Public

Thread: Can masks partially be reused?

12 Discord messages from ℹ️ - Information/⁉️-questions/threads/can-masks-partially-be-reused.md.

Category: ℹ️ - Information Parent channel: #⁉️-questions Thread ID: 1535544860789051413

Messages: 12

message-1535544860789051413

2026-08-08T07:07:03.856+00:00 — rahulbhagwat

Content

https://youtu.be/3MKRjt59hh4?si=43BobvNwE528ZjsH&t=173

In the video he says Taalas builds a base layer but different routing layers per model (eg. metal 5 - 10 is different per design)

I was wondering if something like this reduces non recurring mask set costs since you could reuse masks for the bottom layers?

Embeds

<details> <summary>Embeds data</summary>
[
  {
    "author": {
      "name": "TechTechPotato",
      "url": "https://www.youtube.com/channel/UC1r0DG-KEPyqOeW6o79PByw"
    },
    "color": "#FF0000",
    "description": "Taalas has built a chip with a single AI model wired directly into its metal layers, with no programmability at all, so one part runs one model and nothing else. In the public demo it returns a complete answer in about 33 milliseconds, which works out at roughly 14,200 tokens per second for one user, where an H200 sits nearer 230. When the compa...",
    "fields": [],
    "images": [],
    "inlineEmojis": [],
    "thumbnail": {
      "height": 720,
      "url": "2026-08_media/maxresdefault-B905E.jpg",
      "width": 1280
    },
    "timestamp": null,
    "title": "Why did AMD just buy this REALLY WEIRD chip company?",
    "url": "https://www.youtube.com/watch?v=3MKRjt59hh4&start=173",
    "video": {
      "height": 720,
      "url": "2026-08_media/3MKRjt59hh4-DCCDD",
      "width": 1280
    }
  }
]
</details><details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535545489720610929

2026-08-08T07:09:33.805+00:00 — tnt (@246tnt)

Content

Yes
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/88d17af9d19ab43565b0c46751ef7d12-DFEE5.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535546161794912368

2026-08-08T07:12:14.04+00:00 — rahulbhagwat

Content

are routing layer masks significantly cheaper than base layer masks? (i'm guessing yes otherwise fpga's would be more economical for this)
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535547262007119882

2026-08-08T07:16:36.351+00:00 — tnt (@246tnt)

Content

The higher metals ones yes because they usually have coarser pitch. Also there are less masks for metal. AFAIK it's one mask for metal and one mask for via and that's it.  While the FEOL has finer masks and then a lot more of them for the implants and all that stuff.

Reactions

<details> <summary>Reactions data</summary>
[
  {
    "count": 1,
    "emoji": {
      "code": "thumbsup",
      "id": "",
      "imageUrl": "2026-08_media/1f44d-27259.svg",
      "isAnimated": false,
      "name": "👍"
    },
    "users": [
      {
        "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
        "color": null,
        "discriminator": "0000",
        "id": "488155070052171816",
        "isBot": false,
        "name": "rahulbhagwat",
        "nickname": "rahulbhagwat"
      }
    ]
  }
]
</details><details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/88d17af9d19ab43565b0c46751ef7d12-DFEE5.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535560820833128448

2026-08-08T08:10:29.027+00:00 — Tim 'mithro' Ansell (@mithro_)

Content

Not really a process that can work with a traditional foundry
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535562122694492200

2026-08-08T08:15:39.415+00:00 — Tim 'mithro' Ansell (@mithro_)

Content

I would like to see a metal programmable gate array that can be programmed via something like the hacker fab steppers.
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535582874454196304

2026-08-08T09:38:07.02+00:00 — namibj

Reply/reference

<details> <summary>Reference data</summary>
{
  "channelId": "1535544860789051413",
  "guildId": "1361349522684510449",
  "messageId": "1535560820833128448"
}
</details>

Content

I think it has a good chance of working when the production between FEOL and BEOL is split between companies like for example the copper X-Fab BEOL option on IHP SG13.
But otherwise, yeah, not really.
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/f885abc0dddd0a33e4d4991b3c99091d-7ACD2.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {
    "mentions": [
      {
        "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
        "color": null,
        "discriminator": "0000",
        "id": "169786952432746498",
        "isBot": false,
        "name": "mithro_",
        "nickname": "Tim 'mithro' Ansell",
        "roles": []
      }
    ]
  }
}
</details>

message-1535808189210173450

2026-08-09T00:33:26.244+00:00 — Tim 'mithro' Ansell (@mithro_)

Content

https://bit.ly/ws-gatearray-v1

Embeds

<details> <summary>Embeds data</summary>
[
  {
    "description": "GF180MCU Gate Array https://bit.ly/ws-gatearray-v1  Goal Create a good gate array which is programmable with post manufacturing additional of a metal layer in the  1-4um sizes.  Specifications  Final Metal Layer for Programming  The goal is to have the gate array programmable by adding a single e...",
    "fields": [],
    "images": [],
    "inlineEmojis": [],
    "thumbnail": {
      "height": 630,
      "url": "2026-08_media/AHkbwyJEZ5uvEq4ofXg1xH9zhCBJbAMcNYuL_IoFed-3392A",
      "width": 1200
    },
    "timestamp": null,
    "title": "wafer.space - GF180MCU Gate Array (bit.ly/ws-gatearray-v1)",
    "url": "https://bit.ly/ws-gatearray-v1"
  }
]
</details>

Reactions

<details> <summary>Reactions data</summary>
[
  {
    "count": 1,
    "emoji": {
      "code": "raised_hands",
      "id": "",
      "imageUrl": "2026-08_media/1f64c-7C820.svg",
      "isAnimated": false,
      "name": "🙌"
    },
    "users": [
      {
        "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
        "color": null,
        "discriminator": "0000",
        "id": "488155070052171816",
        "isBot": false,
        "name": "rahulbhagwat",
        "nickname": "rahulbhagwat"
      }
    ]
  }
]
</details><details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535808293442687067

2026-08-09T00:33:51.095+00:00 — Tim 'mithro' Ansell (@mithro_)

Content

The cool thing about doing something like that is you could potentially have a turn around time of days or even hours.

Reactions

<details> <summary>Reactions data</summary>
[
  {
    "count": 1,
    "emoji": {
      "code": "thumbsup",
      "id": "",
      "imageUrl": "2026-08_media/1f44d-27259.svg",
      "isAnimated": false,
      "name": "👍"
    },
    "users": [
      {
        "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
        "color": null,
        "discriminator": "0000",
        "id": "488155070052171816",
        "isBot": false,
        "name": "rahulbhagwat",
        "nickname": "rahulbhagwat"
      }
    ]
  }
]
</details><details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535900448429711391

2026-08-09T06:40:02.556+00:00 — rahulbhagwat

Content

^ would also need a deposition step after litho right?
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>

message-1535901294886260816

2026-08-09T06:43:24.367+00:00 — rahulbhagwat

Reply/reference

<details> <summary>Reference data</summary>
{
  "channelId": "1535544860789051413",
  "guildId": "1361349522684510449",
  "messageId": "1535808189210173450"
}
</details>

Content

This is super exciting btw! Going to start working on the base layer and scripting to reconfigure top metal.

Also, is ‘metal’ programmable gate array a name you just made cz that definitely needs to be an official term
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/e69d5a0ef16fab73702f739d1fde6389-BE72C.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {
    "mentions": [
      {
        "avatarUrl": "2026-08_media/9679bb77da48d47c905b398d98a83780-75B68.png",
        "color": null,
        "discriminator": "0000",
        "id": "169786952432746498",
        "isBot": false,
        "name": "mithro_",
        "nickname": "Tim 'mithro' Ansell",
        "roles": []
      }
    ]
  }
}
</details>

message-1535992397677338674

2026-08-09T12:45:24.965+00:00 — namibj

Content

mask programmable usuaually afaik though using a via layer fairly deep, like via2 or so
<details> <summary>Other message metadata</summary>
{
  "author": {
    "avatarUrl": "2026-08_media/f885abc0dddd0a33e4d4991b3c99091d-7ACD2.png",
    "discriminator": "0000",
    "isBot": false
  },
  "message": {}
}
</details>