So I have been struggling with getting decent framerates like everyone else in VR. I have a Ryzen 3900x, 32 Gig DDR4 3600, Nvidia 3090, and Rift S. As you can guess I’m mostly CPU bound. So I’ve been experimenting with different settings/overclocks and have something interesting that is somewhat known, but I don’t see it talked about a lot.
Anyway, I’ve found that turning off SMT and doing an all core overclock (4400 mhz) gives me about 10% extra performance in FSVR. The test I did was just sitting on the runway and waiting for the framerate to stabilize. It went from ~39 to ~43 fps. At least half of that comes from just turning SMT off. Looked like in stock form with PBO and auto overclock on I was hitting around 4100-4200 mhz while FS was running.
So I’m partly just passing info along, but also I have a question that may be difficult to answer but thought I’d give it a shot.
Given the results above, anyone have a guess at what I might gain by moving up to a 5800x or 5900x? I don’t think the SMT trick helps those from what I found in reviews, so theoretically I’ve made up some of the difference you see in review (like Tom’s Hardware for example that tests FS in 1440p). Using the Tom’s Hardware reviews as an estimate, I came up with only around 10%.
Ryzen 5900x and 3080 here. Cpu usage is between 40-45% and GPU usage is 95%. I play on 1440p with ultra-high settings and fps 30-35 on the ground in handcrafted airports and 40-45 on other airports. 50+ in air.
This is interesting. To add to my SMT off findings, it appears to be true for Intel as well from looking at this article,
And this graph in particular,
Very similar to what I found. That plot also backs up the comment above about the core count not making a difference (if you have at least 6). Although I would at least get an 8 core in case Asobo adds DX12 and better threading in the future.
Just did some tests on a 5900x. Made sure I was Main Thread limited by dropping the resolution to 1920x1080. Sitting on the end of the runway, I was getting:
85-87 FPS with SMT on
95-97 FPS with SMT off
FPS is high because I was testing with a bunch of other stuff turned off like clouds and traffic for consistency reasons and to tweak for VR performance.
Yea I would start there. Then if you decide to try to do an all core OC like I did, you can compare to see if it even helps. You need to exceed whatever it boosts to in auto mode in order to gain anything. Mine was running in the 4.1 to 4.2 range, so the 4.4 I got helps some.
So just finished flying for the evening. Turned SMT off and enabled PBO. I can’t say I notice a whole lot of difference, but I see a couple of extra FPS.
Strangely enough, my GPU (3080) is in the red on Dec Mode, but I’m averaging between 25-30fps on the ground. It does stutter every now and again, but I’m happy overall.
Will try some overclock settings on the CPU tomorrow and see what I get.
I should be able to answer my own original question in regards to Zen2 vs Zen3 this weekend. 5800x is out on the truck right now!
Any suggestions on how to test? No built in benchmark. I’m thinking change settings so it’s mostly cpu bound, but not low low settings. Then just start up at the same airport with constant conditions. Waitfor the framerate to settle in and see what I get with both.
Let the screenshots speak for me. Each screenshot was taken from the same location with the same settings. All I did was swap the CPU and fire it back up again.
First one is 3900x PBO/OC SMT On
2nd is 3900x All Core 4.4ghz and SMT Off
3rd is 5800x PBO/OC SMT On
Well partly because I just managed to snag an Amazon pre-order for the 5800x, but this review suggests it’s the right one for MSFS anyway. No explanation as to why it would be that much better though. Most other benchmarks show it no faster or slightly slower.
For my usage though, I had felt in the past that I didn’t really need 12 cores after buying the 3900x. I don’t do very much video encoding, etc… anymore that would take advantage of it.
Thanks for that info. I had the same questions/concerns that I really out-kicked my coverage with the 3900x. Could have definitely saved money there, because I don’t stream etc.
As far as I can tell, there’s no significant difference between any of the 5000 series AMD processors when it comes to Flight Simulator. So the $300 5600x should be just as good as the $800 5950x. Flight Simulator does not use the extra cores, and they all have about the same single threaded performance, which seems to be the biggest factor.
Yea I agree. They should all perform about the same. I have to think there was some inconsistency in that reviews tests. I think they just reprinted the result from their earlier 5900x review.
Nevertheless there is a big jump from the 3000 to the 5000 series. Much more than I expected.
