Skip to content

Card signals and chips

An overview of card identity and signals: which signals identify a chip, the model mapping and its confidence, and the cards measured so far. Every value comes from an on-card API or from gp.exe; nothing is inferred from a product name. Cross-card performance comparison is on the results page, and the method is on the methodology page.

AttributeNXP P71D600contactless-d600NXP P71D321contactless-d321
CPLC ICTypeD600D321
Interfacecontactlesscontactless
Measured (UTC)2026-11-03 03:512026-11-03 04:01
Applet versionv10v10
Free persistent335,512 B155,124 B
Free transient (RESET)4,609 B3,974 B
Probes45/5833/58
DetailsOpen full data →Open full data →
Signal Identifies the chip? Notes
ATR no a contactless reader returns a PC/SC pseudo-ATR synthesised from the card’s answer; it can change with the reader or the card configuration
CPLC ICType partly the chip vendor’s product code, needs an external mapping; these cards report D600 and D321
CPLC serial / batch identifies one card factory-programmed and read-only; the values are not published here, only used as a signal that separates cards of the same model rather than naming the model
GlobalPlatform Card Recognition Data partly GP version, SCP and the “JavaCard v3” tag as read by gp.exe
JCSystem.getVersion() no both cards report 0x0300; the patch level is not exposed
Free memory no depends on installed applets and packages — a usage state, not a chip property

The CPLC is programmed during chip production, is read-only and does not depend on any applet, which makes it the strongest hardware signal. It only carries a product code (D321), though: mapping that code to a product name still needs external material, so every “usually labelled” entry below is an inference.

Production identifiers that point to one specific card — serial number, batch, pre-personalizer, KDD and SSC — are not written to public data; the published data keeps only the chip product code, production dates and protocol signals.

CPLC ICType Inferred chip Usually labelled Main evidence Confidence
D600 NXP P71D600 JCOP 4.5 / J3R452 CPLC product code; on-card behaviour matches Java Card 3.0.5 (AES-128, explicit P-256, RSA-2048) inferred, high
D321 NXP P71D321 JCOP 4 / J3R180 CPLC product code; the J3R180 product page says “JCOP 4 OS for Java Card 3.0.5 applets on 167736 bytes of available EEPROM”, the same order of magnitude as the clean-card free persistent memory measured on the card in the latest run (155,140 B) inferred, high

Claims deliberately avoided:

  • No RAM size and no CPU clock. Neither is publicly reliable for these parts, and the card API does not expose them; getAvailableMemory returns card-wide free space, not a chip specification.
  • J3R452 / J3R180 are not read from the card. The card reports a CPLC product code; the marketing name comes from external mapping.
  • The ATR is not used to identify the chip. It is kept only as “what the reader saw”.

gp.exe prints a decoded date next to fields such as ICFabricationDate, but tools and bit-field conventions disagree:

  • for 3129 on the D321 card, gp.exe prints 2023-05-09, while the common CPLC layout (7-bit year / 4-bit month / 5-bit day, 1984 epoch) yields 2008-09-09;
  • for 5336 on the D600 card the two disagree as well: 2025-12-02 vs 2025-09-22.

Dates do not affect any performance conclusion, so only raw hexadecimal values are quoted here rather than propagating a date that may be wrong.

Another card was read once through the contact reader (ACS ACR39U ICC Reader 0). Only its ATR and CPLC were captured; the benchmark applet was never installed on it and it took part in no measurement:

  • ATR: 3BFD9600008131FE45534C4A35324778787979797A5224;
  • the ATR contains SLJ52G, a fragment common in Infineon SLE78 product names;
  • yet its CPLC reports ICType D321, which is an NXP product code.

The two signals contradict each other, which is exactly why “ATR identifies the chip” and “CPLC identifies the chip” are both too strong on their own. That card is outside the identity and performance conclusions, and the benchmark never writes to it.