When you do 1-operand division on x86 CPUs, EDX:EAX (64 bit) is divided by the 1st operand (32 bit). The result is stored in EAX (32 bit).
So when you divide by EDX:EAX by EDX, what you essentially get is (EDX * 0x100000000 + EAX) / EDX, which result is always above 0x100000000 and does not fit into the target register or the divisor is zero. In both cases a divide exception occurs.
See also this page (from the Intel developer manuals).
Note that this is not specific to the assembler (MASM), but to the platform in this case.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…