Tom's Hardware MSFS2024 CPU, GPU, and Network Benchmarking

[TenPatrol]

@BigCow74

Tom’s Hardware did a rather extensive set of benchmarks with 23 GPU’s and several higher end CPU’s. They also tested streaming data usage.

I’m wondering about the point of testing a game that has countless bugs and more holes than Swiss cheese. On top of that, the performance also depends heavily on the quality of the connection between us and the servers, because almost everything is streamed. Respect to them for even trying to test a game that doesn’t work as it should.

[BearsAreCool510]

No way, at least not in VR.

[SiphonicMoon7]

My guess would be that they tested it because it made financial sense to do so. A whole bunch of interested simmers clicked on the article to see the results and compare with their own research. That’s the goal!

[MeaningSheep475]

The 1080p results would say otherwise. 3dVcache is still king. Those 4k results could indicate different GPU bottlenecks. You’d have to
Look at the 1080p results to see how the CPU makes a difference.

[BigCow74]

@MeaningSheep475

Look at the 1080p results to see how the CPU makes a difference.

This is definitely how you would show a CPU performance benchmark.

However, practically speaking, how many 9800X3D and 4090 owners are playing at 1080p?

This is always the part about benchmarks everyone needs to consider. They have their place, but always temper that with your specific situation.

I used to be overly concerned about CPU temperatures in online benchmarking until I remembered that I run Cinebench multi thread tests and prime95 exactly zero times a year

[MeaningSheep475]

Very valid point. Although in flightsim you can be cpu limited at 4k in certain scenarious. Unfortunately these benchmarks are typically not done in complex airliners such as the fenix a320. Throw in terrain LOD at 400 and online traffic such as vatsim etc and a 9800x3d starts to show its benefit in 4k. I do believe you can still have a pleasant experience with the way DX12 appears to leverage the GPU performance. It could also be for future proofing as we won’t be so GPU limited at 4k in the sim when the 5000 series gpus release.

[PacificSet90456]

@TenPatrol

I’m wondering about the point of testing a game that has countless bugs and more holes than Swiss cheese. On top of that, the performance also depends heavily on the quality of the connection between us and the servers, because almost everything is streamed.

Benchmarking is helpful in finding hardware issues. For example, there is some info about hardware with potential VRAM problems.

Regarding the networking streaming benchmarking, their testing isn’t to find the best ISP or how to configure a router. Their network benchmarking changed only one thing at a time during the testing, the number of flights and the duration of flights. They may have the worst ISP or the fastest connection. Their results should be about the same.

Since all their tests had the same bugs and holes, their test results should be similar if and when those bugs and holes are fixed.

I think the results would be helpful for anyone considering upgrading their CPU or GPU or for anyone buying or building a PC for MSFS2024.

[SmotheryVase665]

@PilotJedi668

Seb mentioned in the recent dev video that we can overload the GPU with ultra-everywhere. This results in stutters etc.

That is what I’m seeing with my Arc770 (which you will notice is way down near the bottom of the charts in the referenced article). In 2020, it can run 4K at ultra quite well. In 2024, it really, really struggles - even at high or medium settings - particularly when landing. I’ve had to downgrade both the resolution and settings on the graphics tab to get framerates consistently >25. Maybe performance will improve in later builds.