Blog
The FPGA course names its bench board
[the board is the one identified on the bench on 2026-10-08; the still was checked by decoding the PNG, not in a browser] Two lessons of the FPGA course now name the board on our bench, an ALINX AX7203 with the XC7A200T in the 484-ball FBG484 package, and the course lead lists the two X accounts.
Two lessons of the FPGA course pointed at a board that is not on our bench. Lesson 3 gave the chip on our board 676 solder balls, and the task of lesson 1 called the board file of a Wukong the spec of our bench board. The board on the bench is an ALINX AX7203, with the AMD Artix-7 XC7A200T in the 484-ball FBG484 package; a camera read ALINX off the board and its bitstream answered over the UART (t27 fpga/HARDWARE_SSOT.md, 2026-10-08). The 676 balls belong to the FGG676 package of the Wukong board.
What changed
Lesson 1, An FPGA is a grid you configure, now says: our bench chip is an AMD Artix-7 XC7A200T on an ALINX AX7203 board, in the 484-ball FBG484 package. Its task now says that the spec frame is the board file of a Wukong, which carries the same die, and that LUT_TOTAL there holds the AMD logic cell count, 1.6 per LUT6, not a LUT count.
Lesson 3, Pins, balls and the package, now says that on our AX7203 board the XC7A200T is in the 484-ball FBG484, and that the pin map draws the same die in the 676-ball FGG676 package of the Wukong board. The widget did not change; the text now says which package it draws.
The course lead now lists the two X accounts: x.com/t27_dev and x.com/t27_lang.
Where the numbers come from
The board name, the package and the ball count are constants of t27 specs/boards/ax7203.t27. The t27 entry of this post, specs/blog/course_bench_board.t27, imports them and fails its tests if they differ. The same board spec keeps two counts apart that the course once mixed: 134,600 LUT6 and 215,360 logic cells, the AMD figure of 1.6 per LUT6.
Not claimed: that the course is now free of other drift. The drift rules in t27 specs/tri/course/audit.t27 check the triggers they list and nothing else. The still was checked by decoding the PNG, not by looking at it in a browser. The change is in Trinity-S-AI-DNA/trinity#1806.
What this does not settle
- The audit rules in t27 specs/tri/course/audit.t27 check the triggers they list; other drift in the course text is not ruled out.
- The pin map widget of lesson 3 still draws the FGG676 package of the Wukong board; only the text around it changed.
- The still was checked by decoding the PNG; nobody looked at it in a browser for this post.
Receipts
Work with me
Need an FPGA/RTL problem taken to measured hardware?
I work contract and part-time on hardware-AI, FPGA/RTL and ML systems — from specification and open toolchains to reproducible measurements.