Beamdog - Please don't forget about performance/stability!
Squidget
Member Posts: 15
Our PW is loving the updates you've done to NWN so far. The creature and placeable scaling in particular has been an enormous boon for us. However, our players have recently started to encounter more performance and stability issues, to the point where they're becoming a problem in regular play. Many are reporting 3-4 hard crashes per play session, as well as heavy lag spikes in our more traveled areas. Personal experience is that these have gotten a lot worse over the course of the last few EE patches. And it's reaching the point where it's becoming a problem for regular play.
Some of the biggest crash and slowdown issues my players have reported are:
-Frequent hard crashes and major slowdowns that seem to be tied to the Castle Exterior and Rural Exterior areas. These are definitely worse in some areas than others, but we haven't isolated a specific cause. Multiple players have reported crashes in these areas, and it seems to happen most with heavy fog. Some of the symptoms seems to be either gradual framedrops into a hard crash when transitioning, crashes on the first attack, or just long periods not receiving any server updates. We've also seen a much increased number of "Just standing around talking" hard crashes that seem to occur in every area about equally.
-Heavier lag and hard crashes around AoE spells. This seems especially bad with fog effects, but also is a factor around large AoEs like Evard's. This always caused some lag, but seems much more pronounced over the last few patches.
-Long wait times on server updates in general. Enemy positions or models not appearing to the player yet, even when they're attacking you, only to lag back into place at some later point.
Both of these seem much more pronounced in the last EE patch. I mention them here since I'm not sure if other servers are seeing the same. And it's hard to get a sense of such things through the bug tracker alone.
If there's anything I can do, or have my players do, to help track down crashing or stability issues, let us know! They're not as glamorous as new features, but still hugely important to the player experience.
Some of the biggest crash and slowdown issues my players have reported are:
-Frequent hard crashes and major slowdowns that seem to be tied to the Castle Exterior and Rural Exterior areas. These are definitely worse in some areas than others, but we haven't isolated a specific cause. Multiple players have reported crashes in these areas, and it seems to happen most with heavy fog. Some of the symptoms seems to be either gradual framedrops into a hard crash when transitioning, crashes on the first attack, or just long periods not receiving any server updates. We've also seen a much increased number of "Just standing around talking" hard crashes that seem to occur in every area about equally.
-Heavier lag and hard crashes around AoE spells. This seems especially bad with fog effects, but also is a factor around large AoEs like Evard's. This always caused some lag, but seems much more pronounced over the last few patches.
-Long wait times on server updates in general. Enemy positions or models not appearing to the player yet, even when they're attacking you, only to lag back into place at some later point.
Both of these seem much more pronounced in the last EE patch. I mention them here since I'm not sure if other servers are seeing the same. And it's hard to get a sense of such things through the bug tracker alone.
If there's anything I can do, or have my players do, to help track down crashing or stability issues, let us know! They're not as glamorous as new features, but still hugely important to the player experience.
3
Comments
Might be worth checking if that's happening in your case as well.
@JuliusBorisov - can we get a bug raised for this, please?
Also, please note that we're working on the new renderer, which will help improve the performance of the game greatly.
As for OC I thought I had some issues at one time but I have not been able to reproduce since.
This has been reported here: https://support.baldursgate.com/issues/38339
and maybe here?: https://support.baldursgate.com/issues/38162
Currently developing a PW (cormanthor.de) and even there on a small mod we experience major client crashes on a random basis.
Approximately 5 per day per player.
Major fps drops for a second and stuttering randomly aproximately every 30sec-1min.
And unsusually long loading times on maps.
Granted some issues can probably traced back to bad scripting and overbuilt maps, but again these issues also occour on smaller maps, with no scripting and singe player modules and we all are using better computers and graphic cards then back in the days so it deffinately is a client issue.
Graphic adjustments don't improve the situation at all and all experimental shaders are tuned off.
The game unfortuntaely feels very unstable and with a way worse performence than the original one.
The new renderer might help but given the recent dev blog it is up to several months till we even hear something about that. I don't think the game can live through that with this perfomence. So fixes are in dire need no matter that the new renderer is developed. In my opinion at least.
Best whishes,
Talime/Leaf
I ran a long deep-dive test with my (large) testing PW on my LAN, with 2 CD keys/players, for about 2.5 hours total on maximum/enhanced graphics settings. I let the characters sit in a large TNO01 city area for an hour, and then moved around the 250-area world for the remaining time (w/ combat). Most of these areas use either the Castle Rural or Rural (TTR01) tilesets. In the entire time, I only saw an increase of roughly 75MB per game instance, and no increase while both PC's sat in the city area. VRAM did not show any climbs except the usual/minor performance spikes, CPU and GPU usage levels remained constant (except the typical spikes when in combat etc).
I'm forced to conclude there might be some localized factor at play (e.g. a script that can cause accumulation over time, like a heartbeat). Either that, or it's something that only comes out in heavy server play w/ many players on. I gave it a good run, but just couldn't make it happen on my end.
If you can (any/all), would you please provide a simpler/ more specific case of this, that consistently reproduces? Best case a test module/PW that demonstrates the behavior (with repro steps, if any besides "have at it") and includes files we can look through once we can verify it happens for us as well? If so, that would be great. Otherwise, this is going to be very difficult to track down, unless we manage to fix it collaterally.
It also might be related to a very high placeable count (like 5000 on one 24x23 map) or custom tilesets like "medieval rural" which have a higher dip on performence.
Yet it still happens on the OOC Starting area which is only 8x8 and with less placeables.
In terms of scripting, I don't run a single onHeartbeat script other than the ones nwn uses itself.
If it helps feel free to log into cormanthor.de:5121 to experience it (To advance from teh OOC area, just talk to the NPC "Sommerwind" to complete the starting quest). Maybe it helps to figure out where performence or client issues still persist.
The module ha sonly like 10 or 20 maps yet.
Not sure how else I could provide you with more relieable information or data.
The issue is: The crashes and performence drops occour randomly and thus it hard to reproduce them forcefully.
If it would just occour on my server i'd say it is just me being to stupid to get it stabile, but several people are reporting this behaviour so it might be something else.
So if you need anything specific in terms of data or stuff, I try to provide it if you like. The PW module is not publicly accessable though and still in Alpha state maybe I can send it somehow but since it uses haks nwsync might be an easier approach to get them.
This might seem like a step backwards, but keep in mind that they were bugs in 1.69 too - just rather than crashing immediately when they happen, they likely caused some issue in another part of the system. _Usually_, this would be nothing, or maybe it would cause a strand of grass to lean the wrong way, or possibly mess up a description for an object, or not handle a click event. Or it could emit a "drink potion" event when you really didn't want it to, or it could steal your credit card data.
And I am serious here. I can make a broken custom model that will crash EE, but that will actually install a virus on your PC in 1.69.
How to handle "bad" tilesets is a different issue. If EE just decides to crash or to not render those, it breaks "backwards compatibility". But each item could be "bad" in its own way, so fixing them on the fly is prohibitively difficult. And obviously, 1.69 behavior is out of the question.
I'm sure we could umm and ahh about the complexities all year but I think most people agree that when you buy a product you expect that it is able to function in a certain way. Whether instability is introduced by patches or whether it is because now it functions properly and can't handle improper input seems more like an argument for pedants. I think it's reasonable for people to expect that they can use the product for the purpose it was built, perhaps with trouble shooting and diagnostics tools to help correct issues and so 3D models and the like can be certified by a utility program.
Then again it could be entirely unrelated to custom content and even to NWNX plugin/script functions. I don't think it's helpful to just tell people to stop using that stuff in the meantime, a solution should be found instead.
Keep in mind I'm just spitballing and wildly speculating here, because it doesn't seem like anyone really knows what's going on.
I was adding a portrait of a player to our hak packs but the portrait was strange to beginn with, it was upside down, yet for him shown norally. I still added it to the haks and after that the server crashed all the time.
Fiddleing around and readding it several time made the server stale againe but the every client crashed 4 out of 5 times on logging in.
(Funnyly exactly the issue said player had himself before)
So I removed the portrait again and things went back to normal.
Conclusion:
While I do think there are some instabilities in the game, it might be impossible for Beamdog to find out as they are related to custom content and how the game reacts to minor issues and errors.
The game seems to react more sensetive than the original one, yes, but Beamdog can't be held resposible for custom content. Especially content that was created way before NWN:EE
It is a bit sad to say that, but we probably have to live with it or using content that was explicitly created for nwn:ee and tested with it.
Best whishes,
Leaf
(I need to change my nick somehow ...mhmf)
The issue would presumably not be backwards compatibility but also forwards compatibility. If people knew what the issue was then they could update old content. The fact that it was made 15 years ago or 15 days ago has less to do with it than whether it's compliant with a hypothetical unknown change in standards.
We have figured out what is causing teh crashes, and unfortunately BeamDog can't do an awefull lot about that.
These client crashes are allways caused by corrupted, or badly made custom content that doesn't follow the rules.
It can be a placeable or a script, a head model or a weapon model.
A few examples I personally have had:
Clientcrashed because of a badly made head model that only occoured once the npc got spawned via script.
Crahes because of scaling a badly made placeable.
Crahes because of a awefully designed script.
Unfortunately CEP, Project Q and pretty much every hak pack out there will suffer from that, as their composed models do have a lot of corrupted ones in there.
The reason is: previous NWN versions were, for some reaon, a bit more tolerant in terms of errors. At least they didn't result in hard crashes but other performence issues.
That led to the problem that these issues with custom content were just overseen and never fixed or noticed. Simply because the issues weren't causing these dramatic effects instandly.
TL:DR
Client crashes are most certainly 100%ly caused by errors in old, sometimes badly made custom content that was previously not noticed by the creators.