Flight Simulator surely has a memory leak. Tonight I flew for a few hours with @FlyingCookie650’s Thursdays weekly event and after that I decided to do a few Landing Challenges.
After a while I started to get stutters. I lowered the Global Resolution and still continued to have stutters.
I started task manager to see what was going on and this is what I saw. It had 19.6GB active and 64.2GB of Committed SRAM. Even though I have 64GB of SRAM, we are now going into paging and Flight Simulator at this stage, when I’m just doing Landing Challenges, should not have 64.2GB of committed SRAM. Besides FSXX I also have other things running in that background that require SRAM.
This is definitely a memory leak and something that hopefully gets fixed in the next update.
omg… such high amount of memory allocation I’ve never seen. Can you also summary when that happens ( only in the Landing Challenges ?, scenery, location, airplane model ) ?
A memory leak is it in case these allocated memory will never be released. I seen commited usage of 40-50 , but only short time over cities like London and if I leave that areas, the memory becomes free.
That the memory usage made trouble, started earlier… it comes more and more on top with the 1.12… for users with 16GIG RAM and with 1.14… also the users with 32GIG RAM report issues ( need increase pagefile ).
Reducing Terrain detail to 10 for a while usually gets the memory to go back down, but you can’t increase it back over 200 without leaving the sim. Not an option if you fly with LOD over 2.0.
Pausing helps as well. Go make a cup of tea, have a sandwich, come back 15 minutes later and see if it has released all that memory.
Reducing sim rate helps with the memory blow up, Increasing sim rate makes it worse. Movement over the map is a big factor in something falling behind and not letting it go when you’re already out of range.
I first encountered that when I still used the rolling cache, flying over PG areas was fine, leaving them caused slow down. All that PG data stayed in memory and slowly (too slow) got written to the rolling cache. No rolling cache solved the slowdown and memory increasing.
However since the last update, whatever cleaning up process there is, seems to have gotten a much lower priority. Leaving stuff in memory that should already be gone. It should however get released when you go back to the world map, or maybe it still hangs on to it until you exit the sim. Exiting can take a long time when it gets in such a high memory use state. (Since I don’t use rolling cache I simply end process now, logbook won’t get updated but that’s a mess anyway. Windows recovers the memory instantly, restart sim fresh)
I was under the impression that it was slowing down once all the texture had finished loading. Once all textures had loaded, the GPU had more to process. Following that reasoning, I would think the textures would load faster with rolling cache; hence the GPU load would increase faster and the sim would slow down sooner…
But I have no idea what’s going on.
Another thought is the VRAM is running out, due to a lack of software optimization, and the bottleneck becomes system ram.
I have a 3070 with 8GB of ram.
Flights are flawless in most areas, but the wooded area near Manhattan brings FPS to low 20s
It depends whether you are CPU or GPU limited. Most of the time the CPU is the bottleneck. I can run at 150 render scale without losing any fps compared to 100 render scale.
Every time I turned on dev mode when this memory problem came up, vram was never an issue, at least not according to the game.
It does happen sooner in heavily wooded areas. However that is not textures or all that much work for the GPU, it is a lot of work for the terrain generator. It has to populate each tile with trees according to where trees are supposed to be. Perhaps something slows down when cleaning up the trees again.
Maybe volumetric clouds are another resource hog that doesn’t get cleaned up fast enough behind you. I’ve had the problem over open ocean, everything on low or off, just live weather, flying through clouds and fog.
The way the terrain is prepared can use a lot of optimization. I made this video 10 days ago, jumping back to the drone camera from far away (left it sitting there) to see how MSFS loads terrain
I just got prompted for the 1.14.6.0 update again… I had already installed it when it came out…
65 mb download.
Performance in the same wooded area seems better
Update:
Spoke too soon, performance is the same, not sure what that was
It looks like a memory leak to me as well, the free memory dropped as the stuttering commenced.
Check out the hard faults/ sec graph
" Hard faults often occur when a memory block has to swap to the virtual memory (hard drive or SSD) instead of physical memory.
It’s a normal part of the computer is processing the memory information, instead of an issue of the quality or brand of memory. However, when you are getting a massive amount of hard faults/sec, it indicates that your computer is in question because of too little memory."
Dont regard that, I hacked the net and sent everyone a little surprise … Oh wait April Fools day has passed, Bwhahahah
Yeah I got it to, started a a320 flight started at 20fps … set it on course came back an hour later 1 fps and the YOU DONT HAVE THE BANDWIDTH FOR PG dialog box sitting there (didnt even have PG turned on), I got 100mbps connection, guess I better call my provider and tell them I want my money back, since Im not getting the bandwidth promised…
Probably the wrong thread, but I left off in the rain this morning, resumed from the same place half an hour ago and found it rather odd there was not a single cloud in the sky anymore.
The weather forecast says it’s overcast, crystal blue skies in the game. The wind is correct but where did the clouds go. Does this happen more often? I normally never check the real forecast, but now found it rather weird how it went from heavy rain to not a single cloud on the horizon.
I restarted the sim from desktop, same result, not a single cloud, only the wind.
There seems to be issues with the servers at the moment (again cough cough) and that might be related to what your experiencing. Im not ■■■■ about how acurate my weather is, and its been close enough to keep me happy for the most part.
Could this also be why my frame rate tanks after I use time acceleration for more than a minute?
The game runs perfectly fine before I time accelerate, but if I let it go for a minute then go back to 1x mode, the frame rate tanks and never recovers.
(I’ve been using time acceleration since launch in every flight, and there is a short 2-6 second period where the framerate drops as it streams in landscape detail data when I got back to 1x. This is different. It drops down to 10-15 fps when I got back to 1x now and never goes back up even after flying for 10 minutes)
This is the first time that I have encountered this memory leak. I was flying a Beechcraft Bonanza G36.
I will do some testing. A few differences from Thursday’s Flights come to mind and I will test out those scenarios. The Flight Plan I used is the one from Thursday April 8, 19:00 Event Calendar.
We normally fly in more remote areas. Thursday we departed from Santa Barbara and flew to Bear Lake. We landed in LAX and Catalina island and a few other places in California. So we flew over more PG areas. I will re-fly the route.
I never slew and on Thursday I was I had a bad approach and decide to go around for a better approach and that put me behind the group. There was a second time I fell behind and slewed a second time. Some have reported that they fee the slewing over PG areas causes the memory leak. I will check that scenario out.
After our group flight a did three different Landing Challenges, but repeated each several times trying to improve my score. I will test if repeating the challenge over and over indices the memory leak.
About 2/3 into the group flight, my AP became inoperative. If activated, the plane would do a violent turn to the left and a nose dive. So based on that I think the problems are more related to PG areas or Slew.
Also about 2/3 into the flight I started to get a lot of stutters.
I will try to narrow it down today and report back my findings.
Test Results: Only Test 1 for now.
Test1: Run same Flight Plan and landings and take offs from same airports. No Slew used or sim rate increase in this test.
Running with Global Medium and locked at 30FPS. I do tun with Rolling Cache of 64GB, so don’t know if that could have alter results.
When I start at Santa Barbara FS had 13.8GB Commit Size (CM) of RAM. I sat at Santa Barbara for a while before take off. Just doing other stuff. Took a memory screen shot right after I load the sim and after coming back from doing stuff, the memory had a relative big increase without doing anything. At one point got up to 16GB CM.
Still have spikes around Santa Barbara.
After Landing at KLAX, CM at 18.5GB. Started to get some stutters on approach. CM at 18.4GB after takeoff and over water heading to Catalina Island. CM dropped to 16.1GB about halfway to Catalina Island. Landed at Catalina with 15GB CM.
CM at 13.8GB halfway from Catalina to KSNA to land. Started to increase as I got closer to KSNA. Landed at KSNA with 17GB CM.
Flew in from KSNA into KPOC and after landing had 16.8Gb CM. Saw and took phot of spike that we are still getting.
From KPOC to KWHP CM about 16.2GB. 17.2GB after landing at KWHP.
From KWHP to 46CN about 15.9GB CM. Landed at 46CN with 15.4Gb CM.
From 46CN to L26 about 15.8GB CM. Landed at L26 with 16Gb CM.
From L26 to L36 (Bear Lake) about 15.8GB CM. Landed at L36 with 16.2 CM and on exit back to Main Menu was at 15.3 CM.
At Main Menu 11Gb CM. So for this flight, there was no memory leak.
A memory leak is when a program continues to request more memory than what the program actually needs or uses. It is caused by poor programming.
Since WU3, many user report high or very high memory usage usually with lower FPS and stutters. This is caused by MSFS reading into memory lots of detailed scenery and objects usually due to high/ultra and/or high LOD settings. This might look like a memory leak but it really isn’t. When MSFS finishes reading and loading into memory all the scenery files, MSFS and Windows cleans out unused memory and the FPS returns to normal. However, the user has to ESC to the Pause Menu for a few minutes. When flying resumes and moves to another area, this provlem of low FPS returns.
The workaround for this problem is to temporarily reduce LOD and Ultra graphics settings enough to keep FPS from dropping. The permanent fix has to come from Asobo. Part of this fix was in the most recent update.
So, I guess you don’t consider this a Memory Leak, and what I had Thursday night, and had never encountered before and trying to determine what caused it.
I saw your original post. I too have had MSFS committed memory over 50gb in my 32gb system. Task Manager doesn’t and cannot show WHY so much memory is allocated. Memory leaks are bad because they eventually cause memory exhaustion causing Windows to stop or the system because Windows is moving too many pages between real and virtual storage that nothing else can run. Memory leaks don’t go away when a flight finishes or a new flight started. Only exiting MSFS and system reboot can stop and clean up a memory leak.
let us not so focus on the term “memory leak”, let us say “memory issue” … else we lost the focus.
In my opinion the memory management of MSFS must be optimized / changed. I mentioned this alread in other thread : an application should not “think” it can allocate unlimited memory. Well optimized applications stop by around 80% of existing physical RAM. The users with 64GIG are already on a the good side, but what should users do with 16GIG RAM - have a 60GIG pagefile ? This can never be an option and these users risc each day that here SSD will fail. In case some features are not possible with the existing amount of RAM, the user can simple not select it or see the amount of details or whatever.
We already find out that memory consume depends a lot from settings like LOD and that MSFS normaly also release former allocated memory if no longer used… also the high ranges around 50gig worked normaly fine.
The only real problem maker for your case which I have currently in mind is accelerated sim rate (slew). MSFS would be not the first simulation which strugle with this feature. May be we need some more test in this direction. ( I normaly not use acceleration speed… all is real time for me ).
But beside of this I allways wonder me about the relation “working-set to commited”. But we can not find out what happend here, that must do the developers with the given infos. ( and in generally why is there not a limitation of used memory ? ).
I asssume at this point the paging started. This is a real huge amount of data and if not a NVME is used also a ‘normal’ SSD slow down the speed extremly ( compared to RAM speed). But also in case you have an application which use only the RAM and this in the >50GIG range, you will notice that it needs time for releasing it, e.g. while exiting the application ( e.g. long DavincyResolve Session)… ( simple speed test is a file copy of a 30GIG file from/to SSD ).
Then is there the cache which windows use for e.g. file-access - so we are by the rolling cache.
Example: open a video on your disc with a size of e.g. 16gig, this 16gig are held in memory from windows. Normaly these kind of memory is statet as “allocable”, but who knows…
In another thread I mentioned RAMMAP. I ask me what happens if I let cleanup the working-set , shared mem, etc… if MSFS in running
Agree maybe best to refer to it as “memory Issue”. I also agree that the big problem with memory may be in sim rate & slew. I don’t normally slew and only slewed on that flight Thursday and the first time I had memory issues.
I have two NVMe drives. FSXX is on C: along with all other program. My rolling cache and pagefile are on a different NVMe drive.
I’ve read some comments that the rolling cache and FSXX should be on the same drive or it causes performance degradation. I don’t buy into that theory and think that using two drives will give better performance.
I have 64GB of RAM and see very little activity to either drive.
I will do slew test next and will most likely run into memory problems.
And know all the performance problems FSXX has, I don’t run on Ultra. I go back a forth between Medium and High and limit to 30FPS which does help a lot with the excessively high main-thread utilization.
