Load both numbers, subtract the one at 2001H from the one at 2000H, and store the difference at 2002H.

Last time you were made to reach memory through HL. This time do the opposite: name the addresses directly.

The one thing: SUB always computes accumulator minus operand, so the number being subtracted from has to be the one sitting in A.

What is being checked

• 2002H holds the difference
• The carry flag is set when the subtraction borrowed, and clear when it did not
• 2000H and 2001H still hold the original numbers
• The program reaches HLT
• You use LDA and STA

Two ways to reach memory, and when each is right

LDA 2000H loads the accumulator from an address written into the instruction. STA 2002H stores it back the same way. No pointer to set up, no INX H — but the address is frozen into the program, so three different addresses cost three different instructions.

That is the trade. Direct addressing (LDA / STA) is shorter when there are a couple of fixed, unrelated addresses, which is exactly this problem. Pointer addressing (HL and M) is the only thing that works when the address has to change while the program runs, which is every loop you will ever write. Neither one is the "real" way; knowing which question you are answering is the skill.

Watch the order

SUB B computes A - B, always. If you load the numbers the wrong way round you will get the two's complement of the right answer — DBH where 25H was expected — and it will look like a strange bug rather than a swapped pair.