SU5 changed the CPU to GPU balance quite considerably, with the GPU now being much more the determinant of performance (especially in VR), so the old adage that you should throw the fastest single core CPU at the game no longer applies.
Rather than speculate on whether a CPU upgrade will be worth it or not, first pull up task manager and see what your CPU cores are doing while using FS2020 in various areas you normally fly on the 60 second graph. If no core is maxed out while keeping the view fixed then a CPU upgrade will do very little for you post SU5. Sure, changing the view will see CPU spikes up to 100% and a faster CPU would help with that, but for the most part most performance is experienced in this game with a stationary view.
eg. the highest core load my i9 10850K has with a fixed view is now about 60%, so in theory I could underclock from 4.8 down to 2.88GHz and still get similar performance with one core finally running at 100%.
TBH, your best performance boost in FS2020 VR with a G2 would come from upgrading your video card to a 3080 or better. Yes, your 3070 is no slouch for most games but in FS2020 with a G2, video card performance still reigns supreme even after SU5 and you would get much better performance and render resolution settings with a 3080+ than from the CPU motherboard upgrade you are proposing.
Firstly su5 ruined the vr experience, not entirely… but visually a mess compared. Anti aliasing on max is far from working at all. And blinding by the white wash. Currently a 3080TI oc z590 mobo i9-11900k. We just have to wait and hope
An update to this. I did two flights today, the first in south west Australia all by my lonesome with no clouds and the second with a group of 20+ flyers over the centre of Tokyo with few clouds, so two extremes of FS2020 load on my system (i9 @ 5.2GHz HT Off / 32GB RAM / 3080 / 1TB SSD). My settings are OXR/TAA 80/100 and FS VR graphics settings ultra for the top half and high for the bottom half.
Example screenie of just how much is being drawn during the second flight follows. No FPS counter on but I would have been getting about 25 FPS and it was smooth and stutter free.
While the second flight was pushing my system much harder, it is very rare that I fly VR in such an area on a group flight. The first flight is much more representative of how I fly in VR and for that my CPU is roughly 50% more powerful than what it needs to be to run that flight smoothly. Admittedly, even with HT turned off, I still have 10 cores to throw at FS2020 but the 4 cores with HT enabled on the i7 7700K should give enough core and thread power to running SU5 with some spare capacity.
TL;DR - run task manager to see if your CPU is regularly maxing out in flight scenarios you normally fly. If it is maxing out regularly, as in worse than my second flight CPU performance graphs, then a CPU upgrade might be worth it but, if not, then the GPU is what you should be looking at because high res VR loves GPU.
I fear thats wrong. In my investigations I see that msfs always use 2 cores to the max (main thread, render thread) while the rest is handling the os stuff:
Now, exactly the same situation but running 6 cores. Big surprize! You see again 2 cores maxed out (the msfs threads) and 4 cores doing the os stuff at slightly lower utilisation.
All the things above where done with SU5H2, same TBM930, same location (west cost africa near Dakar).
Conclusion: As long as msfs only uses two cores for the important threads the fastest single core rule still counts.
Furthermore I found that the CPU is always running about 1ms behind the GPU, no matter how fast the CPU will run. Running them faster give you more and more fps, on my system 1fps per 100Mhz cxore clock. But the GPU is always a ms ahead.
The game is still CPU limited for GPU 3080 and above. I assume that even if there were a 10Ghz cpu there would also be 2 cores permanently busy at 100% while the rest are dawdling around.
What you observe are some tricky placebo’s we got from SU5 injected via a memory usage corruption.
I can’t say I have ever seen even 1 core maxed out since SU5, so I find your results intriguing. Is it perhaps that at 6 cores and lower that 2 cores do indeed get loaded to 100% because FS doesn’t have enough cores to spread the load more evenly? Do you get the same 2 cores maxed when you run your i9 11900K with all its cores (8 I believe) enabled? What about with HT on? Are the tests you are running in high res VR or, at the very least, 4K to represent what should be GPU bound conditions?
It’s late evening here in Oz, so I’ll run some tests with 6 and 4 cores tomorrow and see if I can repeat your results. I will also try to emulate the OP’s CPU config (eg. 4.8GHz 6 core + 6 HT) and see if that makes any difference.
It would also be great if anyone else can post up some CPU graphs of when FS2020 is running for them, to get some more data points. I am genuinely curious as to why we are seeing such a difference in max single core usage.
Do you notice the different load on core 8? It is no more maxed out. It seems that some load is shiftet to core 5. The spike on all cores is probably the moment when switching the modes.
But after switching wmr off core 8 is again under full load (first screenshot).
Now I’m a little insecure to be honest.
Most likely it seems to me, that we are actually GPU limited under VR, here the GPU utilization is at 100%. In pancake mode, on the other hand, it looks as if the CPU is limiting because the GPU utilization is only 76% (fps limited to 60 in ncp), at least with this resolution (3440x1440).
I am using a 5900x 3090 kingpin and 32GB or ram. All settings on ultra and render scales set to 100. Basically everything maxed out. With my G2 its running crazy smooth. I just wish i had a higher resolution headset now lol.
I tried to apply an undervoltage on my GPU, -200 and +1000 on memory, the graphs look like this now, taken on the runway.
FPS are still at 15 maybe little slightly more stable…
but i am at max everywhere, so i may lower some settings also.
Personally as of right now I wouldn’t go spending tons of cash purely for chasing amazing VR performance. By all means upgrade your CPU but don’t expect to get massive gains for VR. I had an i7 7700K CPU paired with an RTX 3080 and earlier this year I upgraded to an 5600X CPU and whilst I definitely saw very noticeable improvements in performance across the sim I didn’t see the huge performance increases in VR I was expecting.
VR in the sim is currently such a strange beast. There are so many variables and people with identical hard ware spec are reporting very different experiences. So far this year its not been unusual to hear stories of people spending the earth on new PC hardware for the sim and VR only to be underwhelmed by what they got for the money. That was certainly my experience. Like I said though I was however pleased with the performance gains I got elsewhere by upgrading my CPU. So I would say its definitely worth it from that point of view.
Incidentally The 5600X is a very good CPU for the sim. I know lots of people here recommend the 5800X but almost all the general PC hardware reviews out there recommend either the 5600X or the 5900X. For almost everything the 5600X (6 core) and 5800X (8 core) have identical performance only the 5600X is cheaper. If more cores are what you think you need you might as well get a 5900X
Since getting my 5600X the only time I’ve wished I had more cores is for recording VR gameplay using the Quest mirror software and OBX. If you seriously need to multi task like that then 6 cores struggle. For just playing the sim the 5600X is very, very good value for money.
tha’s what i did, i upgraded from 2600X to 5600X after having installed this RTX 3070
then i had SU5 and 50% more FPS (22 to 36), and then i had W11 and to my big surprise i got double FPS (36 to 75), nobody understood why…
so now i pushed all my settings to max (Ultra etc…) and i try to just optimize the new limit which is clearly the GPU, and hoping to limit the small stuttering still present, but i can also reduce some MSFS settings for sure…
as proposed by @sehnsedahamses i just applied an UV tuto, not the one u linked, but its difficult for me to understand which values to choose for Volt and freq, i understand how to change the curve, but which values could be the optimum choice ? i will try again with 950mV and 1950 Mhz knowing that the current max freq is 2025 Mhz
and i didn’t understood clearly how to check if it is a stable setting…
Thanks for doing these extra tests. It seems that FS changes the way it load spreads if more cores are available, which is a good thing IMO.
I ran a few extra tests myself, the first one at i7 7700K equivalent mode (6 cores + 6 HT) and the second one with just 6 cores to see if I could duplicate you seeing two cores running at 100%.
As you can see, the CPU load gets quite evenly spread on mine, even with only 6 cores running. Also the max single core load with a fixed view seems to be about 60%, which is the same as I am getting with my 10 core HT off at 5.2GHz configuration. As you say, perhaps it is the FPS cap you are using that is giving the GPU a break and hence allowing one CPU core to get up to 100%.
It looks like this applies to me also. I do not seem to be getting any noticeable stutters on this power throttling. Nonetheless, I will check out the undervolting info you posted and see if I can get rid of that throttling. Thanks for posting this up.
I tried out undervolting my GPU with good success. I have set 860mv as max voltage which has resulted in a power consumption drop of 40-50W, GPU temp is down 4C and no more PerCap limits in-game. Smoothness seems similar, perhaps because I was only just power limiting before, but I’ll take the power saving and lower GPU temp any day!
When upgrading don’t forget to consider available power supply. IMO anything less than 1000W power may limit performance, especially with the i9 and a 3070 on a z490 mobo. A new water cooler maybe as well? Render scaling for G2 on OXR is VRAM limited - on my 3090 I’ve seen VRAM usage at 18/24Gb. While expensive, if reasonable VR desired, get as much VRAM as can afford, then match other compnents to it.
A potent 850W power supply will do well. I have 850W evga G1+ and an max system load of about 600W from my wallmeter including monitor and peripherals. Anyway, haven 1000W wont not hurt.
You need an gfx heavy application like 3Dmark time spy for example.
Run it full screen in loop mode.
Max out your power limit slider in afterburner. Unlock the temp binding and set temp treshhold to 84C.
Start at about 900mV/1900Mhz. Monitor your power limit. If you hit it frequently reduce voltage further in small increments until your system becomes unstable. Then go up one or two Voltage steps to make it stable again.
If you at 875mV still hit your power target, reduce Mhz by 15Mhz. Try again and so on until you find stable settings.
If you like to reduce wattage and card temps try 850mV/1800Mhz.
It is a little trial and error but after a while you get a feeling for that precedure.
I have two settings for my 3090, one for ambients up to 23C with 900mV/1920Mhz and one for hot summerdays with 875mV/1875Mhz. But honestly, fps will change in homöopathic doses between this settings.
Note that your GPU still boost gently by 15 or 30 Mhz when temps allow it.
