[REPORTED]VR Performance OB 2.5.6 massive FPS loose - Page 29 - ED Forums
 


Notices

Reply
 
Thread Tools Display Modes
Old 03-26-2020, 08:30 AM   #281
Worrazen
Senior Member
 
Join Date: Oct 2014
Location: Slovenia
Posts: 1,278
Default

Quote:
Originally Posted by Py View Post
BIGNEWY,

People are trying to help by posting specs and performance measurements and observations, but every one is different and I can't imagine that it is easy to compare and use this information.

I think it would really be worth ED's time to implement a basic benchmarking function into DCS.

Then everyone can perform the same controlled tests, making comparisons actually useful both for users tweaking options and for ED to collect performance data.
You can also record the system information and log all the important stats during the test.
Even if you just play back a reference track file with performance logging, it would be a good start and may help you solve the current performance issues.

If you make the results packageable then users can post their results in forum posts when there are issues (as they do now with track files), but you will have all the information you need in a standard format.
Maybe have an option to send the result directly to ED for gathering info from a wide range of sytem setups.
Indeed, including a draw call counter. What's also the surest way is like a "huge" special mission which would test a bazillion of scenarios and would perhaps run overnight and some, but we can probably tone it down to some key scenarios, perhaps a quick version for cases like this with some key perf data relevant for users.
Combinations combinations combinations. DCS has so many systems, and depending on the mish-mash you're happen to be into you'll get different results even if you're on the same exact specs, you'll get different results if you use MAV or TGP for example, MAV is more expensive on GPU, that small detail would technically invalidate a test for example.
This would be beyond tracks, this has to be an actual component built for it specifically, running the scripted tests and more than just general tests but rather split by components to figure out their impact individually and be able to home down on culprits, automatically gathering perf data into report, calculating and giving you a % of difference versus another report that you took in another version (if that version supported this benchmarking at all). It could perhaps even be a separate utility outside DCS because for a more surer tests it should researt DCS and clear up standby memory before each new test so you can simulate the worst-case cold boot scenario for every test and that's should be the target, because standby memory is just a convenience IMO, but we've grown so used to it that we take it for granted, if it works nice on cold boot it should work even better after you've been playing for some time.
I might post a draft idea in a separate thread in the future, perhaps I may just make a draft mission that tests some limited things as proof of concept.

I got some nice clues up will post it today hopefully it helps ... there may be a common problem to this.

I should have been on hiatus by now, but after successfully putting in another 8GB of RAM finally, so much overdue, to get me to 32GB I just couldn't miss to not test it and the the month of free trials wouldn't let me go, it's great bug hunting opportunity on modules I never planned buying. Things I do for DCS
__________________
Specs: Win10 x64 1607/14393, 1440p@75"32 - CPU: Intel Core i7 3820 @ 4.0GHz - GPU: Radeon RX 480 8GB - RAM: 32 GB, DCS SSD Samsung 860 EVO 250GB, Saitek Cyborg X/FLY5 joystick.
Modules: A-10C, F/A-18C, Mig-21Bis, M-2000C, AJS-37, WW2AP, CA, FC3.
Terrains: NTTR, Normandy, Persian Gulf

Last edited by Worrazen; 03-26-2020 at 08:59 AM.
Worrazen is offline   Reply With Quote
Old 03-26-2020, 09:02 AM   #282
BIGNEWY
ED Community Manager
 
BIGNEWY's Avatar
 
Join Date: Aug 2011
Location: UK
Posts: 20,073
Default

Quote:
Originally Posted by Py View Post
BIGNEWY,

People are trying to help by posting specs and performance measurements and observations, but every one is different and I can't imagine that it is easy to compare and use this information.

I think it would really be worth ED's time to implement a basic benchmarking function into DCS.

Then everyone can perform the same controlled tests, making comparisons actually useful both for users tweaking options and for ED to collect performance data.
You can also record the system information and log all the important stats during the test.
Even if you just play back a reference track file with performance logging, it would be a good start and may help you solve the current performance issues.

If you make the results packageable then users can post their results in forum posts when there are issues (as they do now with track files), but you will have all the information you need in a standard format.
Maybe have an option to send the result directly to ED for gathering info from a wide range of sytem setups.
I agree it would be nice to have a benchmarking mission or track replay.

The tracks and data provided by users is very helpful the more detail we get the better, there are so many configurations out there it can be difficult sometimes to pin down a problem.

Thanks
__________________
BIGNEWY
Community Manager Eagle Dynamics
Windows 10 Pro x64, NVIDIA MSI RTX 2080Ti VENTUS GP, INTEL i7 4790K @4.4 GHz( Cooled by H100i ), 32GB DDR3 @1866 , Asus Z97-AR, TM Warthog, Jet provost rudder pedals, VIVE Cosmos
BIGNEWY is offline   Reply With Quote
Old 03-26-2020, 10:09 AM   #283
stuartaston
Senior Member
 
stuartaston's Avatar
 
Join Date: Jan 2009
Location: UK
Posts: 1,023
Default

so one thing i spotted recently , on 2.5.6 (having turned off shadows, all of them and mirrors) frame rates are pretty good, what is unusual is the *lack* of blurring (double vision) you get... so in 2.5.5 (same set up no mirrors etal) you get some "double images" especially when "motion vector" is operating, running exactly the same scenarios in 2.5.6 rock solid single image all the way through..
__________________


SYSTEM SPECS: Hardware Intel Corei7-9700K @ 5.0 GHz, 32Gb RAM, EVGA 1080ti FTW 11Gb, Dell S2716DG, Thrustmaster Warthog + MFG Crosswinds V2, HP Reverb Pro SOFTWARE: Microsoft Windows 10 Pro x64, VoiceAttack & VIACOM PRO, TacView, CombatFlite

VR Stuff: My DCS VR Settings FPS Bump in VR Shaders MOD for VR Making an IC compliant version of the Shaders MOD

Last edited by stuartaston; 03-26-2020 at 10:11 AM.
stuartaston is online now   Reply With Quote
Old 03-26-2020, 10:15 AM   #284
Fisu_MAD
Member
 
Fisu_MAD's Avatar
 
Join Date: Jan 2019
Location: Madrid
Posts: 197
Default

Quote:
Originally Posted by imacken View Post
Thanks Pikey, but I’m confused. Why are you quoting monitor fps in a VR thread? Why do you say VR is about half those? There is no direct relationship between the 2.

Sorry if I’m missing something obvious.
+1
__________________
Update:MSI Z390-A PRO, i7-9700K 3.6Ghz liquid Cooled, Corsair Vengeance DDR4 64GB, Asus Dual GeForce RTX 2070 SUPER EVO OC Edition 8GB GDDR6, 970 EVO Plus 1TB SSD NVMe M.2, HOTAS TM Warthog, HP Reverb
Fisu_MAD is offline   Reply With Quote
Old 03-26-2020, 10:24 AM   #285
Pikey
Veteran
 
Pikey's Avatar
 
Join Date: Sep 2009
Location: Reading, UK (GMT)
Posts: 4,096
Default

Quote:
Originally Posted by imacken View Post
Thanks Pikey, but I’m confused. Why are you quoting monitor fps in a VR thread? Why do you say VR is about half those? There is no direct relationship between the 2.

Sorry if I’m missing something obvious.

Correct, VR was my next step, but it has a CPU and graphics overhead. So the 2D is your "never will be exceeded numbers" and it's always been at maximum half of those numbers, there is a relationship, it's just not as simple as "/2" but it is useful for working out what you might achieve in VR and where issues arrrive before you add the complexity of VR on top of the configuration. And it's very much close to /2 for me, I can say that I can see where my 45FPS comes from in single player, and how I get less once you add it to MP. If you see less than 90 FPS you know you will begining to see less optimal numbers. Additionally, there is a Multiplayer overhead for some unknown reason, probably due to pushing the network cards offloading to the CPU as well. Your optimisation must start from the 2D, because it's the same scene being rendered twice and the simpler that is, the more overhead room you have for VR. Of course, its not the same, but most folks using VR will know how much it eats and have loaded up in 2D to tinker with settings. And I will do the next step, but I began here and it took so long to do the basic data to compare against, I thought I'd share early, as well as a personal reference.

Also +1 for a standardised benchmark tool or even scene/mission. Even without a "tool" per se, just the process to be followed could be settled on for now.
__________________
___________________________________________________________________________
SIMPLE SCENERY SAVING * SIMPLE GROUP SAVING * SIMPLE STATIC SAVING * PLAYLIST
===========
Pikey is offline   Reply With Quote
Old 03-26-2020, 11:38 AM   #286
Pikey
Veteran
 
Pikey's Avatar
 
Join Date: Sep 2009
Location: Reading, UK (GMT)
Posts: 4,096
Default

OK on to the VR stuff, brought ahead slightly rushed, I narrowed down the modules I looked at to the top and bottom performers and took screenshots of the in game FPS, the Nvidia overlay and FPSvr tool.


Same mission. The FPS results are broadly in line with the way 2D presented, except the Hornet fell down a bit in relation to the top and bottom performing modules. I believe there is a change in performance when switching module, that is most likely to do with textureloading/memory, so take that into account. However the data still shows the 2D>3D relationship and gives a very good idea of

1. The VR overhead
2. Module differences (significant for VR users)
3. Direction of view and link to terrain
4. Something new - CPU frametime on look direction



The 2D and 3D FPS timings showed relationships on module performance. Just looking at the best performing and worst (best being C-101 and worst was JF-17, with Hornet middleground and A-10 being lowish)

C-101 2D: 120FPS, VR: 52
Frametimes both in the yellow. The data point I have great interest in is the sideways look (the picture attached looking right the 45FPS top right) Something I cannot explain is the CPU frametime peaking and going up. Why? Returns to normal when looking forward. A difference of 5.5ms 16ms framtime compared to 21ms. Same plane, it was faithfully repeating this and all planes did.

JF-17 2D: 80, VR: 36
GPU frametime is now in the red, 25ms for the majority of the time, thats 6ms higher than the C-101 or if you like percentages, FORTY PERCENT difference. Why? GPU went down when looking sideways, CPU went up to nearly 26ms. Why?


There's only two things I know for sure from this. Modules are very much in play with the VR differences and both CPU and GPU interact with looking in certain directions.

Points that aren't easy to absolutely say for certian, but GPU appeared to be much higher on planes with complex instrumentation when looking ahead.

I'll get back to work proper, but have a look at doing this properly this evening. At the very least I discovered which modules perform better in VR.
Attached Thumbnails
Click image for larger version

Name:	2020-03-26 11_50_22-Digital Combat Simulator.jpg
Views:	43
Size:	394.2 KB
ID:	230691   Click image for larger version

Name:	2020-03-26 11_52_06-Digital Combat Simulator.png
Views:	42
Size:	3.51 MB
ID:	230692   Click image for larger version

Name:	2020-03-26 11_53_11-Digital Combat Simulator.png
Views:	36
Size:	3.59 MB
ID:	230693  

Click image for larger version

Name:	2020-03-26 11_54_35-Digital Combat Simulator.png
Views:	36
Size:	3.06 MB
ID:	230694   Click image for larger version

Name:	2020-03-26 11_56_28-Digital Combat Simulator.png
Views:	32
Size:	3.42 MB
ID:	230695   Click image for larger version

Name:	2020-03-26 11_58_07-Digital Combat Simulator.png
Views:	29
Size:	3.44 MB
ID:	230696  

__________________
___________________________________________________________________________
SIMPLE SCENERY SAVING * SIMPLE GROUP SAVING * SIMPLE STATIC SAVING * PLAYLIST
===========
Pikey is offline   Reply With Quote
Old 03-26-2020, 11:44 AM   #287
Mr. Big.”Biggs”
Member
 
Join Date: Oct 2019
Posts: 288
Default

I was in MP yesterday flying F14 in several crowded servers with bellow setup and good ping.
When taxiing, it was stutter fest out front windscreen but out side windows it was butter smooth.
As long as I kept my eyes to the side I could taxi. Not sure if this helps but throwing it out there.
Regards
__________________
I9 (5Ghz turbo)2080ti 64Gb 3200 ram. 3 drives. A sata 2tb storage and 2 M.2 drives. 1 is 1tb, 1 is 500gb.
Valve Index, Virpil t50 cm2 stick, t50 base and v3 throttle w mini stick. MFG crosswind pedals.
Mr. Big.”Biggs” is offline   Reply With Quote
Old 03-26-2020, 01:19 PM   #288
MonnieRock
Member
 
MonnieRock's Avatar
 
Join Date: Mar 2005
Location: Neptune Beach, Florida
Posts: 456
Send a message via ICQ to MonnieRock
Default

Quote:
Originally Posted by Py View Post

I think it would really be worth ED's time to implement a basic benchmarking function into DCS.

I agree !

See my post here
DCS World Time Demo Benchmarking Tool

Happy Simming,
Monnie
__________________
Rack Rig: Rosewill RSV-L4000 | Koolance ERM-3K3UC | i7-4960x @ 5ghz w/EK Monoblock | Asus Rampage IV Black Edition | 64GB 2133mhz | SLI TitanXP w/ EK Waterblocks | 2x Samsung 850 Pro 1TB in RAID 0 | Lepa 1600w PSU | Windows 10 Pro 64bit | TM Warthog HOTAS w/30cm Extension | MFG Crosswind Rudders | Obutto R3volution | HP Reverb

Last edited by MonnieRock; 03-26-2020 at 06:11 PM.
MonnieRock is offline   Reply With Quote
Old 03-26-2020, 07:13 PM   #289
imaxx
Junior Member
 
imaxx's Avatar
 
Join Date: Apr 2018
Location: Graz/Zeltweg -Austria
Posts: 87
Default

Quote:
Originally Posted by BIGNEWY View Post
Hi

We understand that it is a problem and I have shared the feedback with Kate and Matt in my weekly report.

We are working on remaining memory leaks, we also have FPS reports open for individual issues where we have found problems and we have a VR focused report open that we are working on regarding 2.5.5. vs 2.5.6.

I have included some of the graphs I have produced for our VR report,

F-18 free flight and F-5E free flight as comparisons between 2.5.5 and 2.5.6

When I have news I will share it with you all, we are planning an open beta update next week if tests go well.

Thanks
OOOh, I am really looking forward to this.

Thank you @BIGNEWY
__________________
i7-8700K / 32GB RAM @3000 / 1TB M.2 SSD / Asus RTX 2080 Ti OC
VR- Gamer: Valve Index
Flight Gear: Thrustmaster HOTAS Warthog / Thrustmaster TPR / 3x Thrustmaster MFD / Jetseat
imaxx is offline   Reply With Quote
Old 03-27-2020, 02:39 AM   #290
Sedlo
Member
 
Sedlo's Avatar
 
Join Date: Aug 2017
Posts: 562
Default

.

Last edited by Sedlo; 03-27-2020 at 02:45 AM. Reason: deleted wrong forum
Sedlo is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

All times are GMT. The time now is 04:37 AM. vBulletin Skin by ForumMonkeys. Powered by vBulletin®.
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.