Performance and Stability

"

Dopebear เขียน:

This has been the case for how many years now? And it only gets worse. So many words, so little change.

"

Kunaak เขียน:

At some point, you really should get rid of the "loot explosion" of useless stuff.

even the default game now has a built in loot filter.
When 99.9% of all items that ever drop in this game are ignored, and never seen, why even have it there?

That happening for a few hundred thousand game sessions (I am guessing) with thousands on monsters per map, 24 hours a day, has got to be a gigantic resource hog in calculations, all for something no one cares about.

Gigantic amounts of loot, that will never be seen, and is entirely ignored, is just a relic from a time long since gone from ARPGs.
People want good gear now, thats why the loot filters exist.

How much could it help to not have all that random caculations on worthless loot that is never even seen?

"

In Blight, there are a lot of monsters and they have a lot of life. What normally happens when you engage a pack of monsters is that your attacks quickly cull out the weak ones, leaving just a few strong targets. In Blight, however, the weak ones last for much longer. As you know, Path of Exile has a very complicated stat system with tens of thousands of different values that your character can have. Many of these are used in combat calculations and dozens of small buffs, debuffs and ailments are applied to monsters as you attack them. The more monsters there are, and the longer they survive your attacks for, the longer these calculations take. Blight combat is an order of magnitude more strenuous on our servers than regular map content.

Once we knew that we had serverside performance issues with Blighted Maps, we wanted to check whether we could improve this issue by adding more servers. We observed that on physical game servers with different numbers of game instances, the number of dropped frames per instance was not higher on more heavily-loaded servers. This showed that the problem is related to specific instances utilising an entire CPU core for more than 33ms, rather than the server itself having too many instances on it. Throwing more money/servers at the problem would not solve it.

We are working on this problem and trying to find a solution that will reduce the calculation complexity without compromising gameplay.

"

We felt it was important to post this before the weekend so that you know that the concern has been heard.