Forums → Games → Gemcraft 2: 4096 ultra-hardcore upgrading
12 | 4905 |
For when you want to squeeze the most out of your gems!
In the kiwi chat, I half-jokingly suggested to 12345ieee to generate a 4096 upgrade as a sort of masochistic test of patience. A few moments later, the scheme was uploaded to pastebin. After witnessing the wall of text and being utterly intimidated by that monstrosity, I then issued psorek a (non) challenge to tree that titanic scheme. After a short while, he produced this. The text file is so large it cannot fit on pastebin or github!
After experiencing a huge burst of motivation in the middle of the night to learn this upgrade scheme, I immediately went to work using Photoshop and Irfanview to help make sense of this Yggdrasil. It took me a few days to condense it into a format I could work with, and an extra day to fix all the errors in the parentheses scheme, but, after all this hard work, I have a workable format.
Here is the keypad I'm using:
For this example the base gem is a G31/37e, and "U" simply means a G+1 in the following plan. Now, here it is:
KEY COST CHECKS
166 = G+7 + G+5 + G+2 + G+1
172 = G+7 + G+5 + G+3 + G+2
594 = G+9 + G+6 + G+4 + G+1
2900 = G+11 + G+9 + G+8 + G+6 + G+4 + G+2
1196 = G+10 + G+7 + G+4 + G+3 + G+2
GEM CODE
3 = (G+1 + G+0)
4 = (G+1 + 2x G+0)
5 = (G+1 + 3x G+0)
6 = (G+1 + 4x G+0)
7 = (G+1 + 5x G+0)
8 = (G+1 + 6x G+0)
8 combine (8c) = ((((2+1)+1)+1)+1)+2 [never used alone, not shown]
10 combine (10c) = 8c + G+1
A = 10 = (G+2 + (G+1 + 4)) = G+3
B = 20 = (G+1 + 4) + ((3+8) + 3) = G+3
C = 30 = 10 + 20 = G+4
D = 34 = G+4 + (G+3 + A) = G+5
E = 48 = 30 + (G+3 + (G+2 + (G+1 + 4))) = G+5
F = 58 = (((3+8) + ((((((3+8) + 3) + 4) + 5) + 6) + 7)) + (3+8)) = G+3
G = 78 = B + F = G+4
G1 = 82 = D + E = G+6
H = 108 = C + G = G+5
I = 156 = H + E = G+6
J = 166 = ((((((((3+8) + (((((((4+8) + 3) + 4) + 5) + 6)))) + 10c) + 10c+3) + ((10c+3) + 4)) + (((10c+3) + 4) + 5)) + ((((10c+ 3) + 4) + 5) + 6)) + (((((10c+3) + 4) + 5) + 6) + 7)) = G+3
K = 172 = (((((((3+8) + 3) + ((((((((G+1 + 6) + G+1) + 3) + 4) + 5) + 6) + 6) + 7)) + 10c+3) + (((3+8) + 3) + 4)) + ((((3+8) + 3) + 4) + 5) + ((((10c+3) + 4) + 5) + 6)) + (((((10c+3) + 4) + 5) + 6) + 7)) = G+3
L = 594 = ((((((51 + 143) + 49) + 61) + 76) + 95) + 119) = G+4
L subgem:
1. (((3+8) + (((((4+8) + 3) + 4) + 5) + 6)) + 10c) = 51 = G+3
2. (((((((10c + (((((3+8) + 3) + 4) + 5) + 6)) + (G+1 + 7)) + (3+8)) + ((3+8) + 3)) + (((3+8) + 3) + 4)) + ((((3+8) + 3) + 4) + 5)) + (((((3+8) + 3) + 4) + 5) + 6)) = 143 = G+3
3. ((10c + (((((3+8) + 3) + 4) + 5) + 6)) + 10c) = 49 = G+3
4. 3 + (4+8) = 61 = G+3
5. 4 + ((4+8) + 3) = 76 = G+3
6. 5 + (((4+8) + 3) + 4) = 95 = G+3
7. 6 + ((((4+8) + 3) + 4) + 5) = 119 = G+3
8. ((((((1+2) + 49) + 61) + 76) + 95) + 119) = 594 = L
2900-gem
1. L + (J + F) = G+5
2. 1 + ((J + F) + G) = G+6
3. 2 + (((J + F) + G) + H) = G+7
4. 3 + ((((J + F) + G) + H) + I) = G+8
5(1). 4 + (((((J + F) + G) + H) + I) + (I + G1)) = G+9
1196-gem
1. (G1) + (G+5 + D) = 148 = G+7
2. (H + E) + (G1) = 238 = G+7
3. 1 + 2 = 386 = G+8
4.((K + F) + G) + H) + I) + (I + (G1)) = 810 = G+8
5(2). 3 + 4 = G+9
4096-gem
1. 5(1) + 5(2) = G+10
Note that there is a 30-gem in there that is different from gem C which is used in gems J and L. It is [(((((4+8) + 3) + 4) + 5) + 6)] instead of [(A+B)]. It is not shown in the key, thus its expanded representation. I will update this guide soon to shorten that 30-gem and include a new image having that gem listed. For now, this is the first version of the guide, which others will hopefully peruse and figure out how to make this upgrade plan simpler and easier to accomplish.
Now, for a comparison at equivalent grades:
G37e:
~15% better than two 64c operations. Pretty good! Let's go to a higher grade.
G61e:
~26.2% more mana per hit. Getting better! Now, for the final comparison.
G97e:
The 1024c+64c gem is 29.4% better than the 64c gem, and the 4096c gem is an additional 12.1% better than that. The 4096 gem sees a ~45.1% improvement over a standard 64c gem at that grade! This plan is horrendously expensive in the early stages of a mana farm, but indispensable for runs lasting 15+ days where every bit of leech earned can cut upgrading time by several days.
As an added bonus, the 4096c gem in the last image beats out a standard U gem of the same cost, even if they fired at their actual speeds.
G97 U: 855,441,483,443,320,556,233.37 or 8.55e+20 mana per second.
G81/97e by 4096: 1,028,410,090,028,280,988,992.552 or 1.028e+21 mana per second.
Overall, the 4096 gem is 20.2% better than a standard U gem at that grade.
THANK YOU psorek for putting the 4096c into an infinitely more understandable tree format, and 12345ieee for creating gemforce and subsequently this upgrade scheme!
- 12 Replies
You see me deeply impressed!
Great job, all three of you!
I wonder, if this will work as well with the crit hit gems, as the damage of the gem goes down by 10%...
So...since there is a 4096 combine should I attempt to make a decent 2048 spec? lol.
I'm certain that I won't be able to do more than 128, if I'd spend much time maybye 256 specs, so feel free to experiment.
Aww...I was hoping at least for a corrected 512spec...I know mine sucks quite a bit :P
Well, you'd need much better computer scientist than me. Like algorithmics PhD... Not worth the effort needed.
A pair of good 512 spec gems, killgem and managem, would be a very nice thing to have.
You corrected your 32 spec managem, going from the one in your high end guide to your 32 spec version 2, that was much more skewed towards black, because the first gem was built with the idea of 16c amps, and the second around 64c amps. Why can't you just take your program/algorithm and just use 512 gems instead of 32, knowing how much a set of six 1024 manaamps provides? And then do the same thing only focusing on the damage and critical hit for a killgem?
it's more complicated... Memory needed for 128 is around 16gb and it'll be like 64gb for 256 - you can't "just get" 512 gems, you have to combine these gems in every possible way to find out the best. Even if you do that in "smart" way, it takes lots of time. You can even combine one set of gems into different stats gems, there is just no "best way"...
You know, i dont understand a single thing about this. Are you guys all math gods or something? What kind of math do you do to figure out all these crasy things haha
Oh, this thread resurfaced again, now we have 512 gem specces (even 1024 ones) and combines as long as 262144, but nobody left that wants them.
There is no need to be a "math god" to understand and apply math for gems.
All the math I used was from at most 3rd year of high school.
i reacy lvl 90 gem with combine 73 . 73 is e21 90 is e20. %1000 . just with 32 spec gem combine 32x32x32x32.....
here's some info i collected for pure orange gems.
hope its somewhat helpful to someone
Ok @Rocksfire, I can see why you looked for me yesterday.
First of all, impressive collection, I hope you didn't do the 1M combine by hand.
Second, there is a 16c for orange gem, as there is a combine for every length between 2 and 1,048,576, it just isn't on the results/leech folder of gemforce because it's already on the omnia recipe for 32s/16c, as the amps combine recipe.
Third, comparing the leech ratios against U is not really enough to tell you that e.g. 256c is better than 64c, because you are comparing gems at different equivalent grades.
The correct way to compare combine recipes (if you are interested in this) is to compare growth = log(power_after_combine/power_before_combine)/log(combine_length)
If you (or anyone else) would like to discuss recipes further you can find me at the IRC #gemcraft channel for a couple of hours now.
You must be logged in to post a reply!