Upcoming Improvements to Blight

"

healzr wrote:

"

colossalaiur wrote:

Shattering sound explanation is just beyond me. The least you can do is the remake that soundtrack using exact same sound and ship it again. I'm not confident on changing that because the old one is just too good...

On blight, changes on undying aura and shields are good. But to improve blight maps, I really suggest that you guys play more tower defense games. There are breaks between waves, indicators of bosses or tough enemies, and building towers should be easier. Even PvZ 1 does a very good job at this.

they could even make the old sound a mtx i would be willing to pay for that as a mtx

"

dachoppa wrote:

"

healzr wrote:

"

colossalaiur wrote:

Shattering sound explanation is just beyond me. The least you can do is the remake that soundtrack using exact same sound and ship it again. I'm not confident on changing that because the old one is just too good...

On blight, changes on undying aura and shields are good. But to improve blight maps, I really suggest that you guys play more tower defense games. There are breaks between waves, indicators of bosses or tough enemies, and building towers should be easier. Even PvZ 1 does a very good job at this.

they could even make the old sound a mtx i would be willing to pay for that as a mtx

it must be so ridiculously easy to scam people, i should do that for a living. they remove a sound effect and people are willing to pay to have it back. please grow a spine and don't accept this bs.

"

Anteo12 wrote:

"

kuciol wrote:

"

Nephalim wrote:

As a whole the game has a lot of issues with massive fps drops when you hit a large pack/abyss/normal blight/breach/harbinger with a fast hitting attack or spell and there needs to eventually be a global reduction on mob size offset by increased exp gain per mob because poe is getting worse performance wise every patch for many players.

While I love instantly shatteirng 68 mobs in .1 seconds, the server and my pc does not and dropping to half my fps each time this happens is far worse than slightly reduced pack size globally.

The problem game faces is that there is just to many possibilities. Number of monsters is not the problem to a degree but the number of things game has to process. According to Jonathan there are about 9000 calculations for every single hit. They need to do something about THAT or reducing number of monster will be just band aid.

^ This 100%
For every patch in the last year since Syndicate performace got a hit. This one was a bigger hit IMHO but I don't think they are prioritizing performance as much as they should.

"

TreeOfDead wrote:

add slots more to delve rab pls!

"

No not this. 9000 calculations per hit is trivial in a modern processor. Let me just put this into perspective. Each frame of a typical first person shooter the GPU alone deals in the millions of calculations per frame.

The functions per hit are likely cached and with SIMD in modern processors most if not all of the entities that end up processing an on-hit calculation are never even leaving the processor. A typical i5 wouldn't even notice 5000 entities with 9000 calculations per hit is only 45 million calculations.

a 3GHz processor is cycling 3 BILLION times a second. Even in the old days of Single Instruction Per Cycle you'd clear 3 billion instructions a second. Even if you eat half of those cycles in overhead and other tasks that still leaves you with 1.5 billion cycles. Again to put it in perspective that whole 45 million example could be handled 3 times over... per second.

Now you have a lot of shit running on a computer yes, but 9000 instructions isn't even remotely the issue.

In a complex system like that your performance impacts are usually the result of an aggregate of issues. From extreme performance tuning I would more question IO bottlenecks rather then CPU. Go ahead and check your system load when playing a game, most with a few exceptions (I'm looking at you Dwarf Fortress) can't pin a modern CPU core if they tried.

When dealing with a lot of entities also, most are GPU clones so performance impacts are usually no-linear.

The problem is Client-Server more than anything. Netcode, syncing, all that is more than likely the problem. Each entity needs to be reported and synced between the endpoints. There, more entities means more data to line up. That falls into I\O territory.

What you aren't syncing is the calculation. One side or the other is doing that and unless the backend servers are single thread, single core, and handling all clients on that single thread and core... you'll scale just fine on the backend.

Scaling a zombie > more zombies but not as interesting. However if it a "cluster" of zombies that system wise is just a single unit but visually a group of them you would have a better time managing performance.

Animate weapon is an example. Rather then have multiple weapons treat it more like Blade Vortex. Each weapon adds to the mass of the animated weapon. Performance wise, it is just a single entity. Animating 20 weapons just means the single entity is scaled up in power.

It is those subtle differences that can give better performance. 9000 calcs per hit per mob? Not even an issue when turning in an FPS results in a million+ calculations. Now hitting 5000 mobs at once... yeah I can see that being noticeable (that is the 45 million example from above) but not detrimental.