Microsoft Flight Simulator 2024 PC performance testing and settings analysis — we tested 23 GPUs, the game is even more demanding than its predecessor

Microsoft Flight Simulator 2024 screenshots
(Image credit: Microsoft)

Flight Simulator 2024 introduction

Microsoft Flight Simulator 2024 was released on November 20, 2024, and it had some serious issues at takeoff. But things have calmed down enough that we've now managed to test it on a wide selection of graphics cards and processors to see what sort of hardware the game requires for an optimal experience. We've tested 23 of the best graphics cards from the past eight years, including nearly all of the latest generation offerings. We also tested a handful of different CPUs to check how those impact performance. Due to the game's requirement for a constant internet connection for data streaming, we also took a deeper look at just how much data the game streams per hour.

We also recently had the launch of Stalker 2: Heart of Chornobyl, another game that can be extremely demanding of your PC hardware. Give that a read if you haven't already. Microsoft's Flight Simulator 2020 already had hefty requirements for max settings, and the new 2024 release ups the ante. It also ditches DirectX 11 support and leverages DirectX 12 for all systems.

The minimum requirements list an AMD Ryzen 5 2600X or Intel Core i7-6800K processor. Basically, you want a 6-core/12-thread CPU. On the GPU front, the minimum specs list a Radeon RX 5700 or GeForce GTX 970, but those are wildly different levels of performance — according to our GPU benchmarks, the 5700 is about 2.4 times as fast as the 970. This is clearly an error and should list the RX 570 4GB, which has about the same performance as the 970. You also need 50GB of storage, and a 10 Mbps broadband internet connection is required at all times.

The recommended hardware consists of a Ryzen 7 2700X or Intel Core i7-10700K, Radeon RX 5700 XT or GeForce RTX 2080, and a 50 Mbps broadband connection. Based on that, we can show why the minimum spec GPU is supposed to be the RX 570 4GB, not the completely different RX 5700.

The 5700 XT is only about 15% faster than the vanilla RX 5700, while the RTX 2080 ends up being over three times as fast as the GTX 970. If we were looking for equivalent AMD and Nvidia picks, a better selection would be an RTX 2070 Super for the recommended GPU (about 5% ahead of the 5700 XT) and RX 570 4GB for the minimum GPU (about 6% faster than the 970). And for Intel, the Arc A380 basically matches the RX 570, while the Arc A750 comes in 3% behind the 5700 XT.

It's important to note that while Flight Simulator 2024 can be very CPU-limited on higher-spec GPUs, that's also mostly with maximum quality settings. 1080p ultra won't get much above 90 fps in our testing, but 1080p medium nearly doubles that limit — or cuts the CPU demands in half, if you prefer. And 1080p low settings can still hit reasonable framerates even on relatively slow processors.

But system requirements are only the start. There are quite a few settings you can tweak, and the initial install size for Flight Simulator 2024 ends up being much smaller than with the 2020 release... and then it streams in a bunch of map data as you play, which is where that broadband connection requirement comes into play. Let's briefly discuss our test setup and then hit the benchmarks to see how modern PC hardware fares with MSFS24.

Latest Videos FromTom's Hardware
Jarred Walton
Senior Editor

Jarred Walton is a senior editor at Tom's Hardware focusing on everything GPU. He has been working as a tech journalist since 2004, writing for AnandTech, Maximum PC, and PC Gamer. From the first S3 Virge '3D decelerators' to today's GPUs, Jarred keeps up with all the latest graphics trends and is the one to ask about game performance.

  • -Fran-
    Thanks a lot for this testing Jarred. As always, it's interesting to see how the cards stack to one another in MS-FS.

    I'll have to pick you on this one though: "This isn't all that unusual. We've seen plenty of games where, if you're CPU limited at lower resolutions, AMD's drivers and GPUs still hit higher levels of performance. This is reputedly thanks to better threading optimizations, and of course, we're also looking at somewhat silly test settings — $700 to $2,000 GPUs running on a top-tier gaming PC at 1080p medium isn't normally the intended workload."
    Sorry, but that feels a bit dishonest. Everyone in the industry knows nVidia doesn't have a hardware scheduler; or at least, not in the same vein as AMD. AMD has implemnted a full hardware scheduler for command queue inside the GPU, so nVidia has their implementation in the driver (which, all things considered, is very good), but has the massive pitfall that when the CPU is hogged, then your command issuing will suffer. The result of that is exactly the graph you have at 1080p. This became a real tangible problem when games started using more threads and MS introduced HAGS, as I'm sure most remember how nVidia couldn't even use it effectively until a few years back.

    Happy to be corrected in the deeper technicalities of this, but again, that paragraph felt more like a Userbenchmark justification than just giving credit to AMD for having their scheduler inside the GPU design instead of leeching off the CPU instead.

    Not to mention how you're throwing everyone that wants infinite FPS at lower resolutions under the bus :)

    Regards.
    Reply
  • RoLleRKoaSTeR
    Wow, no love for the 3090?
    Reply
  • JamesJones44
    The amount of data the game downloads is pretty crazy. At the low end of 250 MB (2000 Mb) a hour, it would be 120 Gb per month at 2 hours gameplay per day. At the high end, 575 MB (4600 Mb) it would be 276 Gb per month with the 2 hours per day usage. Hopefully if a player sticks around the same general regions the data downloads are far less. Otherwise, those with data caps might take a hit.
    Reply
  • JarredWaltonGPU
    -Fran- said:
    Thanks a lot for this testing Jarred. As always, it's interesting to see how the cards stack to one another in MS-FS.

    I'll have to pick you on this one though: "This isn't all that unusual. We've seen plenty of games where, if you're CPU limited at lower resolutions, AMD's drivers and GPUs still hit higher levels of performance. This is reputedly thanks to better threading optimizations, and of course, we're also looking at somewhat silly test settings — $700 to $2,000 GPUs running on a top-tier gaming PC at 1080p medium isn't normally the intended workload."
    Sorry, but that feels a bit dishonest. Everyone in the industry knows nVidia doesn't have a hardware scheduler; or at least, not in the same vein as AMD. AMD has implemnted a full hardware scheduler for command queue inside the GPU, so nVidia has their implementation in the driver (which, all things considered, is very good), but has the massive pitfall that when the CPU is hogged, then your command issuing will suffer. The result of that is exactly the graph you have at 1080p. This became a real tangible problem when games started using more threads and MS introduced HAGS, as I'm sure most remember how nVidia couldn't even use it effectively until a few years back.

    Happy to be corrected in the deeper technicalities of this, but again, that paragraph felt more like a Userbenchmark justification than just giving credit to AMD for having their scheduler inside the GPU design instead of leeching off the CPU instead.

    Not to mention how you're throwing everyone that wants infinite FPS at lower resolutions under the bus :)

    Regards.
    I don't honestly know for sure what the difference is at 1080p medium. It could be as you say, and it's a reasonably regular occurence at lower resolutions and lower settings. There are lots of details that neither AMD nor Nvidia really provide, and I don't want to make blanket assumptions about the root cause. So I do a handwavey "differences in drivers and architectures" and move along, because the amount of time required to properly analyze this stuff isn't small and doesn't add enough to an article to justify my trying to figure it out.

    Ultimately, the performance is what it is. AMD does better with its top-tier GPUs at settings that don't really matter that much for such expensive hardware. It's in the same category as "winning" at 4K ultra performance on a budget GPU that does say 15 FPS while the competition only does 10 FPS. Technically? Yes, that's 50% faster. But it just doesn't matter because no one is using budget GPUs to run 4K ultra. Just like almost no one is using extreme GPUs to play at 1080p medium.

    (And don't say eSports because that's a tiny market overall, and if you're running 1080p medium or 1080p low in an eSports game you're going to hit a similar limit on a $500 GPU as on a $2000 GPU. It's basically a fringe scenario, in other words. Plus I don't test or write about eSports games.)
    Reply
  • JarredWaltonGPU
    RoLleRKoaSTeR said:
    Wow, no love for the 3090?
    Nope. And no love for the 3080 Ti, 3070 Ti, 3070, 3060 Ti, 6950 XT, 6900 XT, 6800 XT, 6800, 6750 XT, 6700 XT, 6700, 6650 XT, 6600 XT, 6600, 6500 XT, and 6400 either. Or most GPUs from the RTX 20-series and RX 5000-series and earlier. Because while the numbers are potentially interesting, gathering all that data can consume multiple days.

    In general, the 4070 Ti Super and 4070 Super should bracket the 3090. In this particular game, based on the 3080 10GB results, it looks like Ampere generation cards may do better vs Ada than in some other games. So, I test all the latest generation (mostly) and a few previous generation that can be used as points of reference.
    Reply
  • ekio
    Optimizations will come soon.
    Don't forget that this is not really a "game" but a simulator, and it gets updated continuously, like a production software, over the years.
    This is just v1.0 for now.
    Reply
  • Roland Of Gilead
    Hi Jarred,

    With MS2024 being VR capable, do you/Tom's ever hope to cover more games that do incorporate that element? I guess it's a whole lot of testing to add to the already heavy load. But, interesting to see.

    I know there probably isn't enough uptake in the VR world to warrant the effort, so can understand why it's not really relevant.

    As a matter of interest, do you test the VR parts of any games you've reviewed just to see how they play?
    Reply
  • edzieba
    If you're looking at that and wondering what happened to the RTX 4060 Ti and RTX 4060, the above discussion about running out of VRAM applies here as well. Probably, framegen would work fine at 1440p and 1080p, but at 4K ultra, the 8GB cards run short of memory, and you get a choppy mess. Wasn't it a great decision by Nvidia to not go with a 192-bit interface and 12GB of VRAM as the baseline, like it did with the (original) RTX 3060? Now, anyone who bought a 4060 or 4060 Ti gets to pay the price.
    What a bizarre interjection.
    Not only are the 4060 non-Ti and 4060 Ti 8GB not present is ANY chart (let alone the framegen one that paragraph sits under), but even if they were then going by the 4060 Ti 16GB's performance they would be getting sub-30 FPS at 4K Ultra regardless of whether framegen was on or off. In other words, those cards would be gong from unplayable to also unplayable.
    As the 4060 Ti 16GB results show (from from 31-34 FPS to 24-37 FPS with framegen) more memory would have done nothing of value, the performance bottleneck is elsewhere - likely the sheer lack of hardware available for both shading and upscaling.
    Reply
  • -Fran-
    JarredWaltonGPU said:
    I don't honestly know for sure what the difference is at 1080p medium. It could be as you say, and it's a reasonably regular occurence at lower resolutions and lower settings. There are lots of details that neither AMD nor Nvidia really provide, and I don't want to make blanket assumptions about the root cause. So I do a handwavey "differences in drivers and architectures" and move along, because the amount of time required to properly analyze this stuff isn't small and doesn't add enough to an article to justify my trying to figure it out.

    Ultimately, the performance is what it is. AMD does better with its top-tier GPUs at settings that don't really matter that much for such expensive hardware. It's in the same category as "winning" at 4K ultra performance on a budget GPU that does say 15 FPS while the competition only does 10 FPS. Technically? Yes, that's 50% faster. But it just doesn't matter because no one is using budget GPUs to run 4K ultra. Just like almost no one is using extreme GPUs to play at 1080p medium.

    (And don't say eSports because that's a tiny market overall, and if you're running 1080p medium or 1080p low in an eSports game you're going to hit a similar limit on a $500 GPU as on a $2000 GPU. It's basically a fringe scenario, in other words. Plus I don't test or write about eSports games.)
    It doesn't need to be a fringe scenario. If the CPU is fully hogged (try testing this is a lower end CPU and any GPU), then AMD should come out on top more often than not because of the CPU scheduler in the nVidia driver. That's the theory at least.

    This being said, I do agree it's way more testing and, perhaps (not sure), a hill not many want to die on, since if you're using a low end CPU and low-ish end GPU, blaming the nVidia scheduler for lower FPS'es is kind of moot.

    I guess the question here would be: which games and/or scenarios would load a mid-range CPU to the point it would really, and noticeably, affect the performance of an nVidia GPU that makes sense in the pairing?

    Like an i3/i5 with a 4060/4060ti vs RX6600/RX7600? I'm sure there's still plenty people using 8700K's and 9700K's, no?

    Regards.
    Reply
  • DS426
    "Microsoft's Flight Simulator 2020 already had hefty requirements for max settings, and the new 2024 release ups the ante. It also ditches DirectX 11 support and leverages DirectX 12 for all systems" - Jarred Walton

    Lol, Microsoft had DX12 on the market for almost six years by the time FS2020 was released. As for FS2024, that's nice that they're using their own latest graphics API... *cough* 10 years later *cough*.

    To be fair, I know game engines don't always benefit significantly in performance terms by going from DX11 to DX12 -- a lot of it depends on draw calls, multi-threading efficiency and heavy multi-threading in general, and of course the hardware itself. That said, additional enhancements and changes since DX12's original launch like DXR (DirectX Raytracing) are value-adds.
    Reply