MSFS 2020 Benchmark, nothing in Community folder, the stock A320, weather settings few clouds, no AI or online traffics etc. as shown in the video, and flying with the internal AI, recording with CapframeX, and the conditions and flight practically always are the same and you can compare different hardware configurations, or between different users etc.. I will add the same exact flight plan file for the benchmarking repetitions flights here.
KJFK BENCH.PLN (3.5 KB)
You have a nice settings benchmark for your future comparisons if you upgrade with faster or higher quality components. Good Idea and I should do the same with my system.
Another with the difference of LODs 200
My 13900KS has approx. the half performance than this 7950X3D
Is that possible?, is there some wrong in the video, or my config.?
Thanks for adding the CapFrameX info. I will definitely add that to my PC toolkit.
Really good, I have found almost 10 FPS in average and 5 in LOW, in the 200 LODs video, the problem was that I have two monitors, one 1080p connected to the iGPU and the other to the GPU, and I have read today that intel in the next gen is going to add L4 cache only for the iGPU, so I tried only with the monitor of the GPU on, disabled in Windows or/and BIOS the monitor of the 13900, with the same result, and eureka, I have had some FPS more. I would have to record the first video again with this manner (re edit: no, with LODs 100 with iGPU off in the other monitor, is the same, maybe with LODs 100 the CPU caches are enough for all).
And DX12 a bit worse:
Here’s a CapFrameX capture graph I just did.
2560 x 1440, DLSS Super Resolution - Quality, HDR on, Frame Sync Off.
Bonanza G36 (Enhanced) flying in Bosnia, SW of Sarajevo @10,000 ft. (~5,000 ft. AGL)
Here is what I got with your Benchmark Flight:
7800X3D
Asus Rog Strix X670e-f
RTX4090
Corsair Dominator DDR5-6000 @5400Mhz due to EXPO limitation
Windows 10
Iam using the LOD Terrain and Objects 200 setting.
Disabling the Memory Integrity like Microsoft recommend only when gaming and then back to on, is a must in MSFS, almost a 20% of gains with it off:
ON:
OFF= almost 20% more of frames and lows:
And even better if disabled in BIOS Virtualization, in cre di ble, it is like to have a new hardware. I had already the Virtual related W11 features as recommended before these last tests and in the first one benchmark test, though now disabling Virtualization in BIOS… a new machine. I enabled the core isolation in Defender, then I rebooted, then I went to BIOS and I disabled Virtualization and VOILÀ, the same sure that directly to have disabled the Core Isolation sure and then Disable the Virtualization in BIOS, but I did it so, just in case.
This time 25% better in average, lows almost the same fps:
With LODs in 200, better too, brutal, almost a 50% better, though still 20% less than the AMDs X3D achieving 90s fps:
Fantastic tool and guidelines! Thank you!
Here’s my result.
F2020 ultra settings 4K
CPU 7800X3D stock
GPU 4900 stock
64 MB RAM@6000mhz
Here my results. At ultra 4K.
First is F2020, second is F2024. As you can see over NY, I get some big spikes, like a micro full stop, then back to 100 fps.
The lowest FPS are right in the center of Manhattan, close to the roofs
System waiting for streamed PG data?
How’s your internet connection?
This is info (along with computer specs) we’re all going to have to get used to providing in these kinds of threads.
I’m been promoting CapFrameX all over this forum. I use it all the time.
300MB fibre
Thanks! I am being called to table, sorry for shortness.
Something is wacky then. I know it doesn’t show the same graph, but here’s a test with a Bonanza @ 1500 AGL over NYC.
This was FS24, pre-SU1
7950X3D, 64 GB DDR5-6400/CL32, 3090 Ti (so no frame gen), 1Gb fiber
Ultra Preset, TLOD/OLOD both = 200, DLSS 4 Quality (so 2K upscaled to 4K)
You should definitely not be running out of VRAM with that 4090.
Have you updated your BIOS, followed by a chipset driver update?
I think it’s good to keep those updated.
Looks very smooth, congrats! But you are flying on a higher altitude.
Then, of course, it’s difficult to normalize benchmarks of 2 “quite different” games… But your personal experience is always the real standard.
Have you tried with stock recommended profiles? (in game, or in Nvidia app).
Then, we should all run this benchmark with 1 monitor, for the sake of comparison, or is multimonitor irrelevant? Once done, run more benches as you like.
So are you also getting AMD FSR3 frame gen working across all 3 monitors or just the 1 main monitor? Others are indicating FSR3 FG working across multiple monitors - see here.
Glad to see there is a thread on CapFrameX. I have a new 5080 and trying to optimize and have it working pretty well but there are some almost stutters on flare to landing.
This is in a 787 on autoland running the ILS on autopilot… a standard approach. Running 4K and with 200 TLOD.
You see the stutters (FPS drops) at the end, but what is bothersome is that my GPU and VRAM are 50-70% with the new 5080 but my Ryzen 7950 seems to throw these spikes of activity… when nothing is really going on. Average CPU load is 19% but can spike to 89% (red arrows).
I think this is the program itself… doing something internal… in relation to the flare to landing and more extensive runway in view…
I would like to understand this and more fully utilize my 5080 maybe pushing my TLOD higher.
From the top of your CapFrameX screenshot I’m assuming your running MSFS 2024. I’ve just upgraded to a 9950x3d and I am seeing similar spikes. From my SU3 screenshot below you can see the occasional spikes but I’ve noticed something very interesting just watching my graph. My 16 parked threads on my non-3d vcached CCD get unparked and then there is a momentary spike across all 32 threads and then the 16 threads get immediately parked again. I agree with you in that there is some rouge sub-process on the sim.
