Jump to content
chiliwili69

Measuring rig performance: Common Baseline (for IL-2 v3.010)

Recommended Posts

12 hours ago, BeastyBaiter said:

Ah, it's 200 seconds, read it as 2 minutes. BTW, before we go too nuts with this, we should change tracks. That one is really far removed from normal gameplay, especially given that it goes 2 minutes straight of almost non-stop shooting thanks to unlimited ammo and invulnerability. A good test track should also start and end with basically nothing going on, to account for minor differences in when the game may start the track (it loads unpaused, then pauses 2-3 seconds, then user unpauses).

 

Regarding the track, yes it is quite intensive and it is only a small fraction (the heavy one) of a normal "mission". A normal mission over stalingrad most of the top rigs (like yours) will be at 90fps around 95% of the time, and only in heavy combat your fps will drop from 90fps. So the purpose of the track is to produce a heavy scenario (even artificially with unlimited ammo or invulnerability) to be below 90 fps with medium settings (so we can decrease/increase settings later to analyze). It is just a measuring tool, not a representation of the normal gameplay fps.

 

Here in stalingrad map, when the fps drops to 45 is mostly due to the smoke/fire of the planes (and perhaps damage model). The only complaint about this track is that there is no mountains and it using a collector plane (La-5FN), so I don´t know if people without that plane can run the track.

12 hours ago, MultiDoc said:

Both times all settings in game were maxed out (should I be using different/lower settings ?)

Thanks for posting your test with just monitor, but in the first post the instructions indicate to run it at a certain settings, so we can compare later the results with other tests.

You also need to report the CPU Mark and STMark and the freq. of your CPU. The stock CPU with no OC could be any value from 3.7 to 4.7(turbo). You can see the freq of CPU during the test using the free tool MSI Afterburner. It has nice trendline where you can monitorize many things (CPU temp, CPU clock, GPU temp, GPU load, GPU fans, etc)

Edited by chiliwili69

Share this post


Link to post
Share on other sites
17 hours ago, chiliwili69 said:

Thanks for posting your test with just monitor, but in the first post the instructions indicate to run it at a certain settings, so we can compare later the results with other tests.

You also need to report the CPU Mark and STMark and the freq. of your CPU. The stock CPU with no OC could be any value from 3.7 to 4.7(turbo). You can see the freq of CPU during the test using the free tool MSI Afterburner. It has nice trendline where you can monitorize many things (CPU temp, CPU clock, GPU temp, GPU load, GPU fans, etc)

I just rerun the benchmark using exactly the settings in the first post of the thread, however I can't see anyone running the game in such a low resolution really (1600x900), or am I missing something ?

 

Anyway here is what I came up with:


Frames: 32432 - Time: 200000ms - Avg: 162.160 - Min: 103 - Max: 254

 

As for the CPU clocks, ofcourse it's at 4.7 all cores.

Share this post


Link to post
Share on other sites
On 3/24/2018 at 3:38 PM, TG-55Panthercules said:

 

[edit] I've been holding off doing any testing/flying until they get around to issuing the hotfix that seems almost inevitable with a patch/update of this size

 

Im with you. Same conclusion. Just going to be patient. A few tweaks in order. Something amiss for sure

Share this post


Link to post
Share on other sites
2 hours ago, MultiDoc said:

I just rerun the benchmark using exactly the settings in the first post of the thread, however I can't see anyone running the game in such a low resolution really (1600x900), or am I missing something ?

 

Since this is primarily a benchmark for VR performance, we run this test at the lower resolution to reduce any extra overhead that goes into mirroring the display on the monitor. The game inside of VR will run at the VR headsets native resolution (plus the level of supersampling applied) regardless of the resolution chosen. That setting for resolution is just going to be what's shown mirrored on your monitor.

 

That said, 1600x900 is the resolution to benchmark VR performance. Since you seem to be only testing in-monitor performance and not VR,  you need to read the first post again because in-monitor testing (as opposed to in-VR) uses a resolution of 1920x1080.

 

Madmatt

Edited by Madmatt
Because I can't spell for shit...

Share this post


Link to post
Share on other sites
3 hours ago, Madmatt said:

you need to read the first post again because in-monitor testing (as opposed to in-VR) uses a resolution of 1920x1080.

That´s it. The instructions to run run in monitor are at the final part of the first post.

Share this post


Link to post
Share on other sites

Ok, for my i9-7900X (@ 4.7 GHz if only uses 5 cores) with GF-1080 (8GB) ON MONITOR with the Samuel track the following:

 

 

chili's settings: (default)

Frames, Time (ms), Min, Max, Avg

 25732,    200000,  64, 209, 128.660

 

max. settings: EDIT: @ 1440p

Frames, Time (ms), Min, Max, Avg

 19734,    200000,  51, 160, 98.670

 

the FPS plot:

il2.jpg

 

 

Edited by ZachariasX

Share this post


Link to post
Share on other sites

Updating my scores with the new reference view and also the latest 8.2 version of lefunestes VR enhancement mod.

 

VR Benchmark with 3Dmigoto Mod Uninstalled:

2018-03-26 10:07:12 - Il-2
Frames: 14468 - Time: 200000ms - Avg: 72.340 - Min: 41 - Max: 91

 

VR Benchmark with 3Dmigoto Mod Installed (all effects at default):

2018-03-26 10:12:55 - Il-2
Frames: 13166 - Time: 200000ms - Avg: 65.830 - Min: 43 - Max: 91

 

Comment: Still seeing that odd anomaly of the minimum FPS being slightly higher with the mod installed as opposed to not. Performance impact of running the mod is still a consistent 10% decrease in average FPS.

 

Madmatt

 

Edited by Madmatt

Share this post


Link to post
Share on other sites
On 3/25/2018 at 2:54 PM, chiliwili69 said:

it using a collector plane (La-5FN)

 

Nope, it is LaGG-3 from BoS. I think everybody has it

Share this post


Link to post
Share on other sites

Thanks to all people reporting the test.

 

One of the things I would like to see in this thread is how the fps is affected from 4.7 to 5.3 Ghz overclock in top CPUs, like i7-8700K and more specially in the i5-8600K.

This year there will be many people thinking to upgrade their rig or their CPU, and I think the i5-8600K could be also a very good CPU for IL-2 VR, and a bit cheaper.

 

In this page you can buy a delided i5-8600K will all range of frequencies up to 5.3Ghz:

https://siliconlottery.com/collections/coffeelake

 

Any body here with an i5-8600K?

Can anybody here run a series of tests from 4.0 to 5.3 with 0.1GHz intervals?

 

In the past I was studying the performance gain from 4.0 to 4.7, and it was a steady gain until 4.7. But didn´t went beyond:

 

Share this post


Link to post
Share on other sites

I have an 8600k and I can't push it past 4.8ghz with a -2 AVX offset and remain stable or cool, but I only have a Hyper 212 Evo. It is a really good CPU for IL-2 though and I am enjoying the upgrade from my old i5-2500k. I run it with 3466mhz RAM and a FE 1080.

Share this post


Link to post
Share on other sites

Here are results with a new track:


Frames: 8691 - Time: 200000ms - Avg: 43.455 - Min: 27 - Max: 87
 

CPU Mark : 8810

CPU Single Threaded: 2379

Share this post


Link to post
Share on other sites

Hello !

 

@chiliwili69 I've got a small addendum to the test procedure to suggest. Modifying that many settings all at once can lead to some confusions, from my own experience my first test was run with ASW enabled.

 

Here is a short procedure for Rift users. Ideally, it should be inserted in the step "2. Settings", after modifying SteamVR SuperSamplings settings and before modifying IL2 Graphics settings, or at the beginning of step "3. Benchmarking".

--------------------------------------------------------------------------------------------------------------------------

Rift users, check ASW and SuperSampling (SS) settings are correctly applied in-game :

- While on IL2 login screen, alt-tab to Oculus Tray Tool* (OTT) and set Visual HUD to "Pixel Density" (see here),

- alt-tab back to IL2 and you'll see the debug tool HUD hovering above everything else. Make sure the last line displays "Pixel Density : 1.32". If not, close IL2, close Oculus Home (if opened), go back to step 2 and check SteamVR SS settings (if everything is OK, then just restart SteamVR, then IL2 and Pixel Density should display 1.32 within IL2).

- alt-tab back to Oculus Tray Tool, and now set Visual HUD to "ASW Status", 

- back in IL2, the overlay HUD must read "ASW Status : Not available". If not, switch back to OTT, set ASW Mode to whatever else than it currently displays, then to off. Back to IL2, ASW should be "Not Available".

- finally, back to OTT to set Visual HUD to "Off". You can now proceed to the test. (phew !)

 

*If you didn't install OTT, you can use Oculus Debug Tool instead (you'll find it in Program Files\Oculus\Support\oculus-diagnostics\), see here.

--------------------------------------------------------------------------------------------------------------------------

Share this post


Link to post
Share on other sites
On 30/3/2018 at 10:11 PM, Talon_ said:

I have an 8600k and I can't push it past 4.8ghz with a -2 AVX offset and remain stable or cool, but I only have a Hyper 212 Evo. It is a really good CPU for IL-2 though and I am enjoying the upgrade from my old i5-2500k. I run it with 3466mhz RAM and a FE 1080.

 

4.8 is good with that cooler.

Have you tried to run the test either with VR or in monitor?

 

If one day you go to liquid cooling I am sure you will be able to reach 5.0

Share this post


Link to post
Share on other sites
On 3/30/2018 at 3:11 PM, Talon_ said:

I have an 8600k and I can't push it past 4.8ghz with a -2 AVX offset and remain stable or cool, but I only have a Hyper 212 Evo. It is a really good CPU for IL-2 though and I am enjoying the upgrade from my old i5-2500k. I run it with 3466mhz RAM and a FE 1080.

 

BoX uses AVX, so that effectively means you are running 4.6 GHz in game. I recommend upgrading the cooler. The Hyper 212 is a solid entry level cooler but really not up to the task with an overclocked Intel chip. Intel's frying pans require a much beefier air or liquid cooler than that. I've been pretty happy with the Arctic Liquid Freezer 120 for my 8700k. It's fairly cheap at only $70-ish USD compared to $25-ish for the Hyper 212.

Share this post


Link to post
Share on other sites
On 4/4/2018 at 7:37 PM, ShugNinx said:

I've got a small addendum to the test procedure to suggest. Modifying that many settings all at once can lead to some confusions, from my own experience my first test was run with ASW enabled

Thanks for your observations.

The fact is that I don´t use OTT at all. I used it in the past but I don´t find a reason to use it anymore when SS can be set directly in SteamVR (also for Rift users) and ASW can be switched off with Ctrl+Numpad1. I like a simpler life.

 

In any case I added that note for people using OTT.

 

 

Share this post


Link to post
Share on other sites

Hello !

 

15 hours ago, chiliwili69 said:

ASW can be switched off with Ctrl+Numpad1.

 

Never liked that shortcut, as there is no visual confirmation of the ASW status (I guess the prop glitch in IL2 is a good enough one but it requires to launch the game entirely , and is not appliable to all VR apps).

Plus OTT does a lot more, but it's a matter of personal tastes in the end.

Edited by ShugNinx

Share this post


Link to post
Share on other sites

I find it very easy to tell. The prop artifacts and jittering go away while the fps counter in the top right goes from 50-ish to 89.

Share this post


Link to post
Share on other sites

That''s assuming you're flying a single engine plane, it's harder to tell with a twin engine one. Plus it requires to launch the game entirely, whereas HUD overlay can be displayed on the logon screen. 

 

I'm just providing an "foolproof" method to make 100% sure benchmark settings are optimal, not trying to convince everyone to use this method instead of their usual one.

Edited by ShugNinx

Share this post


Link to post
Share on other sites
2 hours ago, ShugNinx said:

Hello !

 

 

Never liked that shortcut, as there is no visual confirmation of the ASW status (I guess the prop glitch in IL2 is a good enough one but it requires to launch the game entirely , and is not appliable to all VR apps).

Plus OTT does a lot more, but it's a matter of personal tastes in the end.

 

Agree completely, I use OTT with profiles for each of my games for SS, audio switcher, high performance setting, and ASW off , plus audio confirmation each time the game starts that the profile is loaded. It works a champ for me each and every time.

Note: It does need Oculus Home running to apply everything properly, without it running ASW tends to kick on and off repeatedly. Even checking the box to not require Oculus Home, it does this on ASW. Drives me crazy with it kicking in and out.

 

 

 

 

1 hour ago, BeastyBaiter said:

I find it very easy to tell. The prop artifacts and jittering go away while the fps counter in the top right goes from 50-ish to 89.

 

Yep I can tell pretty quickly when ASW is active.

Share this post


Link to post
Share on other sites
5 hours ago, dburne said:

Note: It does need Oculus Home running to apply everything properly, without it running ASW tends to kick on and off repeatedly. Even checking the box to not require Oculus Home, it does this on ASW. Drives me crazy with it kicking in and out.

Ah, interesting ! That could explain why OTT has been failing so much on my system lately (I set Oculus Home to run with admin permission to prevent it to launch anytime I started IL2).

In my case, OTT does fine with SS and default sound output, but indeed has trouble with disabling ASW, and enabling/disabling Windows performance mode.

Share this post


Link to post
Share on other sites
1 minute ago, ShugNinx said:

Ah, interesting ! That could explain why OTT has been failing so much on my system lately (I set Oculus Home to run with admin permission to prevent it to launch anytime I started IL2).

In my case, OTT does fine with SS and default sound output, but indeed has trouble with disabling ASW, and enabling/disabling Windows performance mode.

 

Yeah, run Oculus Home and that should not happen. Be sure and launch the Oculus Tray Tool prior to Oculus Home starting.

Share this post


Link to post
Share on other sites

I have run a number of test on Monitor just to measure the influence of Overclocking over a wide range. I hear in the past some people said that no benefit from OC beyond 4.7GHz, but I just wanted to test it myself.

 

I have run six test with the IL-2 benchmark (3.001 Samuel track) with monitor settings (1920x1080 and no full screen in my 4K monitor).

Specs: i7 4790K; NZXT Kraken X-52 Liquid cooling; 16Gb 4x4 DDR3 2933MHz ; MSI Z97M-Gaming; Asus ROG Strix GTX 1080Ti;

 

The GPU load was below 40% in all the tests.

 

4.0 GHz Frames: 26376 - Time: 200000ms - Avg: 131.880 - Min: 72 - Max: 212
4.2 GHz Frames: 27159 - Time: 200000ms - Avg: 135.795 - Min: 81 - Max: 216
4.4 GHz Frames: 28140 - Time: 200000ms - Avg: 140.700 - Min: 83 - Max: 223
4.6 GHz Frames: 28841 - Time: 200000ms - Avg: 144.205 - Min: 84 - Max: 226
4.8 GHz Frames: 29635 - Time: 200000ms - Avg: 148.175 - Min: 89 - Max: 230
5.0 GHz Frames: 30480 - Time: 200000ms - Avg: 152.400 - Min: 91 - Max: 238

 

It is a steady gain of about 4fps for every 0.2 GHz for this heavy heavy track.

 

If you put that in a graph you have the following:

5acdb951ef6c1_AvgfpsmonitorvsOC.jpg.f8b7c3aa6e1877c6f741bbf2afed225f.jpg

 

If you want to see the fps for every second and every CPU freq the results are:

 

5acdb95667112_fpsinmonitorvsOverclock.thumb.jpg.7df7a68c13b1c21ed02cf698ead23502.jpg

 

I will run a similar test for VR when I have time for Rift and VivePro

  • Thanks 2
  • Upvote 2

Share this post


Link to post
Share on other sites

Once again, thank you for documenting this Chili.

 

I was one of those believing and even telling others that fps gains lessen the closer to 5.0ghz you get. Formed that opinion from eye balling our old benchmarks and noticing how even avg fps results were when comparing those rigs that managed close to 90fps. While those with lower clocks, fps 'seemed' to fluctuate more. Feelings and assumptions rather than fact.

 

Your graph is clearly linear and I expect VR result to also be. If not, however unlikely, I'd find that very interesting. Been meaning to test the impact of (or lack of) cache frequency for a while now. As I expect it to be very low it's been far down on my priority list though.

Share this post


Link to post
Share on other sites

Well, it is the turn to check the performance with the VivePro.

Using default VR test settings with SS=1.7 (170%) I obtained:

 

Frames: 11974, Avg:59.87, Min :42, Max:91

 

I also tested with SS=1.0 and with SS=1.3 (which is the recommended by SteamVR for my GPU) and the results are almost identical. So it is telling me that the GPU is not the bottleneck for the range SS=1.0 to SS=1.7. 

Using SS=1.7 with the VivePro deliver 15.3 million pixels for render. It is like using SS=3.5 in SteamVR with the rift.

 

The specs for test: 

i7 4790K @4.9GHz; NZXT Kraken X-52 Liquid cooling; 16Gb 4x4 DDR3 2933MHz Corsair Dominator Platinum; MSI Z97M-Gaming (micro-ATX MoBo in ATX case); Asus ROG Strix GTX 1080Ti; 

 

BTW, I don´t know if it is fair to use SS=1.7 for the test when using other VR devices (like VivePro or future Pimax). Perhaps we should use a SS which is equivalent in terms of rendered pixels to the Rift with SS=1.7 (SteamVR). Otherwise devices with more resolution than Rift could be unfairly penalized due to the higher amount of pixels to be rendered.

Share this post


Link to post
Share on other sites

Effective resolution should be standardized and I still think we need to change tracks. The fact that altering resolution so dramatically has no impact for you just goes to prove that. We also need to switch to settings that are actually playable. Anything but high clouds and 2x AA is completely unplayable without icons due to graphical artifacting.

Share this post


Link to post
Share on other sites

Gave uncore (aka cache frequency or ring bus) overclocking a go and these are my findings.

 

Setting the uncore ratio to a given number does not equal what cpu-z reports for north bridge frequency. I don't know why and this makes the comparison a bit more difficult. However, while nothing dramatic, it's clear to see there are a few fps to be gained here for those that overclocked their cpu but haven't touched uncore. 

 

If your running a kaby lake and you upped the voltage of your cpu for overclocking, you have also upped the Uncore voltage as the uncore shares the CPU Vcore rail. (I don't know what other cpu's also share Uncore/cpu voltage). Meaning, uncore is already over voltaged so claim the overclocking margin you already have. As always, small steps and the regular stability testing.

 

These are 60 second monitor tests using the Samuel track. My own graphics settings and not vr. I also have the migoto mod installed so results are not comparable to the standard test settings, only with each other. Each of these results is the average of 3 test runs. The cpu is an i7 7700K.

 

 

cpu@4.8, uncore ratio 42, mem@3000, cpu-z shows NB @ 4200mhz
Avg: 132.417 - Min: 76 - Max: 220


cpu@4.8, uncore ratio 48, mem@3000, cpu-z shows NB @ 4600mhz
Avg: 136.050 - Min: 76 - Max: 221.33

And for comparison let's downclock cpu, which unfortunately also lowers NB freq:


cpu@4.2, uncore ratio 42, mem@3000, cpu-z shows NB @ 4000mhz
Avg: 124.022 - Min: 75 - Max: 208

 

 

Edited by a_radek

Share this post


Link to post
Share on other sites

i7 3770K @ 4.6 GHz, 8GB of DDR 3 1866 MHz, MSI 1070 Gaming Core 2075 MHz Ram 4550 MHz

Avg: 47.205 - Min: 33 - Max: 78

Passmark multi 11734 single 2551

and after upgrade

Ryzen 2700X at a slight auto overclock profile from my Asus board with stock cooler, singlecore up to 4.35 GHz, 16GB of DDR 4 3200 MHz, GPU MSI 1070 Gaming Core 2075 MHz Ram 4550 MHz

Avg: 44.015 - Min: 34 - Max: 86

Passmark multi 17663 (about 50% increase) single 2269 (about 11% loss)

Share this post


Link to post
Share on other sites
7 hours ago, Neithari said:

Ryzen 2700X

Many thanks for reporting the first test with the new Ryzens. I assume both using version 3.002 ¿right?

It is a pity that such a great CPU for multi-thread it, is still quite behind for the performance in single-thread which us what matters for IL-2 VR.

The test with your old CPU was also a bit below than expected due to RAM speed, since 1866 is a bit low nowadays.

 

Many people is thinking to use the new Ryzens for IL-2 VR, so it is good that you put some facts on the table.

 

Share this post


Link to post
Share on other sites
21 hours ago, chiliwili69 said:

Many thanks for reporting the first test with the new Ryzens. I assume both using version 3.002 ¿right?

It is a pity that such a great CPU for multi-thread it, is still quite behind for the performance in single-thread which us what matters for IL-2 VR.

The test with your old CPU was also a bit below than expected due to RAM speed, since 1866 is a bit low nowadays.

 

Many people is thinking to use the new Ryzens for IL-2 VR, so it is good that you put some facts on the table.

 

Yeah it was 3.002 and I am using a Rift, but I had to use 1.72 SS to get a "Pixel Density : 1.32". It was on 1.3 with 1.7. My Ryzen board is Asus Strix-F X470 and my i7 3770k is on  MSI Z77A-GD65. Regarding the old RAM: Yeah it was nothing special and I never tried to OC it more than XMP.

With the new Ryzen I have hoped for at least the same performance on average FPS but sadly not happening. I will submit OC values as soon as I have my custom loop set up and my new RAM will be changed from Hynix A Dies CL16 3200MHz to Samsung B Dies CL14 3200MHz. But don't expect anything soon. My new case is sheduled to ship 16.05.2018  and because it's my first custom loop I expect at least a few days working on it. So maybe end of May or early June, depending on my freetime.

 

I hope another one with a 1070 will contribute so we can see how much impact the better GPU of the others have. Best with a i7 8700K with stock and 5GHz OC.

 

Also a quick question: What is the delta number in the sheet? Performing better/worse than expected? And if so how do you expect it? From clock speed?

Edited by Neithari

Share this post


Link to post
Share on other sites
15 hours ago, Neithari said:

With the new Ryzen I have hoped for at least the same performance on average FPS but sadly not happening

Some people reported the test (using a previous track test known as Balapan test) with the first gen Ryzen and they also got low single-thread index and consequently low fps values.

That´s why I was not recommending Ryzen CPUs for IL-2 VR. It is sad to see the 2nd gen Ryzen has not improved that.

15 hours ago, Neithari said:

I will submit OC values as soon as I have my custom loop set up and my new RAM will be changed from Hynix A Dies CL16 3200MHz to Samsung B Dies CL14 3200MHz.

I think you will not notice a significant improvement in IL-2 test with you RAM latencies upgrade, but let´s see.

15 hours ago, Neithari said:

I hope another one with a 1070 will contribute so we can see how much impact the better GPU of the others have

I was using a 1070 before and I upgraded to a 1080Ti. I didn´t see any performance improvement in the test or in game for the usual SS ratios I was using (1.7 in SteamVR).

15 hours ago, Neithari said:

What is the delta number in the sheet? Performing better/worse than expected? And if so how do you expect it? From clock speed?

Yes, the delta is the difference between achieved and expected.

The expected value is calculated from your reported Passmark Single-Thread index and the correlation of all performed tests.

The correlation correlates linearly the Passmark single-thread index with the achieved avg fps of the test.

Share this post


Link to post
Share on other sites

Thanks for the insight. Maybe I get a 4.4 OC on 1 core and can force IL2 to use only that, we will see.

I would have expected a significant performance boost from upgrading to a 1080Ti. Sad that IL2 is so unoptimised that it is so heavily CPU limited and more CPU freq. results in better performance gains than a significant GPU upgrade.

Share this post


Link to post
Share on other sites

You might be better off leaving it stock. It should boost to 4.35 GHz when only 2 cores are in use, that's better than a 4.1 or 4.2 GHz all core overclock as far as BoX is concerned. Ryzen Master also shows which cores are rated for the highest clock speeds, so you could use process lasso to keep BoX on those cores and remove everything else.

Share this post


Link to post
Share on other sites
On 4.5.2018 at 11:39 PM, BeastyBaiter said:

You might be better off leaving it stock. It should boost to 4.35 GHz when only 2 cores are in use, that's better than a 4.1 or 4.2 GHz all core overclock as far as BoX is concerned. Ryzen Master also shows which cores are rated for the highest clock speeds, so you could use process lasso to keep BoX on those cores and remove everything else.

I've read of good results with slight BCKL OCs. The particular one I read just played with voltage to get it stable and upped BCKL a little and let XFR2 and Precission Boost do it's thing. He had a 4.5GHz on 1 or 2 cores with that. And most games profit a lot from memory OC. Some even 10-15 fps to even out the playfield with an i5 8600k @ 5.2 GHz

These 2 things will be my main try after I have it assembled.

Edited by Neithari

Share this post


Link to post
Share on other sites
On 04/05/2018 at 11:00 AM, Neithari said:

Sad that IL2 is so unoptimised that it is so heavily CPU limited

 

Actually it's because the devs optimized GPU utilization that CPU has now become the bottleneck. A couple years ago it was using dx9, and 45 FPS on the ground at 1080p was what many people were seeing, and both the GPU and the CPU would take turns in being the limiting factor, depending on circumstances. I don't think it's fair to say that IL-2 is "so unoptimized".

Share this post


Link to post
Share on other sites

It used to be poorly optimized for modern CPU's and GPU's. Now it's just poorly optimized for modern CPU's. That's a notable improvement, but still isn't very good.

Share this post


Link to post
Share on other sites

I also recall that Jason stated that there are few if any high performance GPUs in Russia available (like GTX 1XXX series and AMD Vega 56/64) so it takes time to equip team with newest technology. I presume there is still plenty of room to improve GPU utilization considering how much raw compute power 1080/1080Ti/Vega56/Vega64 have. But I share the view that CPU is the limiting factor here, especially in multiplayer environment.

 

Share this post


Link to post
Share on other sites

I just tested the VR benchmark with the following setup.

 

AMD Ryzen 2700X (stock)

AMD RX Vega 64 air (Stock)

Windows 10 v 1803

Samsung Odyssey HMD

 

It should be worth noting that WMR devices have their version of reprojection always on, so max FPS will never go much above 45 if a constant 90 FPS cannot be maintained.  Hence the max FPS being comparatively low using WMR devices and the average being around 45 FPS.  The main thing to take away from this result is that I was stuck at avg 45 FPS for most of the track and it always felt smooth due to the reprojection feature.  It is also worth noting that the results for SS set to 130% and 170% are within margin of error on the min FPS.

 

Supersampling set to 132% to match 7.3 mega pixels equivalent to CV1 tests at 170% SS.

Frames: 8935 - Time: 200000ms - Avg: 44.675 - Min: 40 - Max: 52

 

Supersampling set to 170%

2018-05-09 22:36:37 - Il-2
Frames: 8954 - Time: 200000ms - Avg: 44.770 - Min: 39 - Max: 58

 

Share this post


Link to post
Share on other sites

Flat World Test (Monitor) for my rig. Ill post the rift stuff if it ever comes back from my lads real dad's. 

 

Processor I5 6600K Overclocked to 4.4Ghz on an EVO 212 air cooler. 

MOBO- Asus Gaming Pro Z170

Memory - Corsair Vengence 16MB (2 X8Mb) @ 3000

MSI 1080 Gaming X OC'd +80 on the core  and 560 on the memory. 

 

Tested with V3.0 benchmark

 

FRAPS Results

 

2018-05-28 10:48:14 - Il-2
Frames: 25391 - Time: 200000ms - Avg: 126.955 - Min: 75 - Max: 209

 

Petst Results

 

 

pass mark results.JPG

Edited by BOO

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×
×
  • Create New...