Despite what some people tells, DLSS has been introduced as a part of the RTX tech package first and foremost, and largely, as something to make full scale ray tracing viable as early as the 20xx gen (and even afterward). Ray tracing full scenes and not just what is rendered on screen which is called Screen Space Based method just like in MSFS for example (yes. MSFS DO uses some form of ray tracing already and you can do ray tracing without RTX, only for reflections in the case of MSFS) is so taxiing on the GPU that it botchers your FPS typically by half. DLSS doubles up your FPS on average (assuming no bottlenecks), and there you have it: full scale ray tracing at roughly the same FPS as without.
Without DLSS, full scale ray tracing would have been dead right from the start… and would still not be viable for most people even on the 40xx gen. Nobody would want to sacrifice literally half of their FPS unless the difference would be hugely noticeable (like night and day, which it is not despite how remarkable it can be sometime). DLSS is what permits RTX ray tracing to even exist as a viable marketing asset to begin with.
It’s an 11900k. Of course I have an eye on the 13900k or the 5800X3D but I just upgraded a little over an year ago for comparatively cheap 800€ from 8600k it must stay till next year.
I asked you because I have recently upgraded my PC to i5-12600K (liquid cooled) with an RTX 3090 and although everything has improved significantly I have not achieved exactly what I expected, so I believe that in the next few months I will replace the CPU with i9-13900K.
I’m getting around 30% better performance with the 4090 vs the 3090 based on rasterization alone (all graphics settings the same). Also there’s 3x less low fps with the 4090. With DLSS3 fps is doubled.
This is some benchmark stats for a KAVX-KLAX flight with inibuilds KLAX and live traffic with FSLTL models.
Is the 4090 a pcie4 or pcie5 card. and is it running in which speed slot?
I can get 36 fps with stutters or 20 smooth with a 3060, and I have to wonder if the studders arent because the 3060 is a pcie4 card riding in a pcie3 slot.
Sounds like we’ll need at least a 7000 series X3D then… That’s what I’ll be upgrading before swapping out my 12gb tuf 3080
Ive said it before the GPUs are at least a generation ahead of the CPUs, I think it would be a generation after the 7000 to fully push that monster 4090 to it best so give it a year or 2 of driver updates and optimisation and a chance for CPU tech to catch up… that card will be brilliant and a far better prospect than shelling out over the odds now for it
Interesting that’s what you reacted to from my post re: the 4090 vs 3090 performance. Not the 100% fps increase with DLSS3. Or the 30% increase due to rasterization.
I wonder how significant an increase in stuttering frequence of .9% is?
Just curious sir, I was surprised to see a higher percentage that’s all not having a dig would just expect it the other way considering stuttering is a big complaint if the sim that’s all.
The dlss 3 increase is irrelevant as it just adds in double ai frames in between so not really anything there other than showing ai frame doubling working by adding in duplicate frames. That’s why I didn’t react to that
I’d take a smooth 60 FPS over a 80-90 FPS with some stutters any day
What matters with DLSS3 frame generation is whether the end user notices it when flying. If you get strange artefacts that are quite visible and off-putting then it’s useless.
Whether it’s “fake” or not doesn’t matter too much. I’m more interested in the image quality.
I see one user review on this thread:
However, the frame generation is not great. Even at 75 FPS (150 frame generation) you’ll notice flickering and scrambling quite a bit when you move your head. At lower FPS it still stutters. It’s a weird feeling.
What about you other rich folk with 4090s? Tell us how frame generation looks when flying. Panning the camera inside and out? Cockpit instruments? Glass cockpits? UI elements? How’s it look?
Hello again all! I put together a video where I did a flight test between the two cards in triple monitor mode, as well as in VR incase anyone interested:
Yeah I think Nvidia have reached a tipping point in there design, RTX 5000 series cards must focus less on power improvements and more on efficency, I just can’t get my head around a RTX 5000 series taking up 4.5 slots, 2500 cuda cores, I mean where does it all end…
At this rate you won’t need a computer case. You buy the GPU computer and insert the motherboard inside once you’ve waited 3 days for the radiation and heat levels to drop…
I don’t remember having any stutters with reprojection enabled. OpenXR used to have a fit sometimes when transiitioning between 25/30/45 fps in a 90Hz headset but that’s pretty much gone away, but you do need to scale your graphics accordingly to allow reprojection to work. If it’s setup correctly, any stuttering you have will have been caused by the CPU.
DLSS 3 will come to VR I’m sure of that and this being hardware accelerated will mean you won’t have the initial hit on fps performance you get with the various software reprojection systems we have today.
How did you measure these FPS numbers? Mark from Simhangar showed, that the internal FPS counter in developer-Mode won’t count the extra „virtual“ frames generated completely in the GPU.
I.e. the framecounter of MSI afterburner shows way higher FPS when DLSS3 is working.
As DLSS3 renders complete frames within the GPU without needing help from the CPU for those „in-between“ guessed frames. MSFS should benefit from that even when being heavily CPU limited without DLSS3 on.
