Alrighty, as promised. This is kinda pointless, but I had to know 
TEST SETUP
- CPU: AMD Ryzen 7 5800x
- RAM: DDR4-3200 32 GB
- Mainboard: ASUS TUF Gaming X570-PLUS
- SSD: NVMe Samsung 970 EVO plus
- Case: BeQuiet Silent Base 802 (in high airflow configuration)
- VR Headset Samsung Odyssey+ (1600x1440 per eye)
THE CONTENDERS
- GPU 1: Gigabyte RTX 3080 Vision OC
- GPU 2: XFX Merc 319 RX 6800 XT OC
THE GOAL
I want to see which card is better for running MSFS 2020 and DCS in high resolution and VR.
The numbers will be all about max FPS, so all case fans are at maximum speed for testing. Frequencies are stock.
Software versions are whatever is current right now. SAM is enabled in UEFI. Drivers are stock, except sharpening is set to 0.
The final verdict will be which card I will keep for my own system. This decision will not be based performance alone, seeing that the two cards are different in how they are cooled.
DISCLAIMER: Obviously this test is a snapshot that just compares two video cards, today. It will be useless as soon as software optimisations or other changes happen. Don’t use this to compare the two chipsets, as other models might perform better or worse.
TESTING METHODOLOGY
I tried to test in a reproducible manner. For this, I came up with my own procedures described below.
While testing there is still a small but noticable variation in FPS. I estimate the variation to be around 3 percent.
So if you see 60, think of it as 58 to 62.
MSFS 2020
- Cessna 172 (gauges cockpit)
- KSFO rwy 28R (the default rwy)
- Time 15:00
- Weather preset “Scattered Clouds”
- after loading in or switching views, allow the game to settle and get to a stable FPS number
- Developer mode is used to retrieve FPS numbers. The limiting factor is always the GPU, so I’ll skip frametimes.
- Always check the resolution, not the percentage of supersampling. I’ve seen different numbers for the same percentage.
For testing I am interested in two numbers:
- FPS while sitting in the start position, basically do not move a thing
- FPS with the drone camera up in the air. For this I press INS to switch to drone, then R to fly the drone upwards.
When the center of the first taxiway crossing the runway is exactly on the bottom border of my screen, I stop.
I usually increase drone speed from 4% to 20% so it does not take too long to do that. - For testing VR I do the exact same thing with the headset resting on my table. So after loading in I press CTRL-TAB followed by SPACE.
DCS
This is a little bit harder to achieve. I use FpsVR to record the FPS and frametimes and I manually start and stop the recording.
Basically I run two missions and do the steps below. For pancake the HIGH graphic preset is used, for VR the VR preset. Labels off.
Version is DCS stable 2.5.6.59625.
Mission: Waltz (F/A-18 in Syria, starting on a carrier)
- 3 seconds after loading in, start the FPS recording by pressing CTRL-F11.
- 60 seconds later press F4 6 times to switch to the red planes
- wait 60 seconds then stop recording by pressing CTRL-F11
- Open fpsVR History Viewer to get the numbers (see folder Documents/fpsVR)
Mission: Cryz benchmark mission (the one with action, not the empty one)
- Press a for enabling autopilot. It will fly straight and at a defined altitude
- Start FPS recording when the message on the screen appears
- Stop FPS recording when the message on the screen appears
- Open fpsVR History Viewer to get the numbers (see folder Documents/fpsVR)
THE NUMBERS
MSFS 2020, XFX Merc 319 RX 6800 XT, FPS results. VRAM usage was around 10 GB.
MODE COCKPIT DRONE
Pancake 1440p HIGH 63 75
Pancake 1440p ULTRA 53 63
Pancake 4K HIGH 50 50
Pancake 4K ULTRA 37 39
VR 1651x2064 45 45
VR 2064x2580 44 44
VR 2270x2838 40 38
VR 2476x3096 35 35
VR 2683x3353 31 31
MSFS 2020, Gigabyte RTX 3080 Vision OC, FPS results. VRAM usage was around 5 GB.
MODE COCKPIT DRONE
Pancake 1440p HIGH 72 80
Pancake 1440p ULTRA 60 65
Pancake 4K HIGH 52 56
Pancake 4K ULTRA 42 45
VR 1652x2065 45 55
VR 1817x2271 45 54
VR 2064x2580 41 50
VR 2147x2684 42 50
VR 2476x3096 36 43
VR 2643x3304 35 42
DCS, XFX Merc 319 RX 6800 XT, FPS results.
Pancake 1440p 88 (just sitting on the carrier)
Waltz mission 2 minutes
VR 1768x2212 FPS=69.25, Median=10.6ms, 99th=11.5ms, frametime<11.1=93.1%
VR 1940x2424 FPS=59.78, Median=11.4ms, 99th=12.3ms, frametime<11.1=34.9%
Gryz mission
VR 1768x2212 FPS=86.48, Median=6.9ms, 99th=11.3ms, frametime<11.1=98.5%
VR 1940x2424 FPS=85.61, Median=7.7ms, 99th=11.6ms, frametime<11.1=96.4%
DCS, Gigabyte RTX 3080 Vision OC, FPS results.
Pancake 1440p 105 (just sitting on the carrier)
Waltz mission 2 minutes
VR 1768x2212 FPS=71.25, Median=9.5ms, 99th=12.3ms, frametime<11.1=98.6%
VR 1940x2424 FPS=70.41, Median=10.1ms, 99th=12.3ms, frametime<11.1=95.4%
Gryz mission
VR 1768x2212 FPS=87.05, Median=7.2ms, 99th=12.3ms, frametime<11.1=94.1%
VR 1940x2424 FPS=82.19, Median=7.8ms, 99th=11.7ms, frametime<11.1=95.3%
MY VERDICT
Both cards are fast, both run hot, to a point where it can be annoying. If you’re not into 4K or VR, you might want to go for a 3070 or 6800 for that reason. Nvidia clearly has better FPS and also more headroom in high resolution/details kind of situations. The Waltz mission at SS 180% shows this.
If you are looking to run a flight simulator with maximum FPS, then Nvidia should be the first pick.
That said I really like the XFX card. Its fans run at 1600 RPM when the Gigabyte would do 2800 RPM. The cooler is much larger and so are the fans. Also I have the perception that all games on the Radeon card run smoother. This could be a configuration issue or it could be my monitors freesync capability which works better then the Gsync compatible stuff it uses.
So I’ll decide to keep the Radeon card, even if it means less FPS. It’s a great card and a handful. Plus I can’t stand the whiny noise of the Nvidia card.
The Gigabyte Vision / Eagle / Gaming cards are not exactly the best cards in terms of cooling solution. They just do what is needed. For 1050€ that is not enough. (vs 999€ for the XFX)
My recommendation to VR flight simmers would be to try and get a 3080 with a huge heatsink and huge fans.
