June 2, 2026
Hot chips, hidden bits
On Reading SRAMs in IR Images, and Establishing Bounds on Trust
Chip trust gets the gossip treatment as commenters ask where secret memory could hide
TLDR: The article says infrared photos can reveal whether a chip has more memory than it claims, which is a big deal for checking that hardware isn’t hiding anything shady. Commenters instantly pushed the paranoia further, asking if “error-checking” space could secretly be extra storage in disguise.
The big reveal in Bunnie’s post is surprisingly juicy for a story about computer chips: you may not be able to see every single memory bit, but you can still catch a chip lying about how much memory it has. Using infrared images, the article argues that hidden memory blocks inside a chip would be hard to sneak in without showing up like a suspicious extra room on a house tour. For anyone worried about “trusting” a device, that matters, because secret memory could potentially hide code that normal software checks never notice.
But the real popcorn moment comes from the community reaction, where readers immediately started playing “okay, but what if the chip is even sneakier?” The standout hot take came from commenter adgjlsfhk1, who tossed in a deliciously paranoid scenario: what if the so-called error-checking bits are really just bonus hidden memory in disguise? That one line basically turned the discussion into a mini conspiracy board, with the vibe of, “Nice scan, but did you check the chip’s fake mustache?” It’s the kind of comment that captures the mood perfectly: impressed by the method, but absolutely unwilling to stop being suspicious.
The tone is half security lecture, half detective drama. The article says infrared inspection can put strong limits on secret memory, but the comment section energy is classic internet: trust, but roast. Readers aren’t just asking whether the chip matches the blueprint — they’re asking how a crafty designer might still try to cheat.
Key Points
- •The article says software-based memory measurement alone cannot guarantee trust if hidden RAM is physically present but not exposed through normal system interfaces.
- •IR imaging is presented as a non-destructive way to place an upper bound on SRAM present on a chip, even if individual bit cells cannot be resolved at 22 nm.
- •The article states that extra SRAM macros would be visible in IR scans because SRAM blocks are much larger than the resolution of an entry-level home IR setup.
- •Baochip-1x is used as the example open-RTL system where visible RAM macros can be compared against source-code claims to validate fabricated memory.
- •The article begins a structural explanation of SRAM macros using the dual-port 1024 x 32-bit rdram1kx32 macro, identifying arrays, rows, columns, and the central spine in the micrograph.