One trick that works really well with the A350 is to find a stand that’s far away from the terminal buildings and facing in a direction where there’s nothing to see from the cockpit window. I’m able to fly the plane from the default EDDM where I constantly lose the avionics and my setup if I try to use the gate which the real-life flight started from.
It is instructive and relevant to compare the performance of MSFS2020 SU16.1 with MSFS2024 SU5.1
I used the PMDG 737-900ER KPHL KMSP.
MSFS2020: Fantastic sharp LOD on departure - zero stutters with quite a few parked aircraft. Smooth panning around the cockpit at all times. Dials and switches responded instantly.
Very sharp LOD all the way and very evident on landing - very smooth landing and taxi in to the gate with quite a few wide body aircraft scattered across the ramp. PG was off. So was the time a bit - a 2min loss on roll out at KMSP. Panning around the cockpit was a bit jerky at times and not as fluent as MSFS 2020.
MSFS2024: two minor but noticeable stutters on taxi and take off. Banking round over there was noticeable scenery blurring - quite aggressive. The sound cut out twice as we entered the cloud base. Some of the dial responses were slow or stop/start. Landing wasn’t as smooth as with MSFS2020 - there were some small stutters, worse than with SU4 but not terrible, but there was some pop in and blurred textures. Stop start on taxi in but not horrible. Sounds were a big improvement over 2020 up and so was turbulence - it did feel more immersive than MSFS2020 at times.
Overall - in smoothness and fluidity MSFS2020 feels like a serious upgrade to MSFS2024. On scenery LOD on Xbox MSFS2020 seemed a lot better in cases where blurring occurred in MSFS2024 and trees in MSFS2020 seemed more realistic. When there was no blurring it was close to a tie.
Sounds and turbulence/feel - MSFS 2024 ahead of MSFS2020 but some of that is also due to the MSFS2024 PMDG upgrades to B737.
Given how long it has been since the release of MSFS2024 it is still not a clear enough upgrade to MSFS2020 on Xbox and it is also very evident that not enough emphasis has been given to console testing.
First picture departure KPHL MSFS2020
Second picture departure KPHL MSFS2024
ini A350
SamScene3D ZSAM - WF Scenery ZHCC
The Xiamen scenery includes many enhancements to the surrounding city, all of which were visible at the same time in drone camera view. The ground itself was quite blurry while in close proximity to the airport, but it’s nice that the stuff I paid for actually loaded in, as this has not always been the case historically. Terrain quality improved once the airport disappeared, and there was some surprisingly satisfying scenery on this route considering how “untouched” China is in the simulator. Runway was fully detailed on landing despite ZHCC being a new purchase from the sale, and therefore not yet stored in my cache. High quality buildings with no need for ridiculous LOD model switching, even when I did the drone cam spin around the plane at the gate. Performance in China continues to impress me—the sim once again did everything I needed it to do with very few hiccups. I consider this to be my full experience on the flight, though there were two non-critical issues I’ll add here at the end.
I received the memory warning once when the loading finished, and after being blown away by the performance at ZHCC, I switched to walkaround mode which resulted in a ctd when I turned to look away from the aircraft. The walking was just a stability test as the plane was secured and I was ready to exit the session anyway.
Black Square TBM 850
MM Simulations EFTU - Attben EFKI
First flight in this plane so not much time for scenery, but I did notice that the terrain quality didn’t improve as much as it usually does when flying a smaller plane. The blurries were definitely there even away from the 3rd party airports, perhaps because of the greater systems depth in the TBM. The biggest offender was again the 2024 upgraded scenery—this time it was the Attben Kajaani, which had all the same problems I’ve described in my KLAS and ENGM rants. Nothing on the airport loaded in detail unless you walk right next to it. From where the planes are parked, all the objects have laughably simple LOD models even when observed from the cockpit view. With no performance gain on a tiny backwoods airport, I have no choice but to conclude that these paid upgrades offer no value and are graphically a huge step backwards from the supposedly unoptimized 2020 versions. Although the session was very stuttery overall (wouldn’t be an issue for me if the graphics had been better), in flight the game still remained playable as the actual short freezes occurred only when testing the aircraft’s external interactions while stationary on the ground.
I just don’t understand how it’s so hard to replicate on their end. They have telemetry data too
At Asobo studios they should have 2 retail xboxs. Both turned on and never turned off.
1st Xbox vanilla Sim, vanilla everything
2nd Xbox loaded with all the popular MP items, everything, zinertek lights, airports, scenery, bravo traffic, PMDG, Inibuilds everything
Both xboxes never turned off, flying in most resources heavy planes, streaming 24/7 all over the planet with unlimited fuel obviously.
Eventually surely memory warnings and stutters and all of our legacy flight sim issues will arise for them to see.
The main variables in the Xbox environment is the MP content and internet connection, so they need to start there. Hardware is all the same. perhaps they could also get some honeycomb periphs, keyboard, mouse, USB hub, headset, etc just in case that’s an issue too
All they would need to do is spawn into hand crafted Heathrow in the A350 or A340, and the issue would be obvious almost immediately. The freezes and stutters begin as soon as the zoom into the airport begins when loading into a flight, shortly followed by low memory warnings. I occasionally get it in the 777 too but is generally smoother when loading in. Have been flying on 2020 most of the week and the difference in performance is night and day. One thing that I have noticed is how much better my landings are when I’m not getting the stutters before touchdown. I really do hope 24 can be as smooth as 2020 soon. SU4 was close but still not as smooth.
Agree - personally I think the thing that’s missing is QA testing on commercially available Xbox. They clearly don’t do this - which explains their failure to replicate problems we see within 5 minutes of booting up the sim.
In the store MSFS2024 carries an “optimised for Xbox” tag - it really hasn’t - the welcoming view of the oversized cursor immediately on booting up the sim stands as a symbol of how little time and attention has been given to the sim on Xbox.
It has been a while since the launch of MSFS2024 - 19 Nov 2024. It is really shocking that we are in the place we are today after 18 months.
Compare this with the efforts of some of third party developers who extensively test on Xbox - the results clearly show the advantages of doing this (e.g. PMDG, Just Flight, FS Reborn) even though we may have to wait a few weeks longer after launch to get their products on Xbox.
MS/Asobo really need to take a look at what is possible and learn from it. It’s a strong opportunity for improvement just waiting to happen
What’s most frustrating is how they teased us with SU4, as that showed how good the sim can be on Xbox. Was just a shame we only had it good for a few months!
Exactly - it’s unbelievable….
That would not be a very valid test because one single badly behaving module could cause massive problems. And what do you do when you see that? Start to debug all the 3rd party DLC? Microsoft does not do that. They told us so. There are a ton of badly programmed products on the Marketplace. We all know it, but when they cause problems, we too often point to Microsoft to fix them.
I will tell you a secret, we hardly do any Xbox/Cloud specific testing. Not only because it is extremely difficult and terribly time consuming (you have to wait for Microsoft to make it available after each tweak, you can’t use a Dev Xbox), it is also pretty wasted time. It the product is efficient on Windows MSFS it will be solid on console. You just need to consider the SDK as the law, do exactly what it tells you to do. If you use FS2020 code in FS2024 (for example for cross compatibility) make sure you inform the customer. In our test, the use of FS2020 code in FS2024 can cause problems. https://forum.pmdg.com/forum/main-forum/general-discussion-news-and-announcements/390136-please-assist-us-understanding-wasm-crashes-better-even-if-you-do-not-see-them?p=398575#post398575
Mathijs Kok
PMDG
For sure in my own experience the clean install path is essential - perhaps it sounds a bit nerdy bit I guarantee good results and it has kept me out of trouble others have experienced more than once….
I was flying a lot of aircraft in SU5 with no Low Memory warnings or black avionics or CTDs. (Not all though)
But it’s also the case that there are a lot of stock aircraft in MSFS2024 that aren’t performing very well, whatever the reasons.
Another thing that perhaps we should bear in mind is that SU5 has caused a variety of issues for most users across all platform types so it has been problematic
The big difference between the PC users and Xbox users is that the PC users have more settings to tweak to make intelligent trade offs…
And it is a series of trade offs in the end - smoothness and responsiveness (FPS) vs LOD and extras (grsphics) has been the traditional trade, now in SU5 we have to trade off RAM space as well on Xbox perhaps not unexpected given the minimum specs of MSFS 2024 and beta testers noted that the Xbox was being “pushed to edge” of its capability.
Being more specific, my own view is that the missing piece of testing is the validation.
Now some of that validation can be carried out by beta testers - but in the case of SU5 the feedback from Xbox users wasn’t acted on ( or listened to?) and the PS5 beta was only a few very short days long. This may well be because the feedback is included with PC beta feedback - and as validation should address the specific needs of each platform, perhaps that just hasn’t worked.
So I think it is a relatively quick and easy fix for MS/Asobo to put right for us. Beta threads for each platform and indicators when that beta feedback is being acted on or looked at so testers can move on to something else. It feels to me like a lot of effort on the part of testers was wasted this time around - I didn’t join the SU5 beta because some specific WASM feedback of mine on SU4 was not listened too…
I have all that on Xbox and great performance. Usually my flights are shorter than 1,5 hours. Longer and I get the memory warning. Doesn’t even need to be the same flight.
On PS5 beta I use it only in combination with PSVR2. There it is the vanilla sim. But same problem when using it more than 1,5 hours.
Like the memory gets filled up over time and never released. Also, just sitting on apron seems to count into that time limit.
That’s why we do very, very extensive tests on memory leaks. Almost every weekend I got two machines doing that non-stop. I find it surprising that it is not standard across all DLCs. I have pointed these issues out to several devs who just do not seem to do this kind of test.
Default A320 Bravoairspace livery
FLYT Simulations SBCT - Latin VFR SCEL
Unable to complete preflight due to constantly losing avionics after almost any movement inside the cockpit. Multiple restarts and always the same result, except on the last try I managed to finish the setup only to lose it all just before pushback.
No Flight.
ini A300 cargo
Default EDDK - BMWorld LEVC
Unable to complete preflight due to constantly losing avionics after almost any movement inside the cockpit. Multiple restarts and always the same result.
No flight.
ini A350
Default EGLL - MK Studios EFHK
6 attempts at different gates/stands around the airport: CTD on the first one, then 2 to 5 the game was partially frozen, preventing movement or entering the aircraft. On the last one I got inside, but was unable to complete preflight due to constantly losing avionics after almost any movement inside the cockpit.
No flight.
ini A300 cargo
Bravoairspace MDSD - Latin VFR KMIA
The best clouds I’ve seen in weeks with absolutely gorgeous ground detail from beginning to end. As someone who has screenshotted the Miami/Fort Lauderdale area extensively in msfs 2020 and even more so in msfs 2024 after the Panthers’ back-to-back Cup wins, I can confidently say that this is the best it’s ever been graphically. In the final year of 2020 and also in 2024 up to SU4 or SU5, the whole city was just tiny pyramids on a pool table, but now, the photogrammetry was picture-perfect. The city enhancements by LVFR had incredible draw distance which was particularly satisfying during final approach. I can’t emphasize enough the tremendous emotional impact of witnessing such a beautiful skyline backdrop for a landing in one of the fastest growing hockey markets in the world. This is the sort of thing I’ve only seen in A330 Driver videos.
Unfortunately, that landing is where it ultimately fell apart.
1st, 2nd, 3rd, and 4th attempts: Black screens when only 300-500 feet to go, plane reset, controls locked, engines dead and plane dives down, then back-on-track in the air.
On the 5th attempt I switched to a runway which didn’t have the city in the background—instant success despite the tailwind. During taxi the plane crashed through the ground (A350-style) and back-on-tracked to the runway. I got to park and secure the plane on the cargo apron, but the plane kept losing the avionics all the way, resulting in a lot of waiting as the displays took time to come back on and an engine had to be restarted.
Black Square TBM 850
PMS50 GTN 750
Attben EFKI - FLY X Simulations EFKS
Sunset flight in bad weather so not much terrain visibility. Same stutters as last time, the most severe of them coinciding with the external interaction animations again, though now I realize some of Black Square’s custom systems trigger performance issues as well. Other than that the only noteworthy observation was the typical gigantic performance/graphics disparity between the “unoptimized” 2020 and the “upgraded” 2024 sceneries. The 2020 EFKS had the terminal in full detail with interior and everything when viewed from the farthest corner of the apron, all the while achieving superior frame rate consistency. Once I stopped interacting with the aircraft, I didn’t get a single stutter, not even during a full walkaround tour.
Default A321
Pyreegue EGAA - iniBuilds LEIB
Success on 1st try. I think I improved my chances with a couple of additional tricks I’ll list here for those interested:
- No Bravoairspace livery with iniBuilds airliners.
- If the cockpit textures struggle to load when looking up and down, stop camera movement immediately. Don’t touch anything until everything is fully loaded.
- Don’t attempt a realistic line up because quickly turning scenery kills the avionics. Instead, sacrifice some runway with a gentle and slow turn while keeping an eye out for possible stutters which could signal an impending memory warning. Again, stop movement at the first sign of struggle.
Cruise went straight to the highlight reel: MSFS 2020 quality mountain crossing over the Pyrenees displayed such majesty and splendor that I can’t even remember how to spell “A350 from Heathrow” anymore. Two photogrammetry cities loaded fairly late but the draw distance was good enough to keep all the tiles in view at the same time, and the memory management didn’t have to level the terrain around the horribly optimized BMWorld airports on the Spanish coast.
The usual solid island performance was nullified by the countless multiplayer aircraft on the airport. The whole rollout was depicted in about 12 frames, and to top it off, the first black avionics hit when I was still on the runway. Long wait to get controls back, then another blackout as I turned toward the terminal. Parking and shutdown were problem-free, inspiring me to spend time admiring the fantastic scenery in drone camera view. There’s A LOT of good here despite all the stability issues.
Black Square TBM 850
PMS50 GTN 750
MM Simulations EFOU - MM Simulations EFTU
I can’t get good terrain quality at higher altitudes in this plane. Not sure about the performance impact of that GTN 750 add-on as I’ve only used it a handful of times before buying the TBM.
Every other stage of flight was visually flawless except cruise at FL280. As soon as I descended about 2000-4000 feet, ground was sharp again, like I passed through a threshold. Only my 3rd TBM 850 outing so won’t say anything more just yet.
For those of you interested in a deep dive on the probable reasons behind what happened here, you might be interested in this explanation (from Claude):
Understanding the SU5 Xbox Performance Regression: What Happened, Why, and When to Expect a Fix
There’s been a lot of frustration since SU5, and a lot of it is justified. I wanted to put together a single explainer covering what’s actually going on, why a mature sim can break like this from one update, and what we can realistically expect. I’ve tried to keep the technical parts accessible to people who aren’t programmers (like myself!).
What’s actually broken
SU5 introduced a memory management regression that hits Xbox Series X|S and PS5 hardest. The symptoms cluster tightly: “running low on memory, disabling avionics” warnings mid-flight, black PFD/MFD/EFB screens, WASM avionics rebooting repeatedly, stutters during ATC interaction and approaches, CTDs even in light GA aircraft, and a visible texture/scenery downgrade on console. The heavy airliner missions (A330, 737 MAX 8) are effectively unplayable from a cold boot for many users.
A telltale pattern lots of people are reporting: the first flight of a session works, the second is worse, the third is catastrophic. That “it degrades the longer you play” signature is important — I’ll come back to it.
Why a single update can break a sim that’s been out this long
This is the question I see asked most, so here’s the honest answer.
SU5 wasn’t a small patch. The release notes describe changes across building generation, airport parking, particle effects, NPC aircraft, cockpit code, texture handling, forests, the renderer’s culling system, plus new PSVR2 support, a new weather radar API, and new SimConnect/WASM camera APIs. That’s an enormous amount of surface area touched in one update — the renderer, the streaming/texturing system, the cockpit code environment, the scenery pipeline, and a whole new VR rendering path, all at once.
A few mechanisms can produce the symptoms we’re seeing:
Memory budget shift. The sim divides a fixed pool of console memory across many subsystems — renderer, scenery streamer, terrain cache, aircraft systems, avionics code, ATC, weather, etc. If even one subsystem’s footprint creeps up by a few hundred MB and nothing else shrinks to compensate, you cross the limit. The first thing the sim does under memory pressure is disable the WASM avionics — which is exactly the black-screen-EFB symptom everyone’s hitting.
A memory leak. That “first flight fine, third flight broken” pattern is the classic signature of leaked references — objects that should be freed when a flight ends are still being held somewhere. Each flight leaks a little; eventually you hit the avionics-disable threshold before you even take off. SU5 touched the renderer culling and scenery object lifecycle code, which is exactly where this class of bug tends to appear.
The new VR path forcing a renderer refactor. Adding PSVR2 with foveated rendering almost certainly required changes to the rendering path. VR and non-VR share code, so changes for VR propagate to everyone — including Xbox players who never touch VR.
The MSFS 2020 compatibility layer. Lots of users report that removing 2020-era addons helps. That compatibility layer is a known cost center, and SU5 may have changed how those packages load or stay resident in a way that’s worse for memory.
The short version: this is what happens when a complex product takes on a large amount of new surface area in one update cycle, and a regression slips through that only shows up on the platform with the least headroom.
Why programming for Xbox specifically is hard
This is the part that explains why your friend’s high-end PC runs SU5 fine while your Series X falls over.
PC has elastic resources. Console has hard walls.
On PC, when a program asks for more memory than is ideal, the operating system absorbs it. Virtual memory means physical RAM is backed lazily, and if RAM fills up, cold data is quietly paged out to disk. GPU drivers do the same for VRAM. The program doesn’t crash — it just runs a bit slower. The cost of inefficiency is paid in performance, and on a strong PC you don’t even notice.
Console is the opposite. The Xbox Series X has 16 GB of unified memory, of which roughly 13.5 GB is available to a game (the rest is reserved by the OS). Series S has far less. There is no pagefile, no swap, no overflow to disk. If the sim asks for more than the budget, the allocation fails — and the result is either a crash or the “graceful degradation” of shutting your avionics off. It’s a hard wall, not a soft one.
It gets worse: console memory is unified, meaning the CPU and GPU share one pool. On PC, textures eat VRAM, a separate budget. On Xbox, if the renderer suddenly needs 500 MB more, that has to come out of something else — the scenery cache, the avionics, the audio system. There’s no separate pool to absorb it. CPU time is fixed too — there’s no “buy a faster chip.”
The cruel consequence for testing: a regression that adds, say, 300 MB to one subsystem is completely invisible on a developer’s 64 GB workstation. It’s invisible to PC beta testers on enthusiast hardware. But on a Series X it disables your avionics, and on a Series S it’s catastrophic. The bug is real the whole time — it just doesn’t show up where the people checking the build are looking.
This is why “console is just PC with a smaller machine” is wrong. It’s a fundamentally different engineering discipline, and a flight sim — which inherently wants to consume infinite memory for terrain, textures, and aircraft systems — is one of the hardest possible things to fit inside a fixed console budget.
What could have been done to avoid it
None of this is exotic. It’s the standard playbook for shipping a complex cross-platform product:
-
Memory budgets as a hard CI gate. Each subsystem should have a tracked per-platform memory budget, measured automatically every night on a fixed set of test scenarios. A change that blows the budget should blockuntil it’s brought back under.
-
Long-soak testing on real console hardware. The symptom here is leak-shaped. Leaks don’t show up in a 20-minute test flight — they show up after five back-to-back flights or a long-haul. CI should include multi-hour soak tests on actual Xbox/PS5 hardware that fail if memory grows steadily.
-
Treat beta feedback as a gating signal. The community flagged these exact issues throughout the SU5 beta. When a structured beta produces a clear, repeated, reproducible signal about a console regression, shipping anyway is a process failure, not a technical one.
-
Decouple feature shipment from optimization shipment. PSVR2 support is a marketing-driven date; memory refactors are engineering-driven. Bundling them into one update biases the team toward shipping even when the engineering isn’t ready.
-
Test the weakest target first. Series S has the tightest budget of all. It should be a primary test target, not an afterthought behind Series X and PC.
When to expect a fix
Microsoft has publicly acknowledged the issue, and Asobo has said a dedicated patch targeting the stability and avionics problems is planned to ship before SU6 flighting begins, described as the team’s top priority.
No official date has been given. Based on Asobo’s past cadence — hotfixes have historically landed roughly 2–4 weeks after a problem update — a realistic estimate is late May to mid-June 2026, with SU6 beta likely opening sometime after. Treat that as an informed guess, not a promise. Check the official threads for anything more current.
What you can do right now to improve performance
Workarounds the community has converged on:
Memory hygiene
-
Clear the rolling cache periodically (set its size to 1, then back to your normal value — 16 GB is a common choice). Roughly weekly.
-
Install aircraft you fly often rather than streaming them — especially PMDG aircraft. Streaming is constant memory pressure.
-
Remove or disable MSFS 2020-era content. If it isn’t native 2024, take it out.
-
After a long session, fully exit and relaunch before doing anything demanding.
Settings to reduce memory/GPU pressure
-
Turn off photogrammetry.
-
Reduce or disable live traffic, airport vehicles, static aircraft, and ground crew.
-
Turn multiplayer off.
-
Until the hotfix, favor smaller airports — complex hand-crafted hubs are the worst offenders.
Aircraft selection
- PMDG aircraft are reported to behave noticeably better than the stock airliners. The stock A330 and 737 MAX 8 are the worst affected; GA and helicopters mostly fly fine.
If you fly a well-optimized aircraft into a smaller airport with these settings, you can get a usable experience until the hotfix lands. If you’re running the stock A330 into a major hub with full traffic and live weather, you’ll keep hitting the wall.
The bottom line
SU5 took on a lot of new surface area in one cycle, a memory regression slipped through, and it landed hardest on the platform with the least headroom to absorb it — exactly the place the testing process was least likely to catch it. It’s recoverable; Asobo recovered from MSFS 2020’s similarly rough early console period. But the structural lesson is that the rigor of the testing harness has to scale with the cross-platform ambition of the product. Hang in there — a fix is coming, and the workarounds above will get most people flying in the meantime.
This explainer was written with the help of Claude (Anthropic). The hotfix timing is an estimate.
Excellent, well detailed and informative test report - thanks for sharing.
The main issue to highlight and this may not be the right thread for it, is the PS5 version has very limited MP content so essentially they are running the Sim vanilla and raw and they have been the loudest in SU5 memory issues. SU4 was fine.
Asobo just needs to do better for consoles, I’m sure we will get there, it feels like 2020 all over again despite the multi threading efficiencies and everything else promised the console will struggle if you keep throwing everything at it instead of meticulous exclusive console environment testing
May be time for another console exclusive beta, we had them in 2020.
Personally I believe that what would help enormously would at least be dedicated beta threads for each major platform - I was reading the SU5 beta thread avidly (though decided not to participate this time) and the Xbox observations on Low Memory issues really stood out to me because I was looking for Xbox related observations. But they were buried in a large number of PC observations and so were not noticed (it was a clear sign of impending trouble) or acted on (low priority when buried in a mass of other comments)..
I don’t understand why the PS5 beta was set up only a few days before full release - it wasn’t enough time to get any user feedback and act on it.
Now we are in this position your proposal of a separate beta seems like it would be a good way forward - our specific memory issues on console need more attention.
I saw some of the PS5 feedback too but stopped paying attention when I realized that the Playstation-only guys have never seen the game’s pre-SU4 graphical state. If the current issues on console are resolved with the trade-offs we’ve discussed recently, my greatest fears are realized and I’m pretty sure at least some of the PS5 users would be in the same camp as me, begging for Asobo to introduce some instability to have a world worth flying in.
In the 2021-2023 MSFS 2020 I had many no-fly zones because of my add-ons, but I still prefer the honesty of a CTD over the memory management they introduced in early 2024. I very much disliked the idea of pretending the console could run the sim by removing all terrain and airport detail (including the scenery I paid for), even if it admittedly made the game very stable. I would rather complete only 1 out of 10 flights with decent graphics than have 100% stability in a Roblox world.
All the Xbox Youtube channels I followed switched to PC during that last year of MSFS 2020—I still remember that one guy making his last Xbox video showing how the ground around the airport had zero detail and the photogrammetry wasn’t visible at all. That was the last straw for him after having accepted the instability as a necessary trade-off. I miss the communities they created.
I’ll return with more detail in my future posts when I have more time, but ultimately, this is the core message I wish to send with all my performance reports. This thread isn’t what you guys think it is.