I am posting this from the “I have already done everything you were about to tell me to do” side of the process, so we can skip that part.
ISSUE DESCRIPTION
Since Sim Update 6, the sim dies the moment I click Fly from the World Map. Four crash dumps across four sessions. The callstack is the same every time — not “similar”, not “in the same area”. Byte-for-byte identical, all 30 frames, down to the offsets:
10 = ["CoherentUIGT.pdb", ..., 559489]
11 = ["CoherentUIGT.pdb", ..., 560819]
12 = ["CoherentUIGT.pdb", ..., 544075]
Purpose = "CoherentGTImpl_Z"
LastStates = " MainState:GAME SubState:GAME_WORLDMAP MainState:WORLDMAP SubState:WORLDMAP_HOMEPAGE"
It dies in the UI layer, in the World Map to flight transition. It is currently the most reproducible thing in my install. If QA needs a deterministic repro case, I appear to have built one by accident.
FREQUENCY OF ISSUE
100%. Every single time, without exception, since SU6.
REPRODUCTION STEPS
- Launch the sim (clean install state — see below)
- Go to the World Map, pick any airport, default Cessna 172
- Click Fly
- Return to desktop
Yes, I removed the add-ons. All of them.
I know how this conversation normally goes, so I ran the clean test before posting:
| Before | Clean test | |
|---|---|---|
| Community packages | 175 | 0 (empty folder) |
| Registered packages | 1365 | 1190 |
| Developer Mode | on | off |
| WASM modules | 4 | 1 (fcr-embedded, ships with the sim) |
| Aircraft | F-35B | default C172 |
| Terrain LOD | 300 | 200 |
Empty Community folder. Developer Mode off. Default aircraft. Default airport.
Identical crash. Identical callstack. Not one byte different.
I also changed GPU driver between dumps (591.44 → 610.88). The callstack did not move. So it is not my add-ons, not my aircraft, and not my driver. I accept that the add-on question has to be asked. It has now been answered.
The UI was already coming apart before it crashed
Screenshot attached. Look at the Career tile on the main menu — torn scanlines and a block of garbage pixels where the artwork should be. That is corrupted texture data in the Coherent UI layer, rendered on screen, immediately before the crash in that same layer.
Also from that dump:
CreatedHandlesDuringACH = 14110
Up from 2 and 4 in the earlier dumps. Handle creation rising by four orders of magnitude usually means something has stopped being asked politely.
Memory
SystemMemory = 32 GB
AllocatedMem = 10.9 GB
PhysicalMem = 6.0 GB
SwapFile = 4.84 GB
4.84GB paged out to disk on a 32GB machine, sitting still on a menu, with zero add-ons installed. I was fully prepared to be told this was my 175 packages. It is not my 175 packages.
Then it took the whole machine with it
Following the standard advice, I turned DLSS off and clicked Fly. The entire PC hard-locked — no desktop, no TDR recovery, no response at all. Power button.
Here is what my UserCfg.opt still said at that moment:
AntiAliasing OFF
DLSSMode PERFORMANCE
FrameGeneration DLSSG <-- still enabled
RaytracedShadows Enabled 1
Windowed 1
Resolution 3840 2160
Frame Generation was still set to DLSSG after DLSS had been switched off. The settings menu let me select a frame generator with nothing left to generate frames from, and said nothing about it.
I am not going to pretend I know that this caused the lock-up. I am going to point out that the UI was perfectly happy to let me get there in two clicks, and that “the renderer is in a state your own config file cannot describe coherently” is not somewhere a user should be able to arrive by following common advice.
Ray-traced shadows were also still on — which the SU6 release notes describe as having received a fix for “random crashes when raytraced shadows are enabled”. Encouraging.
YOUR SETTINGS
- Developer Mode: off (and the crash is identical either way)
- Mods/add-ons: none. Community folder empty.
- GPU: RTX 4070 SUPER 12GB, driver 610.88 (also reproduced on 591.44)
- CPU / RAM: i5-13600KF / 32GB
- Build: 1.8.14.0, Steam
- Display: 3840x2160, windowed
- API: DirectX 12 — which is not a setting, since the DX11 option was removed. When the renderer misbehaves there is nothing to fall back to.
- Peripherals: nothing exotic, and none of it connected at crash time
What I actually want to say
I have been flying this series since 2020. Two years of logged time. This same install ran for about a year without a single CTD, right up until an update I did not schedule, did not request and cannot decline arrived and ended that streak the same evening.
I then spent two hours doing unpaid QA on it — reading crash dumps, bisecting packages, diffing callstack offsets, going through config files — in order to produce a clean-room reproduction of a bug in your UI layer, which I am now filing for you, for free, with a screenshot attached.
Three things would have made this evening considerably shorter:
1. A single log line naming the failing component, instead of a silent vanish to desktop. The sim clearly knows enough to write a 30-frame callstack. It could manage one name.
2. A supported way to stay on the previous build. Steam offers beta branches; plenty of titles publish a previous_version branch precisely so an update cannot become a hostage situation. There is no technical mystery here, only a decision.
3. Not shipping this one. I understand SU6 is large and that large updates carry risk. I would gently observe that it also degraded cockpit rendering on PS5 and Xbox, took out server connectivity for a wave of users, and has produced a bug board that currently reads like a changelog written backwards. At some point “it passed testing” stops being an explanation and becomes a description.
I like this sim. I have given it two years and a good deal of money. That is precisely why I am irritated rather than indifferent.
All four dumps available on request, if anyone wants to diff them. I already have.