So, has anyone seen any benchmarks of the new AMD Ryzen X3D CPU:s in VR? The Ryzen 9 7950X3D seems to be the one reviewed so far.
As of writing this some benchmarks for pancake mode have trickled out, but even if the number of pixels in e.g. a Reberb G2 is similar to a 4K display I figure results in VR may be different, on account of having to do geometry twice. Though to be honest I’m not entirely sure it works that way.
Be that as it may, VR benchmarks would be nice.
Yeah, I would like to see some comparisons as well, I want to upgrade from 5900x, and I would like to see how big the bump might be for the RTX4090.
In the commentary Paul mentions the 4K results are GPU Bound even though a 4090 is being used.
Hopefully we will see some real case examples of VR in use for MSFS and DCS using this new CPU.
The April release of the 7800X3D will be interesting.
Can someone suggest to Paul what mods to an ULTRA preset might reduce the GPU stress just enough to let the CPU differences dominate.
I wonder if the test is using DLSS, since it would render the game at a lower resolution and I assume at that point the game would be more CPU bound.
No one test still in VR? This test at 4k could do an approach.
I have no numbers but I can say that for me in VR there is no improvement moving from 7900X to 7950X3D.
With RTX4090 and G2 headset I still must lock at 45fps and in Fenix A320 its a hit and miss. Still woblling MR. So, waste of efort and money.
In 2D with Frame generation I have increase form 120 to 140fps as 1440p. VR is soooo behind 
@Dankovic Motion reprojection runs primarily on the GPU, so the x3d was never going to be able to help with that unfortunately, but you should be seeing some improvement in performance. Have you verified that FS is using the vcache cores? 45 fps with MR is still a tall order at G2 resolution, especially with a complex airliner, even on a 4090, but 30 fps MR should be entirely doable. I’ve seen a sizable jump in overall smoothness over my 7700x with it. It keeps a very steady framerate with higher scenery settings. You might also try running it with SMT off.
The 7950x3d seems to deal with TLOD 200 in the same frame time that my 7700x could run TLOD 100, and is a lot smoother about it.
Agreed though, VR still needs a lot of love.
For those interested here’s a 5 minute frame time graph of a flight in VR in the Extra 300, departing from London City Airport, westbound right up the Thames river, at 1000 feet, towards Heathrow. This is with all settings maxed, except TLOD at 200, and no traffic, broken clouds preset.
This is on a 4090 with Motion reprojection on and locked at 30 fps using RTSS, on Reverb G2 at 200% supersampling with DLSS. Rock solid and smooth for almost the entire run in one of the worst case scenery scenarios.
I don’t have a capture from my 7700x unfortunately but I did the same run just before upgrading and it was full of micro stutters.
Cheers!
Any improvement with TAA with 100% in OXR Tools?
@JGL574 The performance of TAA vs DLSS is really about the number of raw pixels being rendered after everything has been sent from the CPU, so it’s a question of GPU power. The 7950x3d won’t really help with that. It can let you run more complex scenery for the same frame rate though.
Thanks then, though at 2D 1440p the above video with 7950x3d+4090 has 90fps with TAA, with my 12900k 5.1Ghz-4090 with default A320 I achieve only around 50fps in the same 2D scenario with TAA, curious. I have too a 1GB fibre.
@JGL574 Right, but that’s 2D, where GPU load is so much lower, especially at a very moderate resolution of 2560x1440. In that case, A320 over photogrammetry Manhattan, you (and the video maker) are probably cpu limited. Have you checked your dev mode fps display to see how the loads on cpu and gpu are? In this case the X3D could definitely help to increase frame rate. FS scenery and photogrammetry in particular are quite cache sensitive.
Your original question was about TAA with 100% OXR though. VR is an entirely different beast, and you often can’t infer performance in pancake mode translating over so easily.
A reverb G2 at native resolution is asking the GPU to render the equivalent number of pixels as five 1440p screens combined. The load on the GPU is hugely increased and that usually becomes the limiting factor. The X3d here won’t really help to increase frame rate, but it will get more processing done in the same frame time without taking over as the limiting factor in total frame rate. This means you’re still GPU limited but can display better and smoother running scenery for each frame.
Aaam. Yes, I did it in that test some days ago, it was saying that I was CPU limited in red or so.
For my setup I’m GPU limited with a 4090 OC’d and a 7950x3D in VR with using a G2 V2 and everything on Ultra and a LOD of 200-400 (depending on where I’m at). I get 45FPS (locked using RTSS) and again, my 4090 is the bottleneck.
My CPU runs around 7-12ms and the 4090 GPU is in the 17-22ms.
What do you set your render scale to in the OpenXR Tools for WMR? I’m at 200% and wondering if it’s worth having to replace my Mobo and Ram to upgrade from 5800x3d to 7900x3d.
Definitley not 7900x3D, that’s the worst choice! Only 6 vcache cores and possible scheduling issues. 7800x3D is the way to go for MSFS, although probably not worth if given what you already have
I was hoping from better results when I upgraded from my old i7 8086K to a 7950x3d, with a 3090 GPU, but it wasn’t a spectacular increase.
What is the best method to check what cores are used by MSFS? I’d like to do some tests to verify that MSFS is allocated as it should.
Disabling the cores in BIOS with the non X3D cache I guess. Could you do this benchmark recording with CapframeX to compare it with the 13900?
Windows Game Bar will suspend the non V3D cores. Don’t disable them in the BIOS.
