F14 frame rate thread

[Vertigo72]

F-18 7500 ms, 163 drawcalls;

F-14 1720 ms, 91 drawcalls.

 

Can you do the same for the interior? Im not so much interested in comparing with the F18 as having a ballpark idea how much Vulkan might help.

[Vertigo72]

I've disabled vsync in both game and Amd settings. But I think it's an AMD issue. Still wondering that no one else has come forward.

 

Strange that deleting the shader folder had no effect on you. Without deleting the shader I can go from "low" to "high" and the appearance and FPS will remain the same.

 

I have not tested this exhaustively, there may be scenerio's where I would have to do that too, but I got the same results when I tried. I have also noticed changing settings doesnt always seem to take hold, but it seems to be enough for me to restart DCS. Im using the steam edition if that would make any difference.

 

Regarding single core performance and higher FPS, I made a quick benchmark with CPU-Z:

 

Sure, there are plenty of tests out there comparing single thread cpu performance and in general, and particularly in games, Ryzen performs slightly lower than recent i5s per clock, so I did and do expect that to be the case in DCS. Add to that the generally slightly higher clockspeed, and usually slightly better OC potential and intel is a logical highend choice (if you dont account too much for the price).

 

But you never know how applicable one test is to a given app. Is the test or the app FPU heavy or not? Is branch prediction an important factor or not, is it cache latency bound or cache size sensitive, does it use AVX(2), etc, etc. A single number cant tell you all that and DCS performance on ryzen could be significantly worse or significantly better than on those benchmarks, the only way to know for sure is to test. The same applies within the intel product range, there could be large differences between various models like i5 or i7 at the same clockspeed if L2 or L3 cache size is an important factor, or it could make no difference at all.

 

FWIW, i think cache size does matter. It does seem i5s, particularly older ones, do get beaten by i7s with similar ram subsystems. See Radokills results.

Edited May 4, 2019 by Vertigo72

[Supmua]

I usually play with dual monitor at 2560x1080, so my typical FPS would be lower than this.

 

My results with 1920x1080-single monitor/high settings preset/zoomed out cockpit view:

Mirrors OFF: 82 FPS

Mirrors ON: 70 FPS

 

9900K at 5.0GHz (all core)

ASUS DUAL 2080Ti (autotuned with Afterburner)

32GB DDR4 3600 (XML on)

 

Shore-based Takeoff Tutorial mission tanks the FPS more with mirrors on, btw. A really good stress test for VR.

Edited May 4, 2019 by Supmua

[Vertigo72]

I usually play with dual monitor at 2560x1080, so my typical FPS would be lower than this.

 

My results with 1920x1080-single monitor/high settings preset/zoomed out cockpit view:

Mirrors OFF: 82 FPS

Mirrors ON: 70 FPS

 

9900K at 5.0GHz (all core)

ASUS DUAL 2080Ti (autotuned with Afterburner)

32GB DDR4 3600 (XML on)

 

Shore-based Takeoff Tutorial mission tanks the FPS more with mirrors on, btw. A really good stress test for VR.

 

Wow. Those are really high numbers. It might be worth double checking those results after deleting the fxo and metashaders folders as cobra suggested.

 

If those numbers stand, Im a little confused as to why you have such high numbers. From my testing, GPU performance is all but irrelevant in this test. Your CPU is about as fast as it gets today, so high numbers are no surprise, but I wouldnt expect it to be 30-40% faster than my rig and substantially faster than any other 5GHz intel rig so far. Maybe the 3600 memory does play a role?

 

If you feel like experimenting a little by downclocking cpu, gpu and memory and I would love to hear the results.

[Supmua]

My rig benchmarked about average with other 3D benchmark apps. I don't think I'd call 70 FPS a high number either especially in a non-Ray traced situations at 1080p for this CPU/GPU combination. People with similar setup who really push and manually overclocking their CPU/GPU should get higher numbers than mine.

Edited May 4, 2019 by Supmua

[Vertigo72]

My rig benchmarked about average with other 3D benchmark apps. I don't think I'd call 70 FPS a high number

 

Its not a high absolute number, but its still 10-20 FPS higher than others on this test with what seem similar CPUs (4.9-5 GHz core i5s and i7s). Only one other person reported a result so far that is in the same ballpark, but he used different settings, most notably lower settings for shadows, so that result is not comparable, and thats why Im slightly suspicious your settings may not have been properly applied and you would have to delete those folders to apply them.

 

Then again no one else so far uses 3600 Mhz ram, so that may be an important factor.

Edited May 4, 2019 by Vertigo72

[Supmua]

 

Deleted fxo and metashader 2 folders prior to running DCS. I got roughly similar FPS on avg, high 70s to low 80s. I think I would expect folks with 5.1-5.2 GHz i9 and 2080TI with higher OC rate to hit 90 or at least high 80s.

Edited May 4, 2019 by Supmua

[Cornelius]

Afaik, memory latency is more important in gaming than memory bandwith (as long it's not far too little), as large amounts of data are not permanently moved.

Maybe pay more attention to the timings..?

 

@Supmua

Your Hardware Specs are like elysium to me. There's clearly something wrong with DCS at the moment.

 

Maybe we need something like this to get past 100 FPS:

Edited May 4, 2019 by Cornelius

[Tenkom]

I tested the same settings at 4K resolution.

So at 1080p I had 60 without mirrors and 55 with mirrors.

At 4K I have 56-57 without mirrors and 52-54 with mirrors.

[Vertigo72]

Afaik, memory latency is more important in gaming than memory bandwith

 

There is a lot of truth in that, but the two are not unrelated. All things equal, the higher the memory frequency, the lower the latency in absolute terms (microseconds), and the latter is what matters. Typically you have to relax the CAS latency settings in order to hit higher clockspeeds, but CAS latency is expressed in clock cycles, so despite the increased relative latency (CAS latency expressed in clockcycles) you almost always end up with lower absolute latency (microseconds).

[Vertigo72]

I tested the same settings at 4K resolution.

So at 1080p I had 60 without mirrors and 55 with mirrors.

At 4K I have 56-57 without mirrors and 52-54 with mirrors.

 

That pretty much proves my point that in this test, we are all CPU bound (possibly CPU IO bound). The GPU, especially if you have a fast one, is hardly being stressed at all. Hence you get almost "free" 4x higher resolution. And if you dont run 4k or high MSAA settings, its probably not a good idea to spend 1000 euro on the gpu.

 

Once airborne, this may change significantly, and a fast GPU will likely produce higher framerates than a more modest one, but if you have good enough FPS on the ground, it generally only gets better, and I prefer to have better framerates where they are currently too low, than 100 extra FPS where I already have 100.

 

VR is yet another different story. But Im not touching that as long as even a 2080Ti cant get good online framerates on low end VR sets. I would want good framerates on pimax 5K or 8K in wide FOV. That will have to wait, I dont have the space for the "PC" Cornelius linked :)

[Wali763]

Test

 

My "cheap" system gives me 42/35. My PCs specs are I5-3570, GTX 1660Ti, 24 GB DDR3-Ram at 1600Mhz.

 

What I dont understand is, why GPU and CPU dont seem to be pushed to their limits.

 

See the pic with my slightly different standard-settings. GPU 58% CPU 1/2 14/11%. Shouldnt one be maxed? The only thing, that is maxed is GPU-RAM (6GB), so I turned it off in Afterburner.

 

Since I have a 6GB GTX1660Ti, could a move to something compareable with 8GB (like 1070) raise my fps??

 

Sorry if my question seems stupid

 

EDIT on 20-06-2019:

 

I updated my system now and its proove of those claims, that DCS in CPU-dependent.

 

I upgraded from a I5-3570 to a I5-9600K (not oc-ed) with DDR4-Ram at 2800Mhz, still with the GTX 1660TI I had before and my results went up from ~42/35 to ~77/67. My GPU is now much higher loaded, often maxed.

 

Before upgrading I expected some 15fps, maybe even 10fps, but this result is astonishing. Especially when looking at other game-benchmarks, difference between I5-3570 and I5-9600K are maybe 15% or even less.

 

In DCS things now feel much smoother; in Windows10 though, tehre is no feelable difference.

Edited June 20, 2019 by Wali763

[Donut]

If anything, this thread is stressing the need and importance of implementing Vulkan API.

[Vertigo72]

My "cheap" system gives me 42/35. My PCs specs are below.

 

What I dont understand is, why GPU and CPU dont seem to be pushed to their limits.

 

Your cpu is maxed out. Or at least, one core is. The overall average of all your cores is meaningless, details here in this very thread:

 

https://forums.eagle.ru/showpost.php?p=3899840&postcount=17

 

 

Since I have a 6GB GTX1660Ti, could a move to something compareable with 8GB (like 1070) raise my fps??

 

This entire thread is showing that, at least with F14 and on the ground, all of us are CPU bottlenecked. A faster GPU wont help, not in this test.

[Vertigo72]

Since no one else tested this yet, and I still dont have my ram, I just tried to overclock my old ram. From the default 1066 Mhz (DDR4 2133) to 1200 Mhz (DDR4 2400), which is about as high as it will go. CAS latency only took a small hit.

 

Without mirrors: 51 FPS @stock speed

Without mirrors: 56 FPS @2400 ram

 

With mirrors: 43 FPS @stock speed

With mirrors: 47 FPS @2400 ram

 

A 10% performance boost for a 12% ram frequency increase. And probably closer to 10% ram latency decrease if you account for the increased CAS timing. There may be a point of diminishing returns, and this is only one limited test, but it certainly seems like ram speed matters a LOT. Now I understand better Supmua's results with his DDR4 3600 ram. Those sticks cost an arm and a leg, and if you had asked, I would never have thought it would be worth it, but if your cpu can actually run the ram at that speed, it may be totally worth it for DCS.

Edited May 5, 2019 by Vertigo72

[Flighter]

Hi Vertigo and company, very interesting thread.

My score from fraps on a 30 seconds measurement :

 

No mirrors:

Frames: 1949 - Time: 30000ms - Avg: 64.967 - Min: 61 - Max: 68

Mirrors :

Frames: 1639 - Time: 30000ms - Avg: 54.633 - Min: 51 - Max: 58

 

Test system :

CPU i7-6700K @ 4.6 Ghz

RAM 16GB DDR4 @ 3200 Mhz, CL16

VGA RTX2080ti @ 1.7/2.0 Ghz

[Vertigo72]

Averaging over 30s is a good idea, however, there is an E2 taking off and if you wait too long, there is more variance. Shortly after loading, my framerats dont vary more than 2-3 FPS, so I just mentally averaged it.

 

This is not the most accurate or most consistent test ever devised, but I wanted something simple that required minimal effort to paint a rough picture. And I think it has.

 

FWIW, I wanted to upgrade my GPU, but that seems a complete waste of money with my 1080p display. So instead Im going to upgrade my RAM to much faster ram, and maybe wait for ryzen 3 and hope it provides a decent frequency boost on some of the lower core count SKUs (no point in paying for 8 or 12 cores).

Edited May 5, 2019 by Vertigo72

[Kazansky222]

Specs are 8600k 5.1 Ghz, GTX1080ti +65 core +450 mem. And I play in VR with Pimax 5k+

 

I play with textures on high, everything else off except MSAA which is set at 2x. I use the VR performance mod with the MASKSIZE adjusted to fit the 150 degree FoV I use on the Pimax, and I change the MSAA MASKVALUE slightly to fit the larger sweet spot on the Pimax. in Single player I can run fine, game is always locked to 45fps due to ASW/Smartsmoothing, however Steam VR leaves the resolution at my default 100% setting. In multi player though, performance tanks hard, I still maintain 45fps in all aircraft except the f-14 it dips into low 40s sometimes. And in Multiplayer Steam VR auto resolution adjustment really kicks in a looks pretty bad in comparison to single player.

[Vertigo72]

My new ram modules arrived:

 

Without mirrors: 51 FPS @2133 ram
Without mirrors: 56 FPS @2400 ram 
Without mirrors: 60 FPS @3200 ram CAS 16
Without mirrors: 61 FPS @3466 ram CAS 17 

With mirrors: 43 FPS @2133 ram
With mirrors: 47 FPS @2400 ram
With mirrors: 51 FPS @3200 ram CAS 16
With mirrors: 52 FPS @3466 ram CAS 17

 

Havent tried pushing them to 3600 yet.

Edited May 10, 2019 by Vertigo72