mpw-004 / VIMC Via 1 top, RRAM tier

Downloading the full-resolution image

The stitched reticle is 53,432×65,056 pixels — 3.48 gigapixels. The file itself is small, but decoding it needs about 3.5 GB of memory however small the file is, and most viewers refuse it outright. So it is also published as 42 tiles of about 7576×10776 px (4.4 MB in total, largest 137.1 kB), cut in the gaps between dies so that every tile holds whole projects and none is split.

To put the full-resolution PNG back together from them, byte for byte, with nothing installed but Python itself:

python3 reconstruct_reticle.py

Run it in a directory holding the tiles and mpw-004_VIMC_tiles.json and it finds them itself, checks every tile against its checksum as it reads it, and checks the finished file against the manifest. It works because the tiles and the full image are written with PNG's Up filter, which never looks at the pixel to its left — so a row of tiles' scanlines laid end to end is the full image's scanline, and rebuilding moves bytes rather than doing pixel arithmetic.

Whole-reticle files

Every file in this directory is listed in MD5SUMS · SHA1SUMS · SHA256SUMS, in the format the standard tools read:

sha256sum -c SHA256SUMS

The 42 tiles

Position is the tile's top-left corner in the full-resolution image. MD5 and SHA-1 for the same files are in the checksum files linked above.

About VIMC on mpw-004

Resolved: reticle 5CS8010AC, run mpw-004, mask VIMC

Mask layer

VIMC — Via 1 top, RRAM tier

Drawn layer

68:44

Derived from drawn layers as

68:44 AND 201:20

Detail

the second via1 mask of the sky130B RRAM tier (r1v). The RRAM splits via1 in two: the designer draws one 'via' shape, and mask generation separates it into via_bot (cviam = VIM) and via_top (r1v) wherever the via is enclosed by r1c. Rendered as via AND r1c. NOT CTM1/mcon, as once assumed: 8016/17/18 each carry a separate CTM1 plate (400), and 455 sits above MM1 (450) while mcon is below met1

Also

sky130B RRAM tier; r1v, the upper half of the split via1

Used in SKY130

no

mpw-004 — SKY130 MPW-4 (original)

Foundry product

EF MPW#4 (+ 'With ReRAM' split lot)

Reticle set

5CS8010AC

Run date

2021-12-31

Submissions closed

2021-12-31

Projects

52

Fab lots

4216266
4216266
4216266

Identification

confirmed — Lot records: EF MPW#4, lot 4216266 = 'SKY130 MPW 4'; the set is absent from the 2026-08-25 return shipment (destroyed 2025-06) — see 5CS8018AC

Shuttle page

archived copy of https://platform.efabless.com/shuttles/MPW-4 — the platform that hosted it is gone

Foundry data

third_party/shuttle/sky130/mpw-004/foundry

These are renders of the drawn shapes that go on each photomask, taken from the layouts the shuttle's projects taped out — they are not photomask artwork: reticle pitch, the 4x reduction, mirroring and the frame features the fab adds are not modelled. An empty image is a real result; several mask layers are used by no project on a given shuttle. Each die's source layout is named in result.json and lives in the public shuttle repositories under foss-eda-tools.googlesource.com, so every image here can be traced back to the tape-out it came from. Generated with sky130-mask-decoder.