Greatsword gem tags: line three should be a cooldown
✍️ Alice・愛麗沙
The Greatsword Warrior is the one class where we will tell you to make line three a cooldown line, and the recommendation is arithmetic rather than preference. Smash sits on every damage skill this class has — about 99% coverage — and Support covers roughly 88%. The other five tags together do not reach a tenth of your output. When a third damage tag buys +0.16%, spending that slot on shortening the cooldown of nearly nine tenths of your damage is the better trade — and better still once you notice what the season buff for this class is: skill-use speed, not cooldown reduction, so your animations already outrun your cooldown bars. The order is Smash damage, Support damage, Support cooldown. Two warnings: Disrupt is the newest trap, because a new passive carries it, and never spend a Gem Skill Engraver on a crude-tier gem, since promotion re-randomises the tag set.
Update 9/14: Taiwan’s 9/9 update log (thread 3541097) changed three Greatsword runes, all in the direction of this article’s conclusion: “[飾品符文：斷絕] no longer grants the Wound damage-over-time; instead max uses become 2, with longer cooldown and higher damage” — the DoT line that this article counted toward Hinder’s 8% share no longer exists on Taiwan, so Hinder’s share only shrinks; “Vengeance now stacks up to 100” and “no longer disappears when combat ends”; “[飾品符文：反擊] grants 1 Battle Rhythm when hit mid-use, regardless of stance”. On the gem side, gem alchemy was reworked so that “only Starlight Prism synthesis remains”: rough-S single-colour gems combine straight into Rough Starlight Prism S, and Elite-and-above gems other than Starlight Prism go through the new “item conversion” — “don’t reroll rough-tier gems, save engravers for the Elite promotion” stands. Tag swapping is still not on Taiwan.
Coverage: two tags finish the whole class
A tag line pays out only on skills carrying that tag, so a line is worth its number multiplied by the share of your damage carrying the tag.
| Tag | Rotation skills carrying it | Share of output | One damage line, on total output | One cooldown line, best case |
|---|---|---|---|---|
| Smash | 6 / 6 | 99% | +1.98% | −0.64% |
| Support | 5 / 6 | 88% | +1.76% | −0.57% |
| Disrupt | 2 / 6 | 8% | +0.16% | −0.05% |
| Movement | 1 / 6 | 4% | +0.08% | −0.03% |
Coverage is compiled from Korean skill-tag listings and rotation shares, not an official disclosure. The two right-hand columns scale one Elite Star Prism line (+2.0% damage / −0.65% cooldown) by that share; the cooldown column assumes the seconds saved convert fully into extra casts, which they do not — read it as a ceiling, not an expectation.
Smash converts to +1.98%, Support to +1.76%, and then the table falls off a cliff to Disrupt at +0.16%. Smash and Support between them already cover the overwhelming majority of this class’s skill damage; the table is that sentence with numbers attached.
One figure carries a lot of the weight: a single finisher accounts for more than half of total output on its own, and it is tagged Smash·Support. So the top two rows are not merely “many skills”. They are “the skill”.
Why line three becomes a cooldown line
This is the core of the page, and the only place on this site where we recommend departing from “always take damage.”
The first reason is arithmetic. On an Elite Star Prism a damage line is +2.0% and a cooldown line −0.65%; the full ladder runs Crude S 1.7/0.55, Elite 2.0/0.65, S 2.1/0.7, Whole 2.2/0.75, Triple Quartz 2.1/0.7 over two lines. In general damage wins because 2.0 is roughly three times 0.65 — but that comparison assumes the two lines cover the same amount of your damage. Here they do not. Your third damage tag would be Disrupt at 8% coverage, worth +0.16%. A Support cooldown line instead acts on roughly nine tenths of your skills. Same slot, incomparable contents, so the usual conclusion does not transfer.
The second reason is pacing. The season buff this class received is skill-use speed, not cooldown reduction — and once the animations run faster, guide sites observed that some skills’ cooldowns can no longer keep up with the hands driving them. That is precisely the state in which a cooldown line stops being a rounding error: the cooldown has become the genuine bottleneck.
While we are here: skill-use speed is not cooldown reduction. It acts on the preparation phase — casting, charging — and the use phase — animation, channelling — and neither formula contains a cooldown term. This class has already had one official description mislabel the stat as damage-increase before it was corrected, which is all the more reason to keep the distinction clean. The full comparison.
Why Support cooldown rather than Smash cooldown? Because Support’s reduction also reaches a Support-tagged utility skill, covering one more button. The Korean guide site ranks it ahead of Smash cooldown for that reason; the wiki ranks them the other way round. We follow the guide site and record the disagreement rather than averaging two sources into a number neither published.
Disrupt is this class’s newest trap
The new passive carries the Disrupt tag, and a fair number of players concluded from that alone that Disrupt had been promoted. It has not. A Disrupt tag on the new passive is not a reason to prioritise Disrupt damage; the coverage table is the argument, since Disrupt reaches two skills plus a damage-over-time component and totals 8%.
Two more phantom options are Combo and Movement. They exist only under a rune nobody runs any more, one that rewrites your first skill from Smash·Support to Combo·Movement — demolishing the class’s tag coherence by hand. Unless you deliberately run that, both are zero for you.
Your gem plan is downstream of your rune plan. And the “smash enhancement” printed on a rune reads the judgement result of each individual hit, while a gem tag reads the word stamped on the skill — same name, different system.
Reading the “separate multiplier” correctly
Gem tags do occupy their own multiplicative bucket, distinct from attack power, damage-increase and crit. But inside that bucket the lines are additive with each other. All your Smash damage lines are summed into one figure and multiplied in once. “Each 2% line compounds” is wrong; “1% of Smash damage is about 1% of total damage, scaled by coverage” is right. Ceiling at +132% by addition, not by product.
A corollary specific to this class: because Smash coverage is so close to 1, you can estimate your gem returns as flat, universal damage. Which also means that as you fill twenty-two sockets your displayed power rating barely moves while your real damage climbs the whole way — one of the cleanest demonstrations that a combat-power number is not a strength number.
Crude gems are a trap, and not because they’re weak
The mechanism: promotion re-randomises the tag set. Fusing three gems into one rolls fresh tags on the result — unless you spend an extra 5 Gem Skill Engravers at promotion to carry the first-slotted gem’s tags across. So any engraver spent on a gem that is not your endpoint is wiped at the next promotion. Leave crude and Crude S alone, fill all twenty-two sockets by volume, and reserve engravers for tag carry-over only. Taiwan’s supply is the adventurer pass at ranks 23 and 24 — five per rank free, ten paid.
This class has it easier than most, because you only need two damage tags — the odds a random roll gives you something usable are decent, so fusing and picking works especially well.
Cheaper still: fresh Star Prisms roll random tags anyway, so fuse several and socket the best roll rather than forcing one stubborn gem into shape. Officially a re-roll is just “random”, and no odds distribution has ever been published — we checked namu wiki, Inven, arca, dcinside and the Korean gem simulators.
Korean-server lookahead: the tag swap, stated correctly
Korea added a tag swap on 2026-04-23, and most versions of it in circulation are wrong. It is not two gems trading tags. You pick two tag types, and every gem on your account — stored and socketed alike — has those two exchanged. Pairs compose into a single permutation within one operation, so three-way cycles are reachable, though the primitive is still a pair. 1,000,000 gold per permutation; different permutations for damage and cooldown lines means two operations and 2,000,000 gold. Hard limit: damage lines swap only with damage lines, cooldown only with cooldown.
Unlocked by a one-time quest from a bank NPC in one Korean-server town, then usable at any town bank. Korean-server only; not shipped on Taiwan, and no Taiwanese announcement has promised it. That limit cuts both ways for this class. You need only two damage tags, so converting stray damage lines into Smash is cheap — but your third line is a cooldown line, and it can only be exchanged with other cooldown lines, so moving it to Support cooldown costs a second operation.
What to do now, what to ignore, what would change this
Keep, in order: any Smash damage line — it covers every damage skill you have. Support damage second, covering roughly nine tenths including the finisher that is over half your output. Support cooldown third — the only slot on this site where we actively recommend a cooldown line. Smash cooldown is the alternative, and our two sources disagree on which of the pair comes first. Ignore Disrupt, Survival, Movement and Combo; the new passive carrying Disrupt changes nothing.
Tune against your real rotation. If you run the fringe rune that rewrites your first skill to Combo·Movement, every number above needs recomputing.
Three things would invalidate this page. If Taiwan does not ship the season buff granting skill-use speed, line three should revert to a damage line — a cooldown line only pays when cooldown is genuinely the bottleneck, and that bottleneck is a creation of the new version. If Taiwanese testing reverses Support cooldown and Smash cooldown, line three changes accordingly. And if Taiwan opens a steady engraver supply or ships the tag swap early, hoarding purely for promotion stops paying.
