I’ve been saving up to update my Radeon 6800XT GPU and Ryzen 5600X later this year (assuming one will actually be able to get a hold of new CPUs and GPUs when they come out). I’m resisting the new 5800X3D because of that.
If I could get a stable locked 45fps with mostly high or ultra, plus a few shadow settings higher and light shafts on, I would be very happy, even more so with MR on. For me and my G2 that would probably be more that enough.
What say people here? Anyone else planning to upgrade in Q3 or Q4 this year?
i9 10850K, 3080 and G2 here. I would want to see a marked improvement over what I can already achieve now with the OXR Toolkit, which is 30-40 FPS, half ultra, half high, 60% OXR, 100% TAA, 100% FSR sharpening 75% (which makes 60% OXR look pretty good), FFR quality and wide and FOV 90%).
If DLSS, supposedly coming mid year in SU10, improves supersampling quality such that I can run higher OXR and hence improve visual quality to what 100% OXR would get me with the same FPS, then I won’t really feel the need to upgrade anything.
The main thing that would make me upgrade is a higher resolution OLED headset with eye tracking, which won’t be showing up until next year at the earliest I believe.
I’m currently running 100% OXR, 100% TAA and 100% FSR in rural areas and can mostly maintain a locked 30FPS (using Radeon Chill) in GA aircraft and gliders, without MR, though with micro stutters close to the ground looking sideways. In photogrammetry cities and busy airports I have to lower that to either FSR 80% or less, or use resolution override in the toolkit. I’ve tried MR, but Radeon cards don’t seem to be as happy with it as nVidia. Whatever, it makes things blurrier for me with way too much jello, as I can’t maintain 30fps, and 22.5 fps just sucks. I’m using a mixture of some ultra, with clouds, buildings and trees high, grass medium and others a mix of high or off.
I really want both cockpit and scenery as sharp as the headset can produce, and I’m not quite there yet in all scenarios.
I don’t expect 90fps with the next gen cards, but, as I say, with 45fps MR would be very usable and if that’s with all ultra, which the newer CPUs would no doubt help with, that would be a blast. Of course I could also use FFR in the toolkit, which I can’t at the moment in DX11 on a Radeon card.
Newer VR headsets would be nice too, no doubt about that.
I was thinking of upgrading my 5 year old i7 7700k cpu but the stats tell me not to. Open XR Toolkits advanced fps monitor and msfs fps counter both tell me i am gpu bound only (i have a 3080ti), so why upgrade? I’m pretty certain i will upgrade to a 4xxx series gpu when they’re available, as long as benchmarks indicate a decent (20+%) fps increase.
Thats going to be the big issue I think. I’ve got a 5600X and I’ve decided that I think I will bite the bullet and get a 5800X3D because getting a 4000 series GPU and the new CPU generation will be next to impossible later this year. The global supply situation isn’t any better now than it was at the end of 2020 (in some sectors its even worse). I’d do what you can now to make your current hardware last well into next year if you can.
Plus don’t fall for the click bait hype around the 4000 series cards. I know every frame counts in VR but don’t assume they’re going to be (insert improbably high number here) times better than the 3000 series.
I upgraded from an 7700K to a 5600X last year and the difference was huge, particularly in VR. I run in 4K with a 3080.
I’d be happy with a 50-60% framerate increase in VR, which seems possible with a 4080 card (or whatever it will be called) and a next gen Ryzen which will no doubt also have 3D cache as well (plus a new motherboard, sigh). Hopefully not a new PSU too…I have an 850watt Gold Corsair unit now and would be really annoyed to have to replace it with a 1Kw unit.
Plus, as you say, what about availability, not to mention scalper prices.
The 5800X3D certainly does improve things now in MSFS, especially smoothness due to better 90% lows, which are where a lot of stutter comes from. But they are expensive here in Oz right now at around $AUD720
I’m not disputing your experience.
Personally I am getting fps within a few fps of those with newer cpu’s and similar gpu’s/ram, same hmd. Online 4k gaming benchmarks show older cpu’s don’t effect fps much, and as i said, the Toolkits advanced fps counter tells me its my gpu that’s holding me back. I’m able to run msfs at my hmd’s max resolution (varjo aero) which is more than 4k (combined eyes). If i upgraded my cpu, in my eyes, it would gain a couple of fps, to maybe give me parity to those with newer cpu’s.
It’s not worth it for me to gain a few fps. I’ll spend the money on a gpu np, and on a better hmd (when that comes).
Disabling multi threading and using something like Process Lasso to close or migrate some processes (like WMR) onto lesser used cores can make up for older CPUs, as well as maximise performance on new ones. Always worth a try.
Edit: I hope mbucchia or Cpt Lucky can share some thoughts on this thread, as there are some settings in MSFS that impact more on CPU usage, that could be increased with even minor tweaks as mentioned above, let alone upcoming hardware.
You may find the next zen CPUs struggle to match the 5800X3D in MSFS.
If the rumours are true, the v-cache versions won’t be there at launch and will be released later… perhaps as a refresh type cycle a year later. Given typically we could expect 20-30% from a new architecture, a 6800X (or 7800X as apparently they may be moving to 7000 naming) would probably fall behind a 5800X3D in games where the cache has a huge impact like MSFS. It likely then won’t be until we see a v-cache version of Zen 4 that the 5800X3D would be dethroned.
So the good news is that you can probably just pick up a 5800X3D for your existing AM4 platform and avoid investing in a new DDR5 platform for another year or two by which time availability, prices and speeds of DDR5 should have matured somewhat. Use the money saved avoiding the new platform on a higher tier GPU.
5800X → 5800X3D is an insane jump in this game, more than a typical generational change:
You may be right. I haven’t read much about next gen Ryzens and I’ve only watched a couple of vids on the 5800X3d performance with MSFS.
1080p graphs don’t really relate to VR performance increases, which have been minor in 4K in MSFS (VR is likely similar) from the little I have seen so far. The bigger gain was in increase in 95% and 99% lows bringing them much closer to average FPS, which, as I mentioned earlier really help with stutters.
Still, this all bears careful thought, as I had naively expected 3D cache to be present on next gen Ryzen launch, so thanks for that insight.
That chart is for relatively low res 1080p gaming where cpu does really count (typically done on a 2D monitor that can’t do 4k), but for those of us using high resolution hmd’s it more like 4k (and greater) gaming, this is the area where gpu really counts, as the gpu is responsible for rendering all those extra pixels at high res, while the cpu just twiddles its thumbs.
We need to see in game optimisation not the cycle of throwing money at the next best hardware, for me 12900, RYX3080ti! 32GB it looks reasonable in the aircraft, they just need to resolve clarity and shimmer. Let hope DLSS and DX12 bring improvements that Asobo can build on……. assuming they care about VR which I’m questioning at the moment.
the comparison to 4K doesn’t hold, as it ignores what else is going on in VR. I do most of my gaming in VR on a G2.
When playing in VR you have high resolutions indeed and the GPU is working hard, but the CPU is also much more loaded than in flatscreen with having to draw the scene twice. This is why you hit a CPU limit at a much lower frame rate in VR than you would with the same CPU and settings in flat screen.
That means that the CPU ability is much more relevant in VR at high resolution than in flat screen at high resolution.
The 5800X3D was a very solid upgrade even from my 5800X in VR, I don’t play in 2D. The lows are definitely much improved but I also have much better CPU frametimes in general which allows more headroom on LODs and other CPU intensive settings.
Regardless, the point stands that it is highly unlikely that zen 4 will beat the 5800X3D in MSFS specifically until v-cache versions arrive (they will of course be better all round, but likely not specifically in the games that really hum with the larger cache like MSFS, ACC, star citizen etc). If it is true that the first CPUs are indeed not going to include v-cache variants, then that means that (if one feels the need for more CPU performance) a 5800X3D would potentially save the expense of an entire platform change for a couple of years and potentially allow the budget to go towards a higher tier GPU… plus allow buying into the AM5 and DDR5 ecosystem at a little more mature point in time.
I’m sorry but i don’t agree.
It’s the gpu that is working much harder in drawing the scene twice, the gpu renders pixels.
When i had a G2 it wasn’t getting a whole lot less than me running msfs at 4k on a monitor, maybe 15fps? Atm i’m running a varjo aero which is 2880x2720 per eye, so over 5k combined, no wonder i get less fps than if gaming at 4k on a monitor.
Look at the screenshot:
Look at how at 35ppd (G2 is about 24ppd, quest2 about 20ppd) my cpu load/headroom is half/twice that of my gpu. That’s a 5 year old cpu. I have 0 reason to upgrade my cpu where msfs is concerned.
I’m not going to get into an argument about it, but you need to understand the CPU is also involved in the construction of a scene and is highly affected by things such as number of objects (which doubles in VR of course).
There are also other non pixel-related rendering steps, just one example - calculating visibility. There is no point in rendering objects behind the camera only to have the GPU automatically cull them - Adds a cost in both CPU time adding those invisible objects to the display list and GPU time since the GPU has to do run the vertex shaders, just to have the clipping phase classify them as behind the camera. So the CPU normally does a ray-trace like step to see if objects are visible. This and other CPU scene drawing related work is also doubled in VR.
I agree of course that the higher you go in resolution the less likely you are to hit CPU limitations but even with a Varjo aero now that you can have dynamic foveated rendering, with a next gen GPU bumping up your frame rates you may find you start hitting main thread or render thread limits more often in certain scenes. It will somewhat depend on your settings of course but you could run higher LODs, traffic, fly in areas with lots of objects etc and start seeing CPU become much closer to limiting.
And all this still misses the point that the OP was talking about Zen 4 - the point is simply if games such as MSFS are his focus, then perhaps it’s worth considering why upgrade to AM5 and DDR5 with all that expense when an AM4 drop in 5800X3D will likely outperform it on the CPU front anyway until a refresh brings v-cache to the line up. This ties into what you are saying where I completely agree that the saved money would then be better spent on a higher tier of GPU.
Well, I’ve decided to put this conversation regarding the benefits of increased cache for MSFS to the test myself. I currently have a 5600X, 6800XT and the G2 headset. After reading more about the 5800X3D I decided to see how it performs for me. Thanks to a federal election going on here in Oz, I have an extra $250 in my bank account, due to a “cost of living” one time pension payment (pork-barreling? nah, just a coincidence I’m sure). So I have the funds to buy one, which I snagged last night at the MRSP price.
Should arrive in a week and I’ll see just what a difference it makes. In the meantime, I’ll record a set of comparisons using my current set up with CapFrameX and we shall see if the extra cache and cores makes a worthwhile difference in VR using a G2 and WMR plus the toolkit, despite the slightly lower clock speeds.
Do you know if Intels upper range 13th Gen will have a similar smoothing effect on stuttering as the 5800x3D? Will they also have augmented cache?
Can you provide guidance on which process to place on which core, using Process Lasso?
They won’t have extra cache bonded to the die like AMD does with the 5800x3D, but they likely will have more L3 cache on the main die. From what little I have read, it seems AMD will not be using 3D cache on their next gen Ryzens, at least at the start, but will also have more L3 cache.
As for Process Lasso setting, I’m no expert, but for VR you can move WMR away from core 0 and 1 using affinity settings. I suggest you PM forum user RomanDesign, who knows a lot more than me on the subject and has posted his settings somewhere (which I can’t find at the moment).
My 5800X3D arrived this morning and as promised, I now have some comparison test results against the 5600X.
The tests were done in a rural area, with clouds off, flying the Stearman biplane, using the G2 headset. Obviously a best case scenario. My GPU is a Radeon 6800XT, no OC, resizable BAR on (SAM in Radeon speak). I have a 100Mbs fibre to the premises connection.
OXR 100%, in game render 100 and Toolkit FSR as shown on the graph.
As you can see, average frame rates are about the same, but 1% and 0.1% lows are better, especially at FSR100%.
Subjectively, just flying about and sightseeing, the 5800x3D is noticeably smoother.
I haven’t done any runway comparisons as yet, nor New York or London.