would be interesting to see real 4k in your case... 

my TV monitor does "only" 4K: 3840x2160. if I select 4096 x 2160 pixel in MSFS I suppose my GPU interpolates that down to my TV max resolution of  3840x2160.

@ 4096x2160 = 144 fps. I don't usually need all that anyway because I fly only in VR, @ 2160x2160 per-eye, totalling approximately 4.67 million pixels per-eye, or 9.33 million pixels overall. which is still more than those 8.3 million pixels of the highest resolution 4K @ 4096 x 2160 monitor.

3 hours ago, Cpt_Piett said:

@turbomax it seems that Noel currently holds the Avsim FTV record with a value of 99.73 😄

he needs to be better than 100% FTV to set a new record. 🤣

Default Beech Baron, 4K, TAA, DX12, LOD/TLOD 200. framegeneration, SU15 beta

1 hour ago, turbomax said:

he needs to be better than 100% FTV to set a new record. 🤣

Impressive… most impressive 😄 And the FTV award goes to…. Turbo!! 🏆

Of course I’m not going to ask if that was a final approach to EGLL filled with AI traffic 😉

Edited February 19, 20242 yr by Cpt_Piett

4 hours ago, turbomax said:

perhaps you could repeat your CapFrameX tests one more time now using SU14? I hesitate to leave this SU15 frametime wonderland as I haven't had a crash yet, but then again I fly typically only 1/2 hour GA VR.

Been thinking about doing this (but then forgot about it), thanks for the reminder 🙏 

8 minutes ago, Cpt_Piett said:

I’m not going to ask if that was a final approach to EGLL filled with AI traffic

close! 🤣: France LFVM - LFVP

"final approach to EGLL filled with AI traffic"  I'll leave that to you to benchmark in SU14 😃

but then again, Noels monitor is only: 3440x1440= 4.9 megapixels

my 4K TV monitor: 3840x2160= 8.3 megapixels

7 minutes ago, turbomax said:

close! 🤣: France LFVM - LFVP

"final approach to EGLL filled with AI traffic"  I'll leave that to you to benchmark in SU14 😃

Hey what’s that program you’re using again? Looks like a useful overlay!

Re: resolution and contributing to the mines bigger than yours discussion - I did for a while run in 6144/3240 using DSR. That’s 19,9 mexapixels 😁 Went back to native resolution of 3840/2160 though as I could hear the GPU screaming…

4 minutes ago, Cpt_Piett said:

contributing to the mines bigger than yours

that wasn't my point: but graphics performance testing needs similar conditions. after reducing my resolution down from 4K  I got up from 144 fps to 200 fps with everything else the same. 

a nice side effect after all this testing with different resolutions: it finally dawned on me I might permanently reduce my full screen PC 2D resolution from 4K to something like 1K since I don't need that anyway when all I fly is in VR. will have to see if that saves some fps/frametime.

the reason I went through all of this was to make users aware of the importance of frametime when all we traditionally do is focus on fps frames/per second.

my testing was inspired by some good articles, especially:

Inside the second: A new look at game benchmarking by Scott Wasson (who started frametime analysis 13 years ago)

https://techreport.com/review/inside-the-second-a-new-look-at-game-benchmarking/

 

"After almost 20 years, GPU benchmarking is moving past frames per second"

https://www.extremetech.com/gaming/154089-after-almost-20-years-gpu-benchmarking-is-moving-past-frames-per-second

Analysing Stutter – Mining More from Percentiles:

https://developer.nvidia.com/content/analysing-stutter-–-mining-more-percentiles-0

presentmon: Intel's latest benchmarking tool which interfaces perfectly with CapFrameX:

https://game.intel.com/us/stories/intel-presentmon/

 

 

Edited February 19, 20242 yr by turbomax

3 minutes ago, turbomax said:

that wasn't my point

Thought so, just messing around. And I believe the app you’re using is called Intel^® PresentMon

the new metric buzz word introduced by Intel's Presentmon is: "GPU busy" in contrast to "CPU busy", overlayed in one single graph to identify CPU/GPU bottlenecks. that's a NEW in Presentmon.

13 minutes ago, turbomax said:

the new metric buzz word introduced by Intel's Presentmon is: "GPU busy" in contrast to "CPU busy", overlayed in one single graph to identify CPU/GPU bottlenecks. that's a NEW in Presentmon.

Interesting! Was briefly looking at the intel site and it wasn’t immediately apparent what GPU busy refers to. Is it similar to the dev mode overlay? Time to do some (more) reading, thanks for all those links!

39 minutes ago, turbomax said:

the reason I went through all of this was to make users aware of the importance of frametime when all we traditionally do is focus on fps frames/per second.

Exactly. I think we’ve been discussing it here on Avsim MSFS since autumn 2022. It’s a much better value of smoothness and “holy grail” performance than fps, and certainly relevant for beta testing. I’ll post a SU14 benchmark after I finish my current flight. 

Edited February 20, 20242 yr by Cpt_Piett

we can consider ourselves lucky to now finally have both: great fps in the range of 150 AND constant frametimes up to 99% FrameTimeVariance FTV at the same time plus superb graphics to top it off. it used to be either/or. Asobo has done their job, it's now upon the 3rd party devs to optimise some of their inefficient scenery addons.

in closing: we all know stutter-free state-of-the-art graphics are inevitably tied to an unrealistic flight model on rails. 🤣

Edited February 20, 20242 yr by turbomax

SU15 beta vs SU14 performance comparison

RV-14 final approach rwy 16R YSSY. FT YSSY + Orbx Sydney cityscape + FSLTL. Weather preset: few clouds @ noon. 

Ultra settings w/ 4K | HDR | DX12 | FG | DLSS super resolution @ ultra quality (0.80x native resolution i.e. render resolution 3072x1728)

SU15 beta

SU14

Summary: 

Avg. FPS: SU15 beta 77.2 | SU14 106.9

Frame time variance (FTV) <2ms: SU15 beta 97.94 | SU14 76.11

Even though the average FPS was significantly higher in SU14, the frame time variance was considerably less in SU15 beta i.e. more smooth / less stutters. 

I'll finish off with a quote from a Digital Trends article that perhaps will give more insight into the difference between frame time and FPS:

Quote

It’s important to look at both frame rate and frame time when evaluating a game, and benchmarking tools like CapFrameX make it easy to chart frame times. One is not objectively better than the other. They’re looking at two sides of the same coin. Frame rate tells you the overall performance level, while frame time says more about the experience of actually playing the game.

 

Edited February 20, 20242 yr by Cpt_Piett

Not that it changes your conclusions, but the axes scales are not the same for your comparison. This makes the SU14 spikes twice as large what is used for the SU15 beta graph. (They will still be large compared to the SU15 beta spikes, but you should at least show them using the same scale.)

11 minutes ago, Donstim said:

Not that it changes your conclusions, but the axes scales are not the same for your comparison. This makes the SU14 spikes twice as large what is used for the SU15 beta graph. (They will still be large compared to the SU15 beta spikes, but you should at least show them using the same scale.)

Thanks for pointing this out! I edited the post with both captures showing the same scale.