Feuerblau

Older drivers don't show that though. With an old 22.x.x driver (22.1.1 in my case) or early 23.x.x versions that overflow won't happen at all. The only problem is that newer games sometimes deny to run telling you your driver version is not supported because it's to old.. But i can work with that, while no game i play right now requires the newest version. So, seems it is entirely driver related at least in my case. While it shows also on other OpenGL related software. Not only SL, counting on other OpenGL based forum topics recently i read to pinpoint the cultrip of this issue. Sure the viewer may also has some issues. But i can't replicate your observation with certain previous driver versions. Also it happened sometimes immediatelly. Not just slowly over time. In my case the amount of VRAM surely changes up and down, that's expected. But never to an extend like i saw with the new Driver versions (16GB+ while even allocating to system memory). It dangles around 3 to 4 never going higher than that anymore. While the performance takes a slight hit, also early 23.x.x drivers work, without the performance drop from early 22.x.x drivers. While both are an immense performance boost for AMD Graphics cards. Previous that OpenGL boost my 6900XT was slower than my old 1080ti. That was abysmal.

Anyone using AMDs new cards also running into this issue: https://gyazo.com/cb37ee682549c3d3c519d5f6288ae713 (VRAM isn't the whole 16GB because other stuff needs a bit as well. So yes, the VRAM ran full entirely here with the viewer. No matter if dynamic memory is chosen or the vanilla 2GB option.) The first is the dedicated VRAM. Randomly Firestorm/SL sucks up all VRAM and starts to allocate into the shared system memory, if set. Doesn't matter what driver since last year (there was a version that ran stable though, but that one is so old now that new games refuse to run, if you use it). Fresh installation, DDU used... it happens sometimes or not. Sometimes on events, or not, time is also random. 10 Minutes? 2 Hours. More often though when you visit stuff and for some reason Firestorm/the Viewer of your choice decides to not empty up not used stuff. Using a 6900XT here. But i read that some others even with 7000s series run into that issue occasionally. (As said, it doesn't happen every time. But often enough that it gets annyoing.) Never had that issue with my 1080ti or read about that on nVidia cards so far. So either a driver issue or Firestorm+Driver issue... can't pinpoint. So i would consider twice to go with an AMD card if you use SL a lot. For short visits it may work well. But for longer sessions it runs into a memory leak it seems. Normally the VRAM hovers around 6GB. But always when the performance tanks down to a third of the normal FPS and i look the VRAM usage up... voila, VRAM ran full and the viewer needs a restart. Otherwise you get a diashow. (I talk 200+ FPS at my home down to 40...) That was my experience with SL on one of the newer AMD cards so far. I also talked to other persons in SL itself who had similar issues though with 6000 or 7000 cards. Everything else runs just fine. In terms of other games. As said, if you visit SL only on some occasions, go for them. Price > Performance is still good.

And that issue is what drives my viewer in some occasions into 16GB of VRAM till the point it allocates stuff to the shared memory using system memory. That leads to a performance loss from above 200 at home to 30FPS at home. It's annoying, doesn't happen all the time, just "random" at certain points. It also happens in any point of time. Visiting a crowded place also sometimes doesn't change the VRAM usage at all. It swings around 8GB VRAM. But when the performance loss happens, it's always above 14 up to 16GB using all VRAM it gets. And it starts with texture trashing as well before crashing the performance. (Also maybe an issue of Firestorm and the dynamic Memory usage that FS gets the available memory wrong or just screws it up stuffing more and more into VRAM.) I also had once the issue that at an event the whole PC got laggy when i zoomed in on a vendor screen, or more so the wall the vendors were mounted to, on said event. When zooming out, all back to normal. Means: My second screen while watching youtube, the videos went stuttery and laggy when zoomed in on said vendor and went back to normal when zooming out. Never had that happen before. And i got a 5950X in my system, while it was idling around at ~8% with FS active plus the youtube tabs. While my systememory doesn't matter much (64GB) i mostly ignore a high RAM usage but it doesn't change much if i remember right. But my GPU spiking to 16GB is insane and makes said VRAM leak very visible. It's either an AMD issue or a Firestorm issue. With my lack of knowledge i can't pinpoint the cultrip though. I just can remember that at a certain poin in the past, that stuff wasn't the case. And as far as i know, my old PC with a 1080ti never ran into that issue at all in the past. My thought is that AMDs "new" OpenGL fix in their drivers is a mess or at least something is off with it. At least, if it runs fine, AMD users got mostly the performance nVidia users get. But with that issue... nah.

Well, either the AMD Driver is borked (again) or something weird is going on. When visiting crowded areas the performance tanks and won't go back to normal. It stays bad, worse at it should be. I capped the performance to 60FPS without VSync (that cuts the performance in half, as it seems Firestorm uses standard VSync that cuts to 30FPS if it runs lower than 60FPS.) When i visit said places, the performance at my home (normally said 60FPS, or ~220 without cap) will go down to ~20FPS. That is weird. I tried three driver versions now and the problem still persists. When i log out and back in, performance is back to normal. It happens sporadically out of nowhere and seems to happen when certain graphical stuff is loading. Edit: Seems for some reason my GPU goes into a sort of hybernate mode. The GPU clock goes down to ~500MHz. Way to low, so the performance goes downhill of course. Weird behavior. Normally the GPU runs around ~2000MHz in SL with dips lower to 600ish. But now it doesn't go up again when it drops for some reason. Testing now an older driver version.

Well, normally Firestorm as thread uses ~10GB of dedicated VRAM. It doesn't hit the 16GB border of my GPU though, when it starts to slow down. Will test another driver. The one before the actual had not that issue.

Someone else experiences that behavior: At some Moment Firestorm takes 14.5GB of dedicated VRAM of 16 my Card can provide. The Viewer starts to respond slowly when it happens. From one Moment to the other. Will try another Graphics driver (i bet it's the issue because i updated it recently and it may fall to the same date). Very odd. A restart solves the problem. It does happen when specific textures get loaded it seems, what textures i can't tell so far. Graphics card is an AMD RX 6900XT. I know that AMD fixed their OpenGL performance issue. So i guess it's something around that topic.

Seems it was indeed the driver itself. The new version at least for me, works just fine so far. (Version 23.2.2)

I changed the drive and so far all is fine. Will see how it runs with higher populated regions. Won't change that quickly. The card isn't that old and paying another grand just for SL me not likely. It seems to run properly now.

Recently... or well at least 2-3 months now Firestorm keeps freezing for several seconds up to a minute sometimes (and if to long looses connection), mostly it handles the freeze and comes back. I've seen it happens when loading certain textures, at least that SEEMS to be the problem. When i changed texture layers like BoM stuff it happen regularly. But also not every texture. For example when at a club, all runs fine, then it starts to freeze for a second a few times and the big freeze happens. Happens more often when there are more people. Because more textures are loading in. And at some point it seems to "much" or some stuff isn't to my viewers delight. What i've done: Firestorm clean installation. Deleted everything, config files, cache, the program itself of course. Exceptions for firewall, app and folders. Exceptions for Antivir software, app and folders. (Both are windows own security app, no third party stuff and others were never installed). Connection WAS unstable in the upload. Provider fixed that recently. Now upload AND ping is rock stable. Tested with Plotter and other programs. (Modulation problem from the node because another houses connection interfered with the whole node network where i live.) Changed decoding set from 0 to 1. (No effect whatsoever) Tested the RAM, no errors. What i didn't do: New windows installation. I want avoid that and others had no fix after it as well. Change the GPU driver so far, i think (AMD had issues with the 6000 GPU series so i was avoiding an update.) Using another viewer, what i will test. What i've noticed? When i only look at me, not the crowd it seems not to happen or not as frequent. Texture streaming problem? Asynchronous image decoding with high multi core CPUs? My hardware: AMD Ryzen 5950X AMD RX 6900XT 64GB RAM If more infos needed, lemme know. Hardware shouldn't be the problem, performance wise. AMD seems also to have fixed their OpenGL performance last year. (doubled it to be specific) Will try another driver version now. I know SL isn't any game. But so far had no problems with other 3D software so far. And the CPU should be more than enough to handle SL properly. Except Firestorm got issues with handling AMDs multi core CPUs. Who knows... wouldn't be the first time stuff gets not optimized for AMDs stuff... The problem also wasn't present end last year. It appeared the last 3 months or so. It seems that some specific stuff freezes the viewer or lets it get stuck in some workload, so it halts the entire process and it disconnects/looses connection due to no answer towards the server anymore. It does NOT crash. It disconnects when the freeze is to long. NO crash, that's important. It literally gives me the disconnect screen and i can quit the viewer normally. The problem was before the unstable connection and after it. Didn't changed anything btw. So either texture streaming from my viewer towareds AWS Servers got some weird issue or something local with my Viewer together with something keeps it stalling on some texture processing. PS: Grammer issues please let me know. Not native english speaking person here. PPS: Seems a driver issue. Testing different versions now. Fun...

It does. The original is still there, in some form. But it is not really attachable. When i attach it, in the original folder it does not attach (bold letters). But in the appearance Window there the outfits are saved, it shows as worn. That's some quite strange behavior. But in reality the shape i worn before switching to it for example is still active. Like, when wearing shape 1 and shape 2 is the broken one. I attach shape 2, it shows in the outfits as bold and wearing, but still shape 1 is active. In the inventory and current outfit though it shows not in bold letters so not active. I hope LL can restore it somehow. I can't believe that it is gone from their servers. I bet it's some strange "synchronisation" problem due to that weird disconnect>loading> disconnect issue. (Like pulling power cord from a booting PC like.) Who knows. That issue is quite worrysome to be honest. It can happen everytime again for everyone. Not want to sound panicked but it IS something people should know can happen. I missed to make a backup for that shape unfortunately. And that's what bothers me the most. Normally i do exactly that.

I tried everything now. Cache clearance, other viewers, copy/paste, XML export (it exported the shape i was using before of course). It shows the shape as wearing in the appearance window but not when i look at the original. So the LINK gest activated but it can't find the file obviously. Tried the Standard SL viewer and Blackdragon with clean install. Deleted everything in the cache folder manually. It's like the shape file was totally blown from the servers itself. Like... literally obliterated. I should get an award for that... *laughs with frustation*

May try that. But i tried it with another two viewers. Lumiya and Dragon. Both had a totally empty cache. Same error message on Dragon though. So my assumption was that it was really just destroyed. Will test it with firestorm. Thank you. I also wrote a ticket. But i guess they are not able to fix that from their side to be honest. *sigh*

Not a new issue, i read about it. Problem is/was: I wanted to teleport from Doux to my home. Got stuck in teleport. Abort teleport and quit the viewer because after such a failed teleport i always disconnect in a few seconds. Tried relogging, got stuck in "connect to region". Quit the viewer/disconnected (Closed window, no taskmanager kill). Tried a second time, login worked again. Got the message that my shape was "missing from database". The all known error message. Now my question. It seems the file is corrupted and its gone to nirvana, from what i've read. A unique selfmade shape i'm not able to redo and wanted to make an backup the next day (Karma is a b...). Or is there ANY solution that could "rejoin" the file/link from my viewer to the file on SLs servers? Or is that file on the SL servers just "poof" and gone for ever, destroyed by that unlucky disconnect, reconnect, region teleport error stuff? I'm frustrated... very angry, disappointed... (I know, there are things who are worse. But i worked on that shape for a long long time to get it how i want it.) Thanks for replies. *takes a deep breath*