Purpose of thread: To compile a working definition of what haste does for each class, and then make some conclusions about it's value.

Constraints: We will concern ourselves only with haste effects that do not impact the global cooldown (This appears to be all of them.). We will concern ourselves only with haste effects on players. (No frenzy discussions) From here forward, Haste refers to any effect which increases your real attack rate without affecting your global/internal cooldowns - be it obtained from procs/talents/static haste rating.

Disclaimer
This is my personal understanding of haste as it applies to warriors, rogues, and hunters. I am not familiar enough with casters to venture an opinion about spell haste, although I understand the basics (dots hate it) etc. I would very much appreciate a caster that feels comfortable talking about _all_ of the casters weighing in with their understanding, and will add it to the first post once it's expressed in the thread. Added below under "Haste for casters".

Haste and Procs:


Any proc without an internal cooldown (the most noticable being windfury totem) receives increasing returns from haste. However, recent testing suggests that most post-tbc procs have internal cooldowns - these items do not see above normal gains from haste, the most obvious example would be windfury weapon's 3s cooldown. (They will still have higher avg uptime as a result of haste effects, but their gains are generally discountable.) Additionally, some procs in wow have a modified chance to proc that varies as your hastes vary and your procs per hour may be impacted as a result of this interaction.


Haste for Melee:

Haste has no cap for classes who's primary attack is auto-attack. Unlike all other attacks investigated, auto-attack appears to have no internal cast time or cooldown, and thus adding more haste will always increase your actual rate of attack. Additionally, melee using instant attacks and "on next attack" specials only do not push back their swing timer. As long as you limit yourself to those type of attacks, haste provides full value (or better) for each additional point. Lastly, haste adds white swings, and can change your ratio of white:special attacks. This in turn can impact other abilities in odd ways. Note that some gear/enchanting/talent decisions (see above about procs) may make haste more or less effective for specific melee classes.

Additionally, as soon as you start branching off into effects with actual cast time, especially skills which reset your swing timer such as the warrior skill Slam, you will be reducing the effectiveness of your haste effects for similiar reasons to the hunter auto-shot pushback discussed below. Therefor 2H warriors may find other stats more appealing.

Haste for Hunters:

Haste has a cap on effectiveness as a result of auto-shot having a .5 second cast time. Sufficient haste would push a hunter to a displayed auto-shot cast time of less than .5 seconds, yet their auto-shots would only fire every .5 seconds due to their internal cast time. Thus additional haste would do nothing. Note that at this theoretical cap any non-auto-shot would clip your auto-shots, resulting in minimal dps gains. Thus for dps purposes, there is an even lower practical cap on haste. Additionally hunter auto-shot/special weaving requires some human reaction time for best effect, so leaving yourself a comfortable time between auto-shots is advisable. I've found the following rule helpful:

Attack speed after all hastes should not go below 1 + .5 + latency seconds. The 1 is to give me a second in which to fit a special attack. The .5 is the auto-shot clipping buffer, and the latency is to account for travel time to the server. Your mileage may vary, but this has worked for me.

Haste for Casters:

Haste has a cap on effectiveness for casters as a result of the global cooldown being 1.5secs. Sufficent haste would push many spells to less than 1.5 secs of cast time but they would still only be castable every 1.5 secs due to the gobal cooldown requiring a minimum of 1.5 secs between casts resulting in a net gain of little actual damage once that state of one cast per global cooldown is achieved..

However, haste will still reduce the actual time spent casting which can be advantageous when dealing with spell push back or trying to avoid interupts. Haste will also reduce the amount of time before the first cast which is also situationally advantageous when trying to move and cast at the same time or other situations that encourage narrow time windows for casting.

Channeled spells are effected by haste and will see a reduction in their time spent channeling the spell but no change in overall damage resulting in a DPS boost, additionally haste scales at the same rate for channeled spells and spells with a standard cast time.

Instant cast spells gain nothing signifigant from haste, and most DOTS gain little as well.

Closing Thoughts

Generally, haste will provide the largest real return on damage or threat output to any class up until that class reaches their personal cap on returns. For some classes, that cap is almost immediate due to talents and spell choice. Some classes may find other stats which provide equivalent returns, and should carefully weigh the advantages those other stats have vs the secondary burst potential haste provides. As an example, hit rating for warriors provides similar returns, along with a secondary benefit of stabilizing rage flow and threat flow at a cost of less burst potential.

Haste and the Warrior

Dual Wield: If your primary mode of DPS is DW then the effects of haste have no cap. As such your only concern need be balancing your haste rating versus other stats in a sensible manner, much the same as a rogue.

Two-handers: Haste has much more interesting consequences for 2h-wielders due to the mechanics of Slam. A typical Autoattack > Slam > Instant dps cycle is only theoretically maintainable for weapon speeds over 2.5 seconds due to global cooldown, however the limits of human reactions and network latency will make this cycle untenable at higher delays than this. If you're at 2.5 seconds or faster you can, of course, swap to simply spamming instants as able and only using Slam when you're out of global cooldown, have MS/BT and WW on ability cooldown and find yourself right after an autoattack. This strategy is generally less rage-efficient and will result in more time spent waiting on ability and autoattack cooldowns than a planned Slam cycle, however it may be necessary if you find yourself sufficiently hasted.

It's worth noting that as haste decreases the cast time of Slam and autoattack by equal proportions, the hidden rage and damage cost of Slam does not increase with haste.

I think this is only true in absolute terms for melee DPS.

I know for warlocks that +1 haste ~ +1 crit rating ~ .5 hit rating ~ .5 +dmg ~ .5 + shadow dmg in a destruction heavy build (with a lot of time spent spamming direct damage nukes). For affliction haste is worth about the same so it is fairly resiliant to spec differences, unlike +crit.

I'm not sure about mages or other caster DPS but any class that deal heavily in instant casts or DoTs, like shadow priests, will probably see fairly small returns to haste as well.

Haste is also helpful in dealing with lag if you are not a /stopcasting maestro.

Haste for the Mage

Well, for a Fire Mage (10/48/3 or slight variation) Haste is a pretty direct dps increase with no gain in dpm and a higher burn rate obviously. Fireball is well outside the GCD and making it go faster is better. There is a small latency penalty that makes it not scale perfectly but it is indeed an increase in overall dps.

In mana-constrained situations it is obviously no net gain though and I am very concerned that our default dps will again be balanced around haste effects, leaving us only with worse time-to-oom. Additionally, spells that are knocking on the GCD or at it (Scorch, ramped up AB and so on) gain nothing and there is no white damage being enhanced at all. Still, Fire Mages scale well with haste.

EDIT: In my testing spell haste does nothing for the GCD but if you have a small amount of haste it may be unnoticeable. True chain-casting is really quite hard to pull off. Note well that Aura mods (Bloodlust notably) are different and do indeed reduce the GCD as well. Unless they have changed the mechanics recently though, generic haste most certainly does not.

Scarab + Blessed Band = 30 + 320 = 350 haste rating, or 350/21 = 16.66% haste (21:1 conversion post 2.1).

On a 1.5 second spell, that reduces you to 1.28 seconds cast time.

If you don't /stopcast (and healers typically don't), it's entirely reasonable to believe that lag is accounting for you not hitting GCD.

edit: From what I've seen, haste *effects* (those effects that have a fixed % boost) reduce the GCD. Haste *rating* does not.

Shadowpriests can spam:

Mind Blast (1.5 casting time, CD 5.5 - 8 depend on how heavyly you specced in imp mb)
Shadow Word Death (instant, 12 sec CD)
Shadow Word Pain (intsant, DoT over 18 -21 -24 (depend on spec))
Mind Flay (channeled, 3 seconds)
Vampiric Touch (1.5 casting time, DoT over 15 sec)
Vampiric Embrace (instant, last 60 seconds)

As you can see MB suffer from the GCD problem, like scorch, but it's on a quite long CD, SWD has a long CD too, SWP usually need to be recast every 24 seconds (who would not put 2/2 in imp swp?). VT is something like MB, but on a longer "CD".

Considering a 5/5 Imp MB, 2/2 Imp SWP and, of course, 1/1 VT we have (ideal case: no latency and immediate human response):

0.00.00 VT
0.01.50 SWP
0.03.00 VE
0.04.50 SWD
0.06.00 MB
0.07.50 MF
0.10.50 MF
0.13.50 MB
0.15.00 VT
0.16.50 SWD
0.18.00 MF
0.21.00 MB
0.22.50 MF
0.25.50 SWP
0.27.00 MB
0.28.50 SWD
0.30.00 VT
0.31.50 MF
0.34.50 MB
0.36.00 MF
0.39.00 MF
0.42.00 SWD
0.43.50 MB
0.45.00 VT
0.46.50 MF
0.49.50 SWP
0.51.00 MB
0.52.50 MF
0.55.50 SWD
0.57.00 MB
0.58.50 MF
0.61.50 xxx end xxx

MF has been cast 10 times, which means a total of 30 seconds over 60 seconds fight spent channeling. For sure there are classes which benefits more than shadowpriests from spell haste, and probably +damage is still the best stat to agument their dps, but still, i wouldn't say that spell haste gives small returns to shadowpriests.


For what concern affliction warlocks, having 4 dots (2 instant, 2 with 1.5 casting time) plus an instant curse, probably 2/2 Nightfall, and some time spent using Life Tap / Dark Pact, probably the effect of spellhaste is a lot less noticable. I agree on this point.


P.S. Assumptions
1) Haste do not affect GCD (otherway MB and VT would be also positively affected, resulting in only VE, SWD and SWP not being affected by spellhaste)
2) Shadowpriest goes all out, otherway it would be more MF and less SWD/MB, meaning more time spent channeling (so more benefit from spellhaste)

For completenesses sake: Dots gain nothing from haste.

Wouldn't this be added multiplicatively, instead of additively?
I.e. 30 & 320 converted to haste percentages seperately (30/21 & 320/21) and the applied as all other hastes?

Giving the casters another day to hammer things out before I go back and revise. I'm going to try and take the "wait a few days, revise, repeat" approach to the first post. I'll make sure to actually reply with a summary of changes when the first post changes as well.

Berserking also reduces the GCD from the limited testing I did. This is to be expected given that it is a Apply Aura: Haste styled effect although the quibble would be that it is a value and not the "Apply Aura: Haste %" that Bloodlust has. The values are a percentage though so it makes sense to me. The proc off of a Mystical Skyfire Diamond appears to be the same category as well but I've never slotted one to test.