Isn't the Hand-Mounted Pyro Rocket be a better enchant at lower gear levels, I've never seen it mentioned on these forums. With the hyperspeed accelerators you get one extra attack a minute correct? (12 seconds = 12 attacks * 13% increase = 13.6 attacks or something on that order, i assume it rounds down because you can't have half an attack) While the hand mounted pyro rockets give you an average of 2449.33 damage a minute ((1654+2020)/2=1837 average damage*1.33333 because the cooldown is only 45 seconds = 2449.33 damage per minute) So doesn't that mean that the hyperspeed accelerators are only better if your average white hit is above 2449.33? Where am I doing my math wrong?

I did a few simulations with FeralbyNight about this matter. My results were that until ~875 ArP without the Banner is better to keep the Mjolnir. Take it with a grain of salt until someone else can confirm this.

Also, near 875 ArP, I get better dps with ArP gems with the Mjolnir, even if is over the soft cap by a great margin. But the dps gain is only ~1 dps per epic gem, and near the soft cap Agi gems are better than extra Arp gems so is hard to tell at what level you'll want to regem to Agi or keep Arp gems.

I am trying to figure out if Armor Pen is worth it with 3.2.2 coming out where the soft cap is harder to hit.

Its quite annoying to have to regem with every new piece of gear I get. I am still using my Grim Toll since the runestone does not drop for me. I would like to just drop all Armor Pen gems and gem straight agility from now on. Anyone know anything about this and know if it is that big of a DPS drop.

Ran some Rawr tests, and it says its about the same dps now. Hitting 2500 agi fully buffed in combat having Idol of Mutilation giving me the 200 agility.

I will need to replace Grim Toll with with Banner of Victory if I can.

Anyone else try not dealing with Armor Pen anymore and seeing a drastic drop in DPS?

As I understand it, FBN and SC run actual simulations of a character in combat, which will be more accurate; whereas Tossks and Rawr use mathematical models to predict DPS based on a "typical" encounter. The disadvantage of using simulations is that you have to run a lot of them to predict things that will be overall or consistent upgrades, whereas you can count on a model to give you the same answer for the same set of initial conditions without all the real-world-like noise of a crazy streak of parries or crits (what the cool kids call "RNG high" etc.).

However, in most cases, you should get similar answers from all four. If you don't, something interesting is going on. If you're being told that an ilvl 200 item is better than an ilvl 258 item, be suspicious. If one tool is telling you that ArP is a better gem stat and another is telling you to gem for agi, that's more interesting. If the simulators are giving you one answer and the models a different one, run more simulations or adjust the initial conditions of the models to see if you can find out why.

For me, it really comes down to features. Rawr is by far the most sophisticated and user-friendly when it comes to actually assembling and comparing different gearsets, so I use Rawr mostly. SC is a commandline utility, so that makes it require a bit more effort than looking at the item you want and clicking on it. Etc. Edit 4: A particular strength of Rawr is that it can find synergies in gearsets that take much longer to find with simulations. (The most obvious example is set bonuses, but it goes way beyond that.)

Edit:
Thank you, I will give it a look. I've always liked the helpful data from the FBN project, but resisted an ingame addon for exactly the same reason you mention.

Edit 2: There's a page on the SC Wiki that says what I just did, but a lot more clearly:
FormulationVsSimulation - simulationcraft - Project Hosting on Google Code

Edit 3: One way to get theoretically good answers is to use two things together. For example, take initial information from some simulations, plug it into Rawr, find some guesses as to good sets with Rawr, and then plug those back into simulations to see if the simulations agree.