yeah i meant msfs not windows lol
Same with high end system. Msfs didn’t use our vram and ram enough. Panning camera after some time result in reloading form cache. See performance monitor ingame. Bad programming mostly because of optimising for consoles.
Not happening for me
still having these annoying stutters with SU6.1, anybody else?
seems like reports went down quite a bit but I still can’t get rid of them, haven’t done a single full flight since the release of SU5 because they are so annoying.
Yep, it’s still there. Though I rolled back to another driver - 610.74 I think - which is a lot smoother for me.
Unfortunately the problem is still present, I’m starting to resign myself that they will never solve it, I would bet that the next super feature they will introduce will be native compatibility with DLSS5, leaving aside more important problems like this.
I already posted this in another thread which points to the same issue, so just in case anyone here is not observing the other threads and is anticipating an update.
Sebastian did talk briefly about pre-caching in the Dev Stream (September 16th).
He said when moving the camera in circles, the objects stay cached for only a short while and get unloaded after a few seconds when looking towards a fixed direction.
He also explained that textures stream from Internet → RAM → PCIe Bus → VRAM.
That point is critical. Moving textures across the bus from RAM to VRAM in real-time is a complex process, which explains the massive frame-time spikes across the MainThread and RdrThread during panning.
This is what I also observed with the built in FPS counter. However, in my opinion the issue comes down to one main factor. I think that the current method flushes assets way too aggressively, even when there is still more than enough VRAM/RAM available. Since the engine forcibly unloads those assets after a few seconds, it forces high-end systems to repeatedly run that heavy RAM → Bus → VRAM cycle every single time you look back up or behind you. If the engine simply allowed high-end rigs to hold those assets in the available VRAM headroom, that bus transfer wouldn’t need to happen continuously, and the stutters would disappear.
I am wondering if there is a fundamental engine/architectural reason why this ~20-second asset flush loop can’t be relaxed or made dynamic on DX12, or maybe an ‘Unconstrained’ precaching option for high-end systems could actually be considered for a future update?
honestly I’m not expecting a solution to this anytime soon.
There was a specific question in the dev stream regarding these stuttering issues that wasn’t really answered IMO.
Sebastian only said something like “we are always improving performance and oh, look here, dev mode has now a new option where you can identify stutters caused by the aircraft simulation!” followed then by the more general explanation of how the LOD works.
It seems that this current behavior of the sim is fully intended with no plans to change anything anytime soon.
Quite sad, been plagued by this since launch of SU5 which actually made me drop MSFS and join XPlane and honestly, I couldn’t be happier right now.
Maybe will come back when SU7 is fully released to see if I was wrong there (hopefully).
I don’t know if anyone here is on the SU7 beta, but the situation has worsened—at least in the latest SU7 beta build—and that worries me quite a bit. I ran the usual tests, and while the significant FPS drop used to occur after waiting about ten seconds before panning the view up or down (or vice versa), it now happens much sooner—in less than 5 seconds. If you’re in the beta, give it a try; it’s very concerning, as the problem is now practically constant.
No fix. YET
In 1.9.16 I feel the panning up and down in Fenix A320 cockpit is smoother than previous. Avoiding, or almost avoiding, the annoying stutter after looking at FMS. Just an initial at gate impression, but feels like so to me.