Performance and Stability

Sorry Chris,

you can´t blame Nvidia or AMD for the crashes, it´s your companies fault. At the moment you went from DX9 to DX11, the game became buggy.

Mind you, I saw the same thing happen to many games, DX11 is just a terrible API to code games on. If you can, go for Vulkan, it´s the best API you can have these days, it runs on all systems and with all GPU´s. It´s fast and reliable.

Dx12 has the problem, that it doesn´t run on WIn 7 and I believe Linus has issues there too.


Something else is, that you store all the data in a 25GB archive, which is just not a good idea. You should have one Data file for all the normal content and then one each for the league specific assets. So the loading times will improve dramatically and the HDD´s no longer have to open the whole archive and look for the specific asset / load it all into RAM.

I believe the Blizzard guys run WOW like that, it´s also much easier for patching, as you don´t need to patch the 25GB file, but the data file which contains the content you change.

Honestly, it´s not that hard m8, just do it!

TL;DR
Game so complicated game server "can't even".

"

Nephalim เขียน:

"

It may seem arbitrary and defensive for us to point out the difference between the engine and the gameplay systems, but the reason is that the engine is going to be part of Path of Exile forever, but the gameplay code comes and goes as things are added and removed. We don't want leagues to have performance issues, and this informs the decisions we make about future content development.

No one with any common sense is asking for an engine rebuild from the ground up. What we do want is an option to reduce or turn off particle effects particular including the frozen ground from ice shot, literally every type of 'burning ground mod', gore from belly of the beast and of course a global reduction of mob size offset by increased exp gain per mob. Nuking 60 ones in one click is great and fun but there is a threshold where that fun is offset by trememendous drops in fps.

This fps drop occurs in nearly every mechanic the game has including abyss, harbinger, densely popuated breaches, beyond, and of course blight. In order to play end game players are forced to stack all these mechanics on top of each on top of high pack size maps and multi sextants that also add mobs. Effectively, the higher up you go in poe the worse and worse your performance gets and that's horrible.

I have likely donated 1000 or more usd to this game and will continue to do so but being able to play this game at a steady 100-120 fps is crucially important to me and I know it can be done or at the very least be improved over what it is right now. I've spent even more on my pc and it is still not enough to get steady 100-120 fps.

This is exactly my story.
I can't play with a steady 142 fps on a 2080 and a i7-5930x :(

"

Ygidua เขียน:

Sorry Chris,

you can´t blame Nvidia or AMD for the crashes, it´s your companies fault. At the moment you went from DX9 to DX11, the game became buggy.

Mind you, I saw the same thing happen to many games, DX11 is just a terrible API to code games on. If you can, go for Vulkan, it´s the best API you can have these days, it runs on all systems and with all GPU´s. It´s fast and reliable.

Dx12 has the problem, that it doesn´t run on WIn 7 and I believe Linus has issues there too.


Something else is, that you store all the data in a 25GB archive, which is just not a good idea. You should have one Data file for all the normal content and then one each for the league specific assets. So the loading times will improve dramatically and the HDD´s no longer have to open the whole archive and look for the specific asset / load it all into RAM.

I believe the Blizzard guys run WOW like that, it´s also much easier for patching, as you don´t need to patch the 25GB file, but the data file which contains the content you change.

Honestly, it´s not that hard m8, just do it!

That sounds a bit condescending.
I did some very basic stuff in Vulcan, and that wasn't "not that hard".
Judging from my limited experience, i'd guess you'd basically have to redo _everything_ from scratch. And its something you can't just "do" in parallel to normal development, since you need to add features to the client while doing the rewrite, resulting in more changes for the rewrite, ad nauseam.
If you want to do this right, you need quite some time without new features to the old client, plus an army of experienced vulcan-devs, and (unlike people stating "it's not that hard, just do it") they are fucking hard to find.

"

harter_morgoth เขียน:

plus an army of experienced vulcan-devs, and (unlike people stating "it's not that hard, just do it") they are fucking hard to find.

Especially when you auto-limit your hiring to AUS/NZ only.

I know I'm one of the users who's brought up performance issues.

Based on an offhand comment on the patch thread, I did a little research, and it came to my attention that POE was apparently not detecting my Nvidia Whatsis Quadro Thing..."K1100M" is the designation that stuck in my mind...graphics card.

Having previously determined that the apparent godawful latency spikes that 3.8 had introduced seemed to get worse the longer I played at a time, and seemed to be significantly relieved by quitting and restarting the program, suggesting a memory/performance issue instead, I then spent two and a half hours downloading, installing, and reinstalling various versions of the Nvidia driver, trying various tricks like the "secondary display" one linked in most of the threads that come up when one googles "make POE use 3d video card" and the like (works to enable selecting the K1100M rather than Intel HD Graphics in the dropdown, but then the game tries to play on the nonexistent "second display" in a forced low resolution), disabling the Intel HD Graphics device (requires a forced reboot of the computer and caused POE to immediately crash with an error about vectors and T scripts), and dealing with Nvidia's website lying through its teeth (please pass on that they should remove the teeth in the next update) about which hardware models given driver versions supported. I finally found a solution:

Upon finally managing to download a combination of Nvidia driver component updates (I think direct from the Dell website) which 1) did not Fail to Install and 2) resulted in an installation of the Nvidia Control Panel program which didn't immediately crash on launch, I found an option within the Control Panel to set a Preferred Graphics Adapter on both a global and a per-program basis. Setting the Global to my 3d video card resulted in it, rather than Intel HD Graphics, being the only option in the grayed-out POE options dropdown, and my performance problems pretty much disappearing. Including the previous-version one where the graphics and game speed would degrade rapidly when the computer got too hot (in spite of the fan cooling pad!), until I'd run my desk fan pointed at the keyboard for a minute or two.

So, the takehome is:

1) you might want to focus on fixing that bug where POE won't detect a 3D video card and bends over backwards to use the sucky on-board graphics, to the extent that's a POE thing and not a Windows thing, though google results imply at least some of both, and

2) somehow Grinding Gear's code, in a new thing as of 3.8, causes POE, when it's trying to use the onboard graphics and the onboard graphics are choking, to behave EXACTLY the way it would if server latency had suddenly increased by a factor of 300.

...how did you manage that? o.o

"

Navandis เขียน:

You guys dug yourself into a hole and every league you dig it deeper. Unless there's a critical shift in your approach all the progress you made in the past 2-3 years will be rolled back and you'll extinguish this game back to a niche title.

This. For what it's worth, I am really enjoying this league, and I enjoyed Legion as well. I don't seem to be having as many game-breaking problems as others, but it looks like I'm in the minority, which is not a good thing. I feel like I'm witnessing POE on the path to its crash and burn. I don't want to see that happen. Please, please consider breaking from the current model of releasing a new league every three months, and take some time to fix the game. There are many of us who will stick around and wait, and it'd be truly disappointing to see this game go to shit.

I also agree with many others that the Blight mobs coming in waves would be super helpful. It's actually bonkers that there are so many mobs at once that -- if you haven't lagged and crashed -- you can't even see what you're killing. :/ Quality over quantity!

"

harter_morgoth เขียน:

"

Ygidua เขียน:

Sorry Chris,

you can´t blame Nvidia or AMD for the crashes, it´s your companies fault. At the moment you went from DX9 to DX11, the game became buggy.

Mind you, I saw the same thing happen to many games, DX11 is just a terrible API to code games on. If you can, go for Vulkan, it´s the best API you can have these days, it runs on all systems and with all GPU´s. It´s fast and reliable.

Dx12 has the problem, that it doesn´t run on WIn 7 and I believe Linus has issues there too.


Something else is, that you store all the data in a 25GB archive, which is just not a good idea. You should have one Data file for all the normal content and then one each for the league specific assets. So the loading times will improve dramatically and the HDD´s no longer have to open the whole archive and look for the specific asset / load it all into RAM.

I believe the Blizzard guys run WOW like that, it´s also much easier for patching, as you don´t need to patch the 25GB file, but the data file which contains the content you change.

Honestly, it´s not that hard m8, just do it!

That sounds a bit condescending.
I did some very basic stuff in Vulcan, and that wasn't "not that hard".
Judging from my limited experience, i'd guess you'd basically have to redo _everything_ from scratch. And its something you can't just "do" in parallel to normal development, since you need to add features to the client while doing the rewrite, resulting in more changes for the rewrite, ad nauseam.
If you want to do this right, you need quite some time without new features to the old client, plus an army of experienced vulcan-devs, and (unlike people stating "it's not that hard, just do it") they are fucking hard to find.

The biggest obstacle might be that the "new guys" need to understand the "old guys" code and this makes it brutally hard.

To work on a code of someone else, is what makes coding so difficult in my books. Especially since most coders these days, are not very "kind" with making it easy to grasp for those that follow.

You can be amazing in your job, yet terrible in a precessor.

That being said,

GGG has expanded so much, they went to asia, brasil, russian markets, they ported the game to consoles... that´s a huge amount of ressources spent on a niche market.
Therefore, I would expect GGG to invest heavy into the main market, which is game performance. They are not a small indie company with 10 workers anymore, they are a multi million dollar company, they just need to prioritize this.


Ofc. NZ does have very strict rules if it comes to hiring workers, I made my own experiences with that. If we now factor in that people in NZ might be anything but hugely invested into computing and software, it can be quite difficult to find someone who can actually understand the old code and rewrite it properly. That´s indeed something not in GGG´s hands, but they could surely move to a new country, don´t they also have parts of their buisness in the US or what about Korea, they got good PC nerds there. :D

I do appreciate it! Thx GGG

"

Engine vs Gameplay Performance
Recently there have been a lot of community posts criticising Path of Exile's engine for being slow or not optimised enough. We have spent several years improving the engine and have deployed these updates as soon as they have become ready. Path of Exile's engine is currently modern, lean and fast. But yet, in many situations, the game does not perform well at all.

Almost all of the performance problems that Path of Exile suffers from are related to gameplay systems and decisions we have made there. For the reasons described above, Blight is a league with performance problems. We made a conscious decision to try this type of league, and it's entirely appropriate to blame us and this decision for the poor performance that you may encounter while playing this content. We weren't expecting it to be this bad, and are working on fixing it.

No matter what engine we are using, Path of Exile would be slow in its current state due to the gameplay decisions we have made: calculating too many things on the server and rendering too many things on the client.

It may seem arbitrary and defensive for us to point out the difference between the engine and the gameplay systems, but the reason is that the engine is going to be part of Path of Exile forever, but the gameplay code comes and goes as things are added and removed. We don't want leagues to have performance issues, and this informs the decisions we make about future content development.

So the engine is lean and well designed, but you are using it poorly. Any tool, no matter how well designed will break if abused.

Many players will understand why, good business decisions aren't always the right decisions, but they keep the doors open. Keep mtx selling, keep new leagues coming.

Nokia, Kodak, Sears.....

It's time for you to make a bad business decision that may hit revenue but improve product quality, so that you have an actual product to sell. One more league like this and players will stop not because they want to, but because the game will simply be unplayable.

Synthesis did it for some. Legion did it for me. Blight will do it for many more.