So I decided to take some time to compare my 3090 vs the 4090. I run an i9 12900k, 32gb ddr, as the base system on a 2tb nvme. I ran the tests on my 3 55" TVs setup for my home sim with RSG G1000’s, as well as in single monitor and VR modes (Reverb G2).
The 2 main tests a ran were with the C172 on the Ground at KFRG parking spot 123, and again on the ground at KLGA runway 4 with the SWS Kodiak, both with the few clouds preset. Any questions about the tests let me know.
The numbers are the average frame rate I saw via the developer console.
Interestingly, the biggest improvement seemed to be in VR. Going from low 30s to almost 60.
My sim settings Im running most things ultra, with High on clouds, 110 on OLOD, and 200 on TLOD.
This seems like the modest sort of improvement you’d expect from a new gen. Certainly not the mind blowing results some seem to be touting. Great for VR though.
You didn’t mention the settings from Nvidia control panel or your in game graphics, especially DX 11 or 12.
Personal opinion :this makes my investment of my 3090ti feel worth it. My next build will align the 4090 ti in a couple of years for my LG OLED 4K 120Hz 65 In TV monitor. But by then 8K prices will drop. Neverending battle.
I’ve never changed any settings in Nvidia control panel. DX11 performance was probably 15% less across the board so I didn’t even continue with any tested and stuck with just DX12.
I agree - the 3090Ti is more than fine for this sim. VR seems to be the only thing taking advantage of the 4090 so far. The shocking thing to me is I see less performance in 2k on the triple screens, where I figured I’d see a boost due to driving all the extra pixels. Maybe as the drivers mature things will improve.
There seems to be a misunderstanding of GPU vs CPU bottleneck in your statement. The reason your VR performance significantly improved is because VR is heavily GPU-limited, far more than without VR. All other scenarios show a modest performance improvement using a 4090, as additional performance gains are limited by a high mainthread frametime. Using a terrain LOD higher than 200 will further reduce gains from a 4090.
This only highlights the fact that MSFS is largely a CPU-limited title, with the exception for VR.
The configuration I thought I’d see a decent jump in would be utilizing the multiple displays, as that would need to drive higher pixel counts much like VR - or am I off base?
It does not look like any of you PC guys are getting the 4090 Ti, as it is apparently melting power supplies and Nvidia is cancelling it. Also, another issue seems to be it takes up four slots on the motherboard, wow!
I was just going to say this. I currently have a 4090 and I never see the GPU usage break past 65-70 percent at 4k. Sim always says cpu limted when fps on ground is 45-50.
That’s what I would have expected. Would be interesting how the 4K performance is with a 5800X3D instead of the 5900X. I guess you can get some extra FPS out of the 4090 there
Seems you’re CPU limited in the 45fps range with multiscreen. It’s more demanding for the CPU than single screen. Triple 4K shows a nice improvement. Goes from unplayable to ok.
The 4090 has a 50% faster clock but I think that its faster memory would not bring miracles to MSFS. That´s interesting for rendering and modelling but in MSFS many assets still need to be loaded from hard disk very often, despite they sit on the VRAM afterwards. Even with your nvme disks (I have them too) you will see that they barely reach a mediocre speed during regular flying because the bigger amount of data is transferred during initial flight loading only and later on data is just updated in smaller batches.
The good point of both cards is still their huge VRAM size. That´s proven to be a benefit as of today with current memory management as it allows you to load and cache more things on a dense scene, so increase details or viewdistance until you find another bottleneck in your system. I would wait to see the performance of the eventual 4090Ti version, or even the 4080Ti, that may be having similar specs (just a bit lower) but with 16 GB VRAM instead, at least if they follow a similar approach as in the case of the 3080Ti vs 3090Ti. That could be a cheaper alternative but still with a crazy and powerfull core. I have owned a 2080Ti and now a 3080Ti and they are like tanks.
Those rumors could be marketing tricks to avoid that they eat the 4090s stock if people know that a Ti is going to be launched afterwards. There will be room for Ti editions for sure, as well as for factory overclocked versions. It has been like that for years and there´s no reason to change it now. The thermal limit protection will be triggered automatically as the 12900k CPUs do when they reach 100ºC. The 3080Ti is also switching automatically to low power mode to prevent it to go above 400 watts. In the worse case they will just downgrade the performance of the vanilla version to have a reason to increase it afterwards on the Ti and OC versions so that the latest appear to be better.
In the case of those high-end developments everything is already in place with a smart strategy and in the case of Nvidia that strategy is that users are always anxious to get next gen cards everytime There´s always a lot of demand and interest before the specs are even published. That´s fed with well put and targeted benchmarks and articles that make the whole thing look like a gem. That also worked for decades so they won´t change it now either.
This is true. In 4k with DLSS # I was seeing 84-85 frames on the ground and anywhere from 90-115 fps in the sky. Testing was done at KSEA using PMDG 737-800 (I know people are saying the pmdg doesn’t work in SU11 but I have not had any issues so far.
I dont see an official post on NVIDA website from their “quote”. It may be true, but they can find the issue and fix it and release it next year. All depends if they need the sales boast.
Word is DLSS3 doesn’t work with VR though. And recent videos show quite some significant artefacts on objects in motion and UI elements, which makes sense because it’s trying to guess what the middle frame would look like and any significant change between first and 2nd frame it will struggle. How noticeable these artefacts are seems to depend on the overall framerate too. If you’re running @ 30fps and turn on DLSS3 to get 60fps the artefacts are apparently a lot more noticeable than if you were running @ 60fps and turning on DLSS3 to get 120fps.
Time will tell I guess whether the frame generation can be improved, or if there are other optimisations to help improve it (like somehow excluding UI elements). As it is now though, I wouldn’t be surprised to see artefacts occurring on glass cockpits, especially in demanding scenarios. I’m sure we’ll know more in a few days.
There´s always a lot of demand and interest before the specs are even published. That´s fed with well put and targeted benchmarks and articles that make the whole thing look like a gem. That also worked for decades so they won´t change it now either.