Skip to the content
2 layersSmall-3, the printed circuit board in 3D

Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

Copper layers
2
Thickness
1.6 mm
Parts
63
Nets
50
Pads
156
Tracks, vias
205, 2
Routed
Yes
KiCad
KiCad 8.0 (file 20240108)
Created
11 Oct 2024
Updated
3 Mar 2026

Overview

Small-3 is an open-source printed circuit board by NVlabs, from the NVlabs/Cypress repository on GitHub, published under the Apache-2.0 licence. It is a 2-layer board with 63 parts on 50 nets, routed with 205 tracks and 2 vias.

Its parts include 45 capacitors, 12 resistors, 3 ICs and 3 inductors.

README

README.md

Cypress: VLSI-Inspired Scalable GPU-Accelerated PCB Placement

Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

Cypress PCB Placement Animation

How to Build

You can build in two ways:

  • Build without Docker by following the instructions of the DREAMPlace build at README_DREAMPlace.md.
  • Use the provided Dockerfile to build an image with the required library dependencies.

Run PCB Placement

sh
cd /Cypress
python dreamplace/Placer.py path/to/your/config.json

Run Hyperparameter search

sh
./tuner/run_tuner.sh 1 1 	test/tune/pcb-configspace.json 	/path/to/aux_file 	test/tuner/tuner-ppa.json "" 	100 4 0 0 10 	./tuner 	./results/result_path

Acknowledgement

Cypress builds upon and extends the following open-source projects:

  • DREAMPlace: A deep learning toolkit-enabled VLSI placement framework that provides the foundation for our GPU-accelerated placement engine.

  • AutoDMP: An automated design methodology platform that inspired aspects of our hyperparameter tuning approach.

We thank the authors and contributors of these projects for their valuable work that made Cypress possible.

Citation

If you find this work useful, please cite the following paper:

bibtex
@inproceedings{zhang2025cypress,
  title={Cypress: VLSI-Inspired PCB Placement with GPU Acceleration},
  author={Zhang, Niansong and Agnesina, Anthony and Shbat, Noor and Leader, Yuval and Zhang, Zhiru and Ren, Haoxing},
  booktitle={Proceedings of the 2025 International Symposium on Physical Design},
  pages={xx--xx},
  year={2025}
}

Parts

63
QtyValueFootprintReferences
1C10_21_4C10_21_4C10_21_4
1C10_22_4C10_22_4C10_22_4
1C10_23_4C10_23_4C10_23_4
1C11_21_4C11_21_4C11_21_4
1C11_22_4C11_22_4C11_22_4
1C11_23_4C11_23_4C11_23_4
1C12_21_4C12_21_4C12_21_4
1C12_22_4C12_22_4C12_22_4
1C12_23_4C12_23_4C12_23_4
1C13_21_4C13_21_4C13_21_4
1C13_22_4C13_22_4C13_22_4
1C13_23_4C13_23_4C13_23_4
1C14_21_4C14_21_4C14_21_4
1C14_22_4C14_22_4C14_22_4
1C14_23_4C14_23_4C14_23_4
1C15_21_4C15_21_4C15_21_4
1C15_22_4C15_22_4C15_22_4
1C15_23_4C15_23_4C15_23_4
1C1_21_4C1_21_4C1_21_4
1C1_22_4C1_22_4C1_22_4
1C1_23_4C1_23_4C1_23_4
1C2_21_4C2_21_4C2_21_4
1C2_22_4C2_22_4C2_22_4
1C2_23_4C2_23_4C2_23_4
1C3_21_4C3_21_4C3_21_4
1C3_22_4C3_22_4C3_22_4
1C3_23_4C3_23_4C3_23_4
1C4_21_4C4_21_4C4_21_4
1C4_22_4C4_22_4C4_22_4
1C4_23_4C4_23_4C4_23_4
1C5_21_4C5_21_4C5_21_4
1C5_22_4C5_22_4C5_22_4
1C5_23_4C5_23_4C5_23_4
1C6_21_4C6_21_4C6_21_4
1C6_22_4C6_22_4C6_22_4
1C6_23_4C6_23_4C6_23_4
1C7_21_4C7_21_4C7_21_4
1C7_22_4C7_22_4C7_22_4
1C7_23_4C7_23_4C7_23_4
1C8_21_4C8_21_4C8_21_4
1C8_22_4C8_22_4C8_22_4
1C8_23_4C8_23_4C8_23_4
1C9_21_4C9_21_4C9_21_4
1C9_22_4C9_22_4C9_22_4
1C9_23_4C9_23_4C9_23_4
1L1_21_4L1_21_4L1_21_4
1L1_22_4L1_22_4L1_22_4
1L1_23_4L1_23_4L1_23_4
1R1_21_4R1_21_4R1_21_4
1R1_22_4R1_22_4R1_22_4
1R1_23_4R1_23_4R1_23_4
1R2_21_4R2_21_4R2_21_4
1R2_22_4R2_22_4R2_22_4
1R2_23_4R2_23_4R2_23_4
1R3_21_4R3_21_4R3_21_4
1R3_22_4R3_22_4R3_22_4
1R3_23_4R3_23_4R3_23_4
1R4_21_4R4_21_4R4_21_4
1R4_22_4R4_22_4R4_22_4
1R4_23_4R4_23_4R4_23_4
1U1_21_4U1_21_4U1_21_4
1U1_22_4U1_22_4U1_22_4
1U1_23_4U1_23_4U1_23_4

More from NVlabs/Cypress

11
  • 2 layersSmall-9, the printed circuit board in 3D

    Small-9

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    476 parts56Apache-2.0
  • 2 layersSmall-2, the printed circuit board in 3D

    Small-2

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    240 parts56Apache-2.0
  • 2 layersSmall-10, the printed circuit board in 3D

    Small-10

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    136 parts56Apache-2.0
  • 2 layersSmall-8, the printed circuit board in 3D

    Small-8

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    115 parts56Apache-2.0
  • 2 layersSmall-1, the printed circuit board in 3D

    Small-1

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    57 parts56Apache-2.0
  • 2 layers31.2 × 21 mmSmall-7 legacy, the printed circuit board in 3D

    Small-7 legacy

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    52 parts56Apache-2.0
  • 2 layersSmall-7, the printed circuit board in 3D

    Small-7

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    52 parts56Apache-2.0
  • 2 layers31.3 × 21 mmS7-cypress, the printed circuit board in 3D

    S7-cypress

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    50 parts56Apache-2.0
  • 2 layersSmall-6, the printed circuit board in 3D

    Small-6

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    46 parts56Apache-2.0
  • 2 layers31.3 × 21 mmSmall-4, the printed circuit board in 3D

    Small-4

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    43 parts56Apache-2.0
  • 2 layers31.3 × 21 mmSmall-5, the printed circuit board in 3D

    Small-5

    NVlabs/Cypress

    Cypress is a scalable, GPU-accelerated PCB placement method inspired by VLSI. It incorporates tailored cost functions, constraint handling, and optimized techniques adapted for PCB layouts.

    41 parts56Apache-2.0

Similar boards

  • 2 layers143.3 × 116.6 mmLeft, the printed circuit board in 3D

    Left

    hackclub/OnBoard

    💡 Join 1,000 teenagers and make your first PCB with a $100 grant!

    63 parts915
  • 2 layers107 × 54 mmInterface board 1, the printed circuit board in 3D

    Interface board 1

    Green-bms/SmartBMS

    Open source Smart Battery Management System

    63 parts756
  • 2 layers81.9 × 39.8 mmCAUTION LIGHTS PANEL, the printed circuit board in 3D

    CAUTION LIGHTS PANEL

    jrsteensen/OpenHornet

    OpenHornet 1:1 F/A-18C Simulator Repository

    63 parts431
  • 2 layers76.3 × 122.8 mmDiscipad pcb, the printed circuit board in 3D

    Discipad pcb

    coseyfannitutti/discipad

    17-key numpad assembled with only through hole components, including usb type-c

    63 parts221GPL-3.0
  • 2 layers53 × 47 mmServo control, the printed circuit board in 3D

    Servo control

    eddyem/stm32samples

    Code samples and ready projects based on bare-metal STM32

    63 parts208
  • 2 layers98 × 38 mmBldc control ver2, the printed circuit board in 3D

    Bldc control ver2

    pekkaroi/bldc-drive

    Cheap and simple brushless DC motor driver designed for CNC applications using STM32 microcontroller

    63 parts199GPL-2.0
  • 2 layers60 × 75 mmMakennbot, the printed circuit board in 3D

    Makennbot

    idlehandsdev/makennbot

    A neural network bot for Make

    63 parts175

Questions

What parts does the Small-3 use?

63 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Where can I download the Small-3 KiCad files?

From the NVlabs/Cypress repository on GitHub, at the commit this page was made from: the board is artifacts/kicad/small-3.kicad_pcb. PCBKit opens the board in the browser, with nothing to install.

Can I use the Small-3 design in my own project?

Its repository publishes it under the Apache-2.0 licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).