Ive actually done the full maths on this, while I was bored yesterday for my blog. Here are the relevant numbers picked out of my article

TLAC Chance of Glance Chance of Pen Chance of kill
VS AV 10 13.88% 41.66% 13.88%
VS AV 11 13.88% 27.76% 9.25%
VS AV 12 13.88% 13.88% 4.44%
VS AV 13 13.88% 0% 0%
VS AV 14 0% 0% 0%
Assault Cannon Chance of Glance Chance of Pen Chance of kill
VS AV 10 11.11% 22.22% 7.41%
VS AV 11 11.11% 11.11% 3.70%
VS AV 12 0% 11.11% 3.70%
VS AV 13 3.70% 7.41% 2.46%
VS AV 14 3.70% 3.70% 1.23%
Melta over half range Chance of Glance Chance of Pen Chance of kill
VS AV 10 11.11% 44.44% 24.07%
VS AV 11 11.11% 33.33% 18.55%
VS AV 12 11.11% 22.22% 12.96%
VS AV 13 11.11% 11.11% 7.41%
VS AV 14 11.11% 0% 1.85%
Melta under half range Chance of Glance Chance of Pen Chance of kill
VS AV 10 1.83% 64.84% 33.03%
VS AV 11 3.66% 61.17% 31.20%
VS AV 12 5.49% 55.68% 28.76%
VS AV 13 7.32% 48.36% 25.40%
VS AV 14 9.15% 39.21% 21.14%
The autocannons performs considerably better than I thought it would do, but you need to give up the second arm to really make it better than the assault cannon. Of course, a rifleman will be giving up its second arm, but you do lose a lot of utility in the dreadnought design. Bearing mind these numbers are per shot. So the autocannon and assault cannon get more shots than the multi melta. My numbers are on a per bs4 shot basis.