2 hours ago, DAD said:

lucky you. now if you say your LOD settings are all at 400+ then I really envy you.

Nooooo alas, TLOD and LOD both set at 150. 

The most important part of the OP's original post at top of thread is his last sentence:

It is hard to understand why all the 1.5.xx Betas up to 1.5.14 could tolerate the increased memory speed but 1.5.15 and .16 utterly died on the same memory.

He had no problem in the Betas up to 1.5.14. It was the same with me. I do not believe his issue was hardware or settings. The problem was caused by the update to the Beta.

With both 1.5.15 and 1.5.16 updates, I cured the same problem the OP had by a simple 2 step process:

1) leave the beta, restart Windows, open msfs, close msfs.

2) rejoin the beta, restart Windows, open msfs, close msfs.

Before doing those 2 things: constant freezes of the sim, terrible fps and CTD's when using planes that are resource hogs in heavy scenery areas. After doing those 2 things: no CTDs, no freezes, always much better fps.

All I know about the issue, is that this worked for me perfectly. Twice.

 

OP Here - I changed my memory settings to XMP 6000  30-30-36-76 (as specified MSI X870 BIOS).  Sim now runs perfectly

Two hour flight in Lear 35 in and out of thunderstorms and variety of other weather.  Extended periods at FL200, 300, 410  with no problems.  Flew a convoluted approach into KSJC (historically a terrible VRAM experience) 12R with a flight over San Francisco (Modern Cities) and Oakland with an ILS approach into fog.  FSLTL was running as was Map Enhancement and Active Sky.

No problems - but, the MOBO did need 20 minutes to train the memory.  

It is still hard for me to understand why ONLY FS2024 suffers random failures running slightly faster memory profiles when every other performance testing software has no problem.  Why can multiple memory testing applications stress the various memory settings for hours with no failures yet the sim gives up in a few minutes.  Some programs take extra steps to catch and deal with hardware errors and seem to find a way around them. 

I run a mem test that randomly writes blocks of random data to memory and then reads it back.  I use 1 GB blocks up to 54GB blocks and let the random/read write run for hours with no problems.  NO errors are reported.

I guess as long as I find a profile that works I should stick with it.  The XMP 6000 profile does allow 10% to 25% faster test results in Passmark Memory results  compared to default settings.  My previous research (did this kind of stuff for 30-years) indicates that such memory improvements might increase frame rate by a very small, and probably not apparent to the eye, results.   

But, it is fun to experiment with this stuff.
 

33 minutes ago, TacomaSailor said:

.  

It is still hard for me to understand why ONLY FS2024 suffers random failures running slightly faster memory profiles when every other performance testing software has no problem.  Why can multiple memory testing applications stress the various memory settings for hours with no failures yet the sim gives up in a few minutes. 

 

How to tell me you’re young without telling me 😛

There was a time back in late 80s and early 90s where affordable “IBM clone” PCs were a thing. At the time, the benchmark for whether your “clone” PC was truly compatible was “Will it run Microsoft Flight Simulator?”. If FS crashed, your PC wasn’t good. It seems nothing has changed in 30-40 years 😛

53 minutes ago, TacomaSailor said:

No problems - but, the MOBO did need 20 minutes to train the memory.  

Goodness, it's back to 1993 I guess.

53 minutes ago, TacomaSailor said:

It is still hard for me to understand why ONLY FS2024 suffers random failures running slightly faster memory profiles when every other performance testing software has no problem.  Why can multiple memory testing applications stress the various memory settings for hours with no failures yet the sim gives up in a few minutes.  Some programs take extra steps to catch and deal with hardware errors and seem to find a way around them. 

I've never found a test app that tested like flight sim does...they'll either test some instruction set that flight sim doesn't use, or they'll not test something that it does use.  Sim +addons is the REAL stress test.   As virtual-chris mentions, that's a de facto carryover from the old-school benchmark...could it run Microsoft Flight Simulator?

53 minutes ago, TacomaSailor said:

I guess as long as I find a profile that works I should stick with it. 

There you go, that is the best advice.  Could you get a few % more?  Perhaps.  But you know what goes with that -- hours of testing and re-testing and re-boots etc. etc. and with the load times of sim + boot time + OS load...ugh, the mere thought of it.

1 hour ago, TacomaSailor said:

Some programs take extra steps to catch and deal with hardware errors and seem to find a way around them.

Again, there isn't anything a program can do to protect or harden itself against bad memory. If a memory read operation supplies a different value to the one that was previously written, there's no way for a program to know or detect this. If the erroneous value is used for, say, the address of something else to access in memory, then unless you're lucky and the value happens to be a valid address, the program will crash.

Anyway, glad to hear that you've found settings that give you a stable sim experience!

6 hours ago, TacomaSailor said:

No problems - but, the MOBO did need 20 minutes to train the memory.

Recommend you turn this feature OFF in the BIOS/EFI … over time it can keep retraining and causing LONG startup delays and the return of application/game issues.   Find the RAM tuning that works/stable, turn training OFF.

I am not a software programmer or hardware engineer, but I'll take a guess that it is the way in which MSFS reads and writes to the memory that is unique and stresses it more than other applications. Reading/writing 1GB blocks is a lot different than reading/writing a bunch of 2KB blocks or constantly reading from one section of memory. You could have a single chip on your memory module that isn't completely stable in the overclock and MSFS just keeps accessing that location. Again, theorizing here, but doesn't seem unrealistic.

Thing is, if your system isn't stable in MSFS with a certain overclock, it's likely to be unstable in other areas as well. Maybe it's Chrome crashing one day, or waking your computer in the morning to find it rebooted for no reason the night before. Software bug or hardware issue? If you're overclock isn't 100% stable, the hardware issue is more likely.

ECC memory (error correction code) is used in critical workloads like servers and workstations to protect against this. I'm sure you could find a database application or code run through a compiler or a 24-hour video render that crashes the same way MSFS does when an unstable overclock is used. Benchmark and testing applications can't test every scenario...

2 hours ago, Funky D said:

I am not a software programmer or hardware engineer, but I'll take a guess that it is the way in which MSFS reads and writes to the memory that is unique and stresses it more than other applications. Reading/writing 1GB blocks is a lot different than reading/writing a bunch of 2KB blocks or constantly reading from one section of memory. You could have a single chip on your memory module that isn't completely stable in the overclock and MSFS just keeps accessing that location. Again, theorizing here, but doesn't seem unrealistic.

Thing is, if your system isn't stable in MSFS with a certain overclock, it's likely to be unstable in other areas as well. Maybe it's Chrome crashing one day, or waking your computer in the morning to find it rebooted for no reason the night before. Software bug or hardware issue? If you're overclock isn't 100% stable, the hardware issue is more likely.

ECC memory (error correction code) is used in critical workloads like servers and workstations to protect against this. I'm sure you could find a database application or code run through a compiler or a 24-hour video render that crashes the same way MSFS does when an unstable overclock is used. Benchmark and testing applications can't test every scenario...

I am a software developer and I agree with most of what you wrote.  As I mentioned in this thread already, I had done an overnight MemTest (probably around 12 hours, until the report was spit out) one year before MSFS 2024 came out, and there were no errors detected in the MemTest I did (I mentioned to Luck38i already that I may have not used the original MemTest program, but I used of the other highly ranked MemTest versions when you Google for "MemTest"). But MSFS 2024 was causing my computer to crash, with my monitor's screen blacking out, and I was unable to get back into Windows when it happened.  Otherwise, my computer was relatively stable throughout the year, before MSFS 2024.  I had an odd and similar crash with Chrome when I was watching Twitch TV, but that stopped when I switched to FireFox to watch TwitchTV.  That crash with Chrome was very hard to reproduce and only happened once every several weeks of watching Twitch TV.

My computer shop's technician detected that my RAM was defective. I got the new RAM, and MSFS 2024 has been stable since. Actually, I had done several career missions in MSFS 2024 while I was using the defective RAM, and I didn't encounter any problems.  It was when I slewed to the top of El Capitan in Yosemite, and I left it running at the peak of El Capitan, where my computer would crash.  And my computer was generally stable in the few years before MSFS 2024 came out (after I reduced my RAM's clock speed to 3200 MHz to match my CPU).

I wish I had asked my computer's technician what software he used to detect my RAM problems, but I forgot to ask at the time.  And unless my RAM degraded further in the year since my last MemTest, then you are right, MSFS 2024 stresses the computer and RAM in a way that MemTest does not.  It's quite possible that the MemTest I was using does not do every comprehensive test, and there may be some cases that it does not rigorously test, that MSFS 2024 is doing to the RAM. Off the top of my head, I wonder if it's the amount of data that is written to, and read from RAM, that MSFS 2024 does, that is more than the version of MemTest that I used does.  

 

 

Edited July 17, 2025Jul 17 by abrams_tank

I’d recommend Prime95 with HWINFO to monitor throttling events.  Run Prime95 stress test duration for however long your longest flight will be.  So if you run 2 hour flights on average, then run Prime95 stress test for 2 hours.

There are several rumors that Prime95 will destroy a CPU, these are false unless you turn OFF thermal throttling or the CPU is defective.

With all that said, it is still possible to CTD MSFS with perfectly stable hardware.  

More serious hardware issues tend to freeze the PC (completely unresponsive) or reboot the PC.  PC spontaneous reboot is often related to power supply issue or power connection issue (not fully seated).  A complete system freeze is RAM or CPU or GPU issue.  CTD can be application code issue or RAM/CPU OC issue.

MSFS 2024 SU2 with many add-ons has been relatively issue free in terms of CTDs.  I’ve had one CTD that was triggered by GSX (coualt) and a complete system freeze triggered by nVidia GPU driver version.  Out of about 1000 flight sessions.

42 minutes ago, SayAgain said:

I’d recommend Prime95 with HWINFO to monitor throttling events.  Run Prime95 stress test duration for however long your longest flight will be.  So if you run 2 hour flights on average, then run Prime95 stress test for 2 hours.

did that as an eight hour overnight test with no failures

 

I've been running 64gb of 6400 MT/s on my AMD 7800X3d with the B clock set to 103mhz and uclock/memclock running 1:1 and have had 0 issues with MSFS 2020 or 2024 in the past year or so. 

Prime 95 isn't really the best test these days for memory, I use Y-Cruncher VT3, That is right now the most stressful memory test you can run. 

At the end of the day if your apps crash due to XMP/EXPO or other overclocks then you have to work on those and the process can take weeks at a time to dial it in.

1 hour ago, sultanofswing said:

Prime 95 isn't really the best test these days for memory, I use Y-Cruncher VT3, That is right now the most stressful memory test you can run

I’ll have to disagree with you, I’ve had failures under Prime95 that never surfaced in Y-Cruncher VT3.  Prime95 last update was June 2024, Y-Cruncher was Jan 2025.  Both are still being updated.

7 hours ago, SayAgain said:

I’ll have to disagree with you, I’ve had failures under Prime95 that never surfaced in Y-Cruncher VT3.  Prime95 last update was June 2024, Y-Cruncher was Jan 2025.  Both are still being updated.

VT3 is purely memory/Memory controller so Having an issue pop up in Prime 95 would lead to the CPU more than likely.

1 hour ago, sultanofswing said:

VT3 is purely memory/Memory controller so Having an issue pop up in Prime 95 would lead to the CPU more than likely.

The IMC is part of the CPU.