Vram leak prolly

dumping shortlog here with no further context. Looks like I ran out of VRAM. Game was running for ~11 hours, and I was going in and out of nether portals around this time. Nvidia 3070Ti (8GB VRAM), 8GB XmX G1GC, Windows 11 23H2. DH 2.2.0a with balanced cpu preset, no other setting changes. Embeddium 0.3.30+1.20.1 / Oculus 1.6.15a, no shaders or resource packs (Modpack is Gregtech New Beginnings, which uses Create, True Darkness and BiomesoPlenty).
39 Replies
Puhpine
Puhpine•2mo ago
im suprised that you are asking about a vram issue, and not that eveyting looks horrible, because oculus 1.6.15a does not have shader support what mc version are you on?
noruzenchi86
noruzenchi86OP•2mo ago
1.20.1
Puhpine
Puhpine•2mo ago
then you should either use DH 2.1.2 with oculus 1.7, or use DH 2.2.1 with an oculus version that is only available on their github for now. as a memory leak was fixed in 2.2.0, i recommend the latter
noruzenchi86
noruzenchi86OP•2mo ago
i remember having some similar issue a couple months ago but it was definitely with physical RAM and DH 2.1.2. Create blocks didn't show up (were invisible) when using Oculus 1.7.0. Hopefully they have fixed it by now?
Puhpine
Puhpine•2mo ago
maybe, i dont know about that if you cant use a oculus version higher than 1.6.15a, its better to remove either oculus or DH. the vram leak may or may not be because of those two mods not supporting each other, its hard to say.
noruzenchi86
noruzenchi86OP•2mo ago
Hmm. Last time Oculus's github updated was 7 months ago. Maybe something will happen if I use Iris's forge build instead.
Puhpine
Puhpine•2mo ago
youre probably looking at the wrong branch
noruzenchi86
noruzenchi86OP•2mo ago
oh, yeah. thats it 1.16.5 is marked as the default branch on their git so 1.20.1-new
Puhpine
Puhpine•2mo ago
yea
noruzenchi86
noruzenchi86OP•2mo ago
I won't touch the Oculus branch as I'm not about to install an IDE to build a .jar file But apparently latest Iris+Embeddium+DH doesn't crash so let's see how it is ingame! yep, all good here matter of time for testing now
Puhpine
Puhpine•2mo ago
you can download the build artifact. no need to compile yourself youre switching to 1.21.1?
noruzenchi86
noruzenchi86OP•2mo ago
no, i'm still on 1.20.1 I think the modpack would die instantly if I tried
Puhpine
Puhpine•2mo ago
oh, then what DH are you using?
noruzenchi86
noruzenchi86OP•2mo ago
DH 2.2.1 (should have specified *latest for this version lol)
Puhpine
Puhpine•2mo ago
huh weird, that shouldnt work with latest iris
noruzenchi86
noruzenchi86OP•2mo ago
create's good
Puhpine
Puhpine•2mo ago
what is the exact iris version?
noruzenchi86
noruzenchi86OP•2mo ago
"1.7.5+mc1.20.1" via CurseForge, Prism Launcher
Puhpine
Puhpine•2mo ago
oh youre switching to fabric now?
noruzenchi86
noruzenchi86OP•2mo ago
not really I think whatever the iris devs have done somehow enables it to work on forge perfectly fine
Puhpine
Puhpine•2mo ago
then it doubt its actually working
noruzenchi86
noruzenchi86OP•2mo ago
or the modpack does, no idea
Puhpine
Puhpine•2mo ago
no can you share a screenshot in game? with f3 open?
noruzenchi86
noruzenchi86OP•2mo ago
No description
noruzenchi86
noruzenchi86OP•2mo ago
gonna guess it's not enabled then
Puhpine
Puhpine•2mo ago
so iris is not loading yea forge just skips loading it because its not for forge go here
Puhpine
Puhpine•2mo ago
Puhpine
Puhpine•2mo ago
and click build artifacts then open the zip
noruzenchi86
noruzenchi86OP•2mo ago
Ok, thanks. I immediately look to the right of any github page to see if there's any releases/packages, did not know this was a feature of Github
Puhpine
Puhpine•2mo ago
had to search for a while as well 😅
noruzenchi86
noruzenchi86OP•2mo ago
[Oculus] Version: 1.7.0
No description
Puhpine
Puhpine•2mo ago
looks like its working, no? shaders are just disabled
noruzenchi86
noruzenchi86OP•2mo ago
looks like
Puhpine
Puhpine•2mo ago
now lets hope the vram issue doesnt come back
noruzenchi86
noruzenchi86OP•2mo ago
hmm. gonna try just not running DH for a bit and see if it goes away if it helps I'm on nvidia driver version 565.90 for win11 might be related to chromium somehow too, since whenever minecraft crashes with the OP error, it sometimes takes my browser out with it journeymap interaction? or journey tries to eat memory that's in use by DH. who knows got Crashpad_NotConnectedToHandler error in chromium
noruzenchi86
noruzenchi86OP•2mo ago
hol up
No description
noruzenchi86
noruzenchi86OP•2mo ago
11.6GB? why is it allocating 3 and a half more GB physical ram than XmX istg if i have to reinstall windows literally the only thing preventing me from going linux fulltime is meta/oculus VR ok cool, last crash also corrupted ftbchunks map cache
java.lang.NullPointerException: Cannot invoke "com.mojang.blaze3d.platform.NativeImage.m_84988_(int, int, int)" because "this.renderedMapImage" is null
at TRANSFORMER/[email protected]/dev.ftb.mods.ftbchunks.client.map.MapRegion.setRenderedMapImageRGBA(MapRegion.java:205)
java.lang.NullPointerException: Cannot invoke "com.mojang.blaze3d.platform.NativeImage.m_84988_(int, int, int)" because "this.renderedMapImage" is null
at TRANSFORMER/[email protected]/dev.ftb.mods.ftbchunks.client.map.MapRegion.setRenderedMapImageRGBA(MapRegion.java:205)
thanks minecrap think its become clear this is probably not a DH issue
Puhpine
Puhpine•2mo ago
Because java has different memory systems, and you allocated 8 GB to one of them. That's not a windows problem, that's just how java works Yea that does not look like something dh would cause
noruzenchi86
noruzenchi86OP•2mo ago
best guess is something to do with adoptium vs oracle JDK, DH's 2.2.1 server interaction, or the threading / viewdistance settings (default DH viewdistance at 128, still has issues lowered down to 96 chunks and a vanilla distance of 8) either way, I disabled DH on my client, and my local server that never had DH installed is running well under 4GB, where before it was topping out above 9GB on an 8XmX instance
[23Oct2024 00:49:24.691] [DH-LOD Update Propagator Thread[1]/WARN] [DistantHorizons-com.seibel.distanthorizons.core.sql.repo.AbstractDhRepo/]: Attempted to insert [FullDataSourceV2DTO] with primary key [6*1,1] on closed repo [jdbc:sqlite:C:\Users\Shadow86\AppData\Roaming\PrismLauncher\instances\GTNB 3.3\minecraft\Distant_Horizons_server_data\69%2E49%2E85%2E3525565\overworld\DistantHorizons.sqlite].
[23Oct2024 00:49:24.691] [DH-LOD Update Propagator Thread[1]/WARN] [DistantHorizons-com.seibel.distanthorizons.core.sql.repo.AbstractDhRepo/]: Attempted to insert [FullDataSourceV2DTO] with primary key [6*1,1] on closed repo [jdbc:sqlite:C:\Users\Shadow86\AppData\Roaming\PrismLauncher\instances\GTNB 3.3\minecraft\Distant_Horizons_server_data\69%2E49%2E85%2E3525565\overworld\DistantHorizons.sqlite].
might have something to do with it this gets spammed when /backing after dying, or changing dimensions often. -# yes the IP is exposed i don't care either way, it hits max RAM usage very quickly this way and after setting all threads to 1 with 90% runtime except for the LOD builder threads which get 8 because of the max queue limit warning it seems like distant generation is trying to occur way too frequently for chunks that haven't changed, or something isn't being released properly no way to do anything about it until the mods I'm using updates to 1.21, (Create…)
Want results from more Discord servers?
Add your server