[Assembly] signed math

Hi,

I'm taking my first steps in assembly and I'm a bit confused over signed versus unsigned math. I'm debugging a simple executable which at one point does a subtraction.

Before the operation, this is what AX and BX look like:

AX CF8F
BX 2300

After "sub ax, bx" the sign flag is set and AX becomes

AX AC8F

When thinking in unsigned numbers I understand what is happening.

53135 - 8960 = 44175

However, when I'm trying to understand what happens in terms of signed numbers, I'm a bit confused.

AX CF8F = 1100 1111 1000 1111 = -100 1111 1000 1111 = -20367
BX 2300 = 0010 0011 0000 0000 = +010 0011 0000 0000 = 8960

-20367 - 8960 = -29327 = -111 0010 1000 1111 = 1111 0010 1000 1111 = F28F

So obviously there is something wrong with my reasoning. Who can help me out?

Thanks

Are AX and BX 16 or 32 bit registers?

Edit:
Also, are you sure the architecture uses sign magnitude instead of 2's complement? If it's 2's complement then
1100 1111 1000 1111 = - 011 0000 0111 0001 = -12,401

The registers are 16 bit. The architecture is x86, I'm using Windows XP on a virtual machine.

I'll look into 2's complement, I wasn't aware of this.

Ok, it's clear to me now. Thank you!