nor2_1 | 7t-5V | 461 | 2-input NOR gate |
| oai21_1 | 7t-5V | 412 | OR-AND-Invert compound gate |
| clkinv_1 | 7t-5V | 391 | Clock inverter |
| aoi22_1 | 7t-5V | 383 | AND-OR-Invert (2x2 inputs) |
| aoi21_1 | 7t-5V | 267 | AND-OR-Invert compound gate |
| and2_1 | 7t-5V | 247 | 2-input AND gate |
| clkbuf_8 | 7t-5V | 206 | Clock buffer (8x drive) |
Logic cells per mm² in each square millimetre of the die. Hatched outlines are SRAM macros; the red outline is the densest mm² that avoids them (8k logic cells per mm²), found by sliding a 1 mm square over the die in 0.1 mm steps, so it can straddle the fixed squares and find more than any one of them. Squares at the edges are partial, so they read low.
Expresso Ethernet switch has been made on 3 shuttles since Jun 2026. Every version, oldest first, with this one in bold:
Tiny Tapeout tile Google shuttle wafer.space die
Tiny Tapeout puts many small designs on one chip, so trying out a block costs a fraction of a whole die: a cheap way to test one piece of a bigger design before it goes on a die of its own.
Before a design is made, the wafer.space platform checks it against the foundry's design rules. CAFE was accepted on this check, and has been checked again since with newer rules.
All 17 steps passed, with 0 KLayout DRC errors, on 2026-07-15.
| Step | Accepted on check 576, 2026-07-15, initial, gf180mcu-precheck 1.7.2 (from its image) PDK open_pdks 1.0.604, LibreLane 5fe3db1 | Latest re-run check 754, 2026-08-02, drc_update, gf180mcu-precheck 1.7.3 PDK open_pdks 1.0.604, LibreLane 5fe3db1 |
|---|---|---|
| KLayout.ReadLayout | passed | passed |
| KLayout.CheckTopLevel | passed | passed |
| KLayout.CheckSize | passed | passed |
| KLayout.CheckPadMask | passed | passed |
| KLayout.GenerateID | passed | passed |
| KLayout.Render | passed | passed |
| KLayout.Density | passed | passed |
| Checker.KLayoutDensity | passed | passed |
| KLayout.ZeroAreaPolygons | passed | passed |
| Checker.KLayoutZeroAreaPolygons | passed | passed |
| KLayout.Antenna | passed | passed |
| Checker.KLayoutAntenna | passed | passed |
| Magic.DRC | passed | passed |
| Checker.MagicDRC | passed | passed |
| KLayout.DRC | passed | passed |
| Checker.KLayoutDRC | passed | passed |
| KLayout.WriteLayout | passed | passed |
| KLayout DRC errors | 0 | 0 |
| Magic DRC errors (advisory) | 0 | 0 |
| Files | steps · log (gzip) | steps · log (gzip) |
The steps of the gf180mcu-precheck LibreLane flow, as run by platform.wafer.space. The design was accepted on the newest finished check before it was submitted; later re-runs check the same file with updated rules. KLayout DRC is the sign-off check. Magic DRC is advisory: its count does not stop a design. In the logs, Magic's "Error while reading cell" lines about fill cells are messages from its reader, not check failures.