Jump to content
chiliwili69

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

Recommended Posts

0. This is perfmon. It is a windows utility. You can run it as simple as press Win+R and type perfmon, press Enter.

 

AceVenturi, try this

1. Run IL-2 and minimize

2. "Add Counters" -> "Thread" -> "% Processor Time" -> type "IL-2" in textbox -> select all thread instances -> click "Add"

 

Difference is that on your 8 graphs there are usages of each core and on my graphs there are CPU usages for each IL-2 thread. However core usage is also a very interesting information.

 

Below is just my interpretation and only devs can say exactly what is happening. However

CPU load in VR may drop due to a fixed time at which a frame should be ready. If a frame could not be delivered on time then it gets ready whithin next time frame and CPU sleeps for the rest of time. This cannot happen on screen because FPS is not limited and as soon as CPU delivers previous frame it starts processing the next, this gives full CPU utilization.

You can try changing pre-rendered frames in NVIDIA control-panel and see what happens. I think core utilization should be higher and FPS may rise in VR.

Thanks. Have updated and made two new graphs.

 

I tried changing pre-rendered frame settings but just got lots of tearing and lower frame rates. Setting of "1" offered no tearing but also no FPS improvement.

 

The graphs are very similar to the per core graphs but obviously without the extra 'noise' from the other cores. The VR thread does move with the same spikes as the screen version but seems to average about 60% vs 80% so about 20% lower CPU utilisation in VR.

 

I'm no programmer but what you suspect is probably right, low frequency CPUs can't deliver the frames in time leading to overall lower FPS. This explains why the person who got to 5GHz didn't see any improvement over say 4.5GHz as once the frequency is high enough the load probably shifts back to the GPU.

 

Now what would be interesting to see is what CPU speed we need to get max frames possible.

 

I was going to try overclock but someone else with a 1600X at 4GHz wasn't able to get any improvement so I suspect a 'full' frame rate isn't possible on anything lower than say 4.5GHz which puts it out of reach of Ryzen owners. Other VR games run flawlessly but BOS unfortunately does not.

 

Do IL2 Devs look at these threads? It would be good to work with them to see if we can find a way to make BOS work well in VR on lower clocked CPUs.

post-77541-0-26909200-1503029630_thumb.jpg

post-77541-0-16513700-1503029641_thumb.jpg

Edited by AceVenturi

Share this post


Link to post
Share on other sites

A while back I tried researching this single thread issue and found other games suffer from the same problem. A quick googling gives lot's of info and opinions/hearsay on this. For example from the world of Warcraft forum:

"Has nothing to do with software. It has to do with how much memory needs to be accessed in order to do calculations. The more memory needs to be accessed (frequently or in large quantities) the less the CPU can work because RAM is far slower than the CPU.

Pure math based calculations work the CPU to it's limit because it doesn't require large amounts of RAM fetches. A good example of this is Simcraft. Theres two reasons WoW doesn't use a significant portion of the CPU. 1.) Frequent ram fetches and 2.) One thread is heavily loaded down and thus one core has to do the majority of the work. You may look at your CPU usage and notice it's spread reasonably even, but that's because the task is being switched around constantly. X core is doing Z calculation and then X core yields to Y core and Y core resumes Z calculation. It's supposed to keep heat down."


Feel I should ad to this that the "wow" engine and netcode is today not the ancient tech it was when it first came out. 

I suppose above post would explain the low CPU usage, high single thread usage but also the large impact memory speed has on il2 fps.

I'm convinced devs are very aware of this. Splitting things up (like mentioned here previously), would indeed alow even much older cpu's to run il2 comfortably. It would be in the devs best interest to open their product up to a broader customer base. I suppose it's just not very easy. 

  • Upvote 1

Share this post


Link to post
Share on other sites

I got to say I am not sure what they heck has happened, but ever since I installed Fraps to run this benchmark my performance has gone into the cellar. 

First problem I had after installing it, is when launching BoS it would take a few minutes before I could use the mouse.

 

So I uninstalled Fraps and deleted the folder that it had created, and still when I launch BoS it takes a few minutes before mouse is active and I can click the enter button for the game.

Plus, my performance in VR is now worse than it was before I had done this - augh. Not getting near the performance I was before all this.

 

None of the above has ever happened to me in BoS VR since it was implemented, until after installing Fraps for this benchmark.

Edited by dburne

Share this post


Link to post
Share on other sites
Pure math based calculations work the CPU to it's limit because it doesn't require large amounts of RAM fetches. A good example of this is Simcraft. Theres two reasons WoW doesn't use a significant portion of the CPU. 1.) Frequent ram fetches and 2.) One thread is heavily loaded down and thus one core has to do the majority of the work. You may look at your CPU usage and notice it's spread reasonably even, but that's because the task is being switched around constantly. X core is doing Z calculation and then X core yields to Y core and Y core resumes Z calculation. It's supposed to keep heat down

 

Thanks radek from bringing this enlighting text here.

This explain exactly why my cores were not loaded at 100% when running the balapan test in my previous post (ie the heavy thread is jumping from core to core):

post-18865-0-18471200-1503096484_thumb.jpg

It also explain the RAM speed influence. And it opens de door to experiment with the Cache frequency,but still not do it myself.

Edited by chiliwili69

Share this post


Link to post
Share on other sites

 

 

Plus, my performance in VR is now worse than it was before I had done this - augh. Not getting near the performance I was before all this.

 

I had no problem with Fraps, either mouse delay at load or worse VR.

 

For the worse VR it might be related to changing BoX settings or SS in OTT and in SteamVR. You can use OculusMirror to verify actual rendering size.

 

Sorry for asking you to run the test and causing you this unexpected collateral damage. I hope you can find a solution for that.

Share this post


Link to post
Share on other sites

A while back I tried researching this single thread issue and found other games suffer from the same problem. A quick googling gives lot's of info and opinions/hearsay on this. For example from the world of Warcraft forum:

 

"Has nothing to do with software. It has to do with how much memory needs to be accessed in order to do calculations. The more memory needs to be accessed (frequently or in large quantities) the less the CPU can work because RAM is far slower than the CPU.

 

Pure math based calculations work the CPU to it's limit because it doesn't require large amounts of RAM fetches. A good example of this is Simcraft. Theres two reasons WoW doesn't use a significant portion of the CPU. 1.) Frequent ram fetches and 2.) One thread is heavily loaded down and thus one core has to do the majority of the work. You may look at your CPU usage and notice it's spread reasonably even, but that's because the task is being switched around constantly. X core is doing Z calculation and then X core yields to Y core and Y core resumes Z calculation. It's supposed to keep heat down."

 

Feel I should ad to this that the "wow" engine and netcode is today not the ancient tech it was when it first came out.

 

I suppose above post would explain the low CPU usage, high single thread usage but also the large impact memory speed has on il2 fps.

 

I'm convinced devs are very aware of this. Splitting things up (like mentioned here previously), would indeed alow even much older cpu's to run il2 comfortably. It would be in the devs best interest to open their product up to a broader customer base. I suppose it's just not very easy.

Thanks Radek. That might explain the problem.

 

However, my thread cpu usage in the main thread is much lower in vr compared with screen version which means there must be something else going on in vr that limits cpu utilisation, perhaps what Samuel mentioned to do with frame rendering?

 

My rig can run max settings at solid 60fps on my 4k monitor but just struggles in vr.

 

Also, other users with slow ram are able to achieve much better performance with high clocked cpus. My ram is at 2933.

 

They only released vr implementation earlier this year so do you think there's maybe potential to see more optimisation?

 

I don't really want to buy a while new rig but will do so if it gets me better vr performance. Tried playing in 2d yesterday and although it's much clearer and performs flawlessly, the experience really can't stack up to vr, even with the vr fuzzy image and poor frame rates.

 

I'm coming round to the idea that perhaps there's something we or devs can do to shift the load in vr back onto the gpu and not have this cpu bottleneck.

Share this post


Link to post
Share on other sites

I got to say I am not sure what they heck has happened, but ever since I installed Fraps to run this benchmark my performance has gone into the cellar.

First problem I had after installing it, is when launching BoS it would take a few minutes before I could use the mouse.

 

So I uninstalled Fraps and deleted the folder that it had created, and still when I launch BoS it takes a few minutes before mouse is active and I can click the enter button for the game.

Plus, my performance in VR is now worse than it was before I had done this - augh. Not getting near the performance I was before all this.

 

None of the above has ever happened to me in BoS VR since it was implemented, until after installing Fraps for this benchmark.

I think I had the same problem awhile back. Getting into the game took what seemed like a couple of minutes. It was so bad I couldn't alt tab or alt esc out of il2 without pc crashing. I don't know what caused it but reinstalling il2 seemed to fix it.

 

Maybe try that, make sure to backup your settings.

 

I also have variable vr performance when doing an unknown to me sequence of settings changes. A restart should clean up though.

 

The oculus tray tool and devtool caused me all sorts of problems. I use the steamvr settings to control SS and ctrl+numpad 1 to turn off asw.

Share this post


Link to post
Share on other sites

 

 

My rig can run max settings at solid 60fps on my 4k monitor but just struggles in vr

 

When you run at 4K, you are requesting about 8.3 millions pixels to be rendered at 60fps. The render work is done by your GPU (a 1070 can do this quite well).

 

When you run in VR at 1.7 SS, you are requesting about 7.2 million pixels to be rendered at 90fps. Here is when a low CPU clock speed (3.65 GHz) could bottleneck the data transfer from RAM to GPU.

As you know, IL-2 uses OpenVR standards to call Oculus processes. I wonder if a Vive would work better than the Rift in a Ryzen platform.

 

If you decide to change your rig, think you still will get a decent amount of money for your Ryzen system.

Share this post


Link to post
Share on other sites

When you run at 4K, you are requesting about 8.3 millions pixels to be rendered at 60fps. The render work is done by your GPU (a 1070 can do this quite well).

 

When you run in VR at 1.7 SS, you are requesting about 7.2 million pixels to be rendered at 90fps. Here is when a low CPU clock speed (3.65 GHz) could bottleneck the data transfer from RAM to GPU.

As you know, IL-2 uses OpenVR standards to call Oculus processes. I wonder if a Vive would work better than the Rift in a Ryzen platform.

 

If you decide to change your rig, think you still will get a decent amount of money for your Ryzen system.

But even at ss1 my CPU can't do the job :(

 

Increasing super sampling for me doesn't change fps too much. I am running steam vr ss at 2 which improves the picture dramatically but only seems to hurt my fps at the top end, I.E. When I'd be getting 70 i only get 50. At the low end (45) I guess it's already so slow that it doesn't make a difference?

 

The in game AA has a much bigger impact on frame rate for me. I can't run it at 4 without lots of frame drops. For me the ss has less performance impact per gain in appearance than AA or high vs low graphics setting.

 

The low frame rate (45) isn't actually too much of a problem for me, it's when I get dropped frames and particularly the dropped compositor frames that it gets painful. Almost feels like one of my eyes gets pulled sideways or goes out of focus.

 

The double vision when moving head quickly is a problem too which occurs at low frame rates, I guess as it delivers the left eye frame faster than right?

Share this post


Link to post
Share on other sites

I got to say I am not sure what they heck has happened, but ever since I installed Fraps to run this benchmark my performance has gone into the cellar. 

First problem I had after installing it, is when launching BoS it would take a few minutes before I could use the mouse.

 

So I uninstalled Fraps and deleted the folder that it had created, and still when I launch BoS it takes a few minutes before mouse is active and I can click the enter button for the game.

Plus, my performance in VR is now worse than it was before I had done this - augh. Not getting near the performance I was before all this.

 

None of the above has ever happened to me in BoS VR since it was implemented, until after installing Fraps for this benchmark.

 

Ok so I have flown a few hours and missions this morning, and now everything is back to the way it was.

When I launch the game the mouse is immediately responsive like it used to be, and my performance went up significantly - back to where it was. Everything running great and really smooth.

As mentioned, this all happened to me after installing Fraps to do this benchmark. Yesterday I uninstalled Fraps, but everything remained the same. Took a few minutes for mouse to become responsive, and performance was still not what it used to be.

 

So what changed today? The only thing I did this morning first thing before launching BoS, was to go in and delete that trk file we were using for this benchmark. Since doing that first thing this morning, everything is back to the way it was. No issues with the mouse whatsoever and performance is very good. I have been running BoS all morning long, and have launched it several times. I am convinced the issue is gone for sure.

 

Now who knows if this actually had any effect on it or not, I sure don't. But I sure am glad things are now back to normal that is for sure!

  • Upvote 1

Share this post


Link to post
Share on other sites

Glad to see as well that you back to normal.

 

This is a weird case. Maybe nothing to do with Fraps or the trk file.  Only way to know it is to put them again.... :P

Share this post


Link to post
Share on other sites

Glad to see as well that you back to normal.

 

This is a weird case. Maybe nothing to do with Fraps or the trk file.  Only way to know it is to put them again.... :P

 

There is not a chance in hades of that LOL! 

Been flying in Ivan's War campaign most of the day, enjoying great and smooth performance again. 

Share this post


Link to post
Share on other sites

Well, after a couple of days of running tests that were showing no noticeable improvement at all regardless of a variety of overclocking attempts, I was just about ready to give up and just reset everything back to defaults.  But inexplicable things tend to bug me, and while I was resetting everything back to normal default I noticed a new set of OC presets in the BIOS that I hadn't seen before.  I tried the 5.0 preset, and saw the following improvement (using latest 1st post VR benchmark parameters as of 8/19/17; SS set at 1.7 via SteamVR):

 

From standard MB defaults (4.2/4.5 turbo) with 3200 RAM:  Frames: 5065 - Time: 60000ms - Avg: 84.417 - Min: 53 - Max: 91

(Passmark = 6622; CPU mark = 13044)

 

To 5.0 OC preset with 3200 RAM:  Frames: 5231 - Time: 60000ms - Avg: 87.183 - Min: 64 - Max: 91

(Passmark = 7087; CPU Mark = 14131; CPU ST = 2913)

 

Only about a 3% increase in average FPS, but a 20% increase in minimum FPS, and at least it shows something good happening.  Assuming this stays stable with these OC settings I think I'll stick with them.

 

Now - back to trying to learn how to fly the Bf-110 by touch :)

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites

Greetings,

First post on this forum even though I've been following it for awhile.  Got my OR a month ago and been playing around fine tuning that and finally got around to doing  Balapan's test following Chili's instructions.  I also just built a Ryzen 5 1600 rig so I tested that along with my i5 6600K both using the same card (GTX1080) and 16GB of DDR4 3200 RAM.

 

First the i5 6600K OC's at 4.4Ghz (min/max/avg)

 

 

    44/ 91/ 68

 

For this test my CPU Passmark was 10179 and ST was 2498 (my tests, not from the public)

 

My monitor results were:

    115/ 182/ 144

 

I also tested at a lower speed, 4.2 GHz with RAM at 2933 (before I tweaked everything up):

 

               
 

 

 

(VR)


    44/ 91/ 61

 

(Monitor)

      111/ 178/ 140

 

with a Passmark/ST of 9850/2388

 

For the Ryzen 5 1600 video card and RAM were identical, with clock  speed at 3.200 Ghz

(VR)


  42/ 63/ 46

The 45 avg seemed too close to ASW, even though I'm pretty sure I had it turned it off.

 

(Monitor)

  86/ 146/110  

with a Passmark/ST of 12650/1905

 

So no surprises, the CPU makes a difference.   Looking at the spreadsheet, the i5 6600K results seem to be in line with the other Intel processors and the Ryzen is consistent with the one other Ryzen 7 test.

 

The biggest surprise to me was on the Intel, using Resource Monitor, all four processors were even loaded at ~50%.   The Ryzen, which showed 12 virtual processors (six actual), only one was loaded going around 80%-90%.   In both cases the GPU load stayed below 60%-70%.   Only other difference between the two tests is with Ryzen I used Steam VR SS=1.7 and with the Intel I used OTT SS=1.3 with Steam VR=0 per dburne.

 

Sorry for throwing all this data out, but thought you guys might be interested.  If someone wants me to fill out the spreadsheet I can.

  • Upvote 1

Share this post


Link to post
Share on other sites

 

To 5.0 OC preset with 3200 RAM:  Frames: 5231 - Time: 60000ms - Avg: 87.183 - Min: 64 - Max: 91

(Passmark = 7087; CPU Mark = 14131; CPU ST = 2913)

 

 

:good: 

 

Congrats, that is a sweet system you have there!

Share this post


Link to post
Share on other sites

Just a note for those considering ryzen. Despite the low performance of my rig in the benchmark I don't suffer in multiplayer. My fps is normally at 90 unless I'm on the deck and I don't get dropped frames unless HUD is on or there's server lag.

 

So I'll prob stick with this rig as I don't play single player anyway.

 

Might look at coffee lake when it's available or at least wait to see if 7700k prices drop in the lead up.

Share this post


Link to post
Share on other sites

 

 

Avg: 87.183 - Min: 64 - Max: 91

 

Wow! this is a very nice performance. Take into account that the ideal avg is 90fps, so going from 84 to 87 is a big improvement since the closer you are to 90 the more difficult is to increase the avg. The effort to get closer to 90 increases exponentially.

 

If you have a chance to run it at monitor this will set the reference bar for all as well.

 

Enjoy now you new toy!

Share this post


Link to post
Share on other sites

 

 

The Ryzen, which showed 12 virtual processors (six actual), only one was loaded going around 80%-90%.

 

First at all, many thanks for reporting your interesting tests in VR and monitor. It is good to have i5 and R5 CPUs in the spreadsheet.

 

It is surprising that the R5 only load one core. 

Have you tried to OC the Ryzen?

Share this post


Link to post
Share on other sites

Yea, I was surprised too, not sure if that's a BOS thing or a Ryzen issue.  I was more surprised that the i5 had all four cores running, I thought I read BOS was optimized for single core performance. 

 

For the test I did OC to 3.6 GHz, the 3.2 GHz I stated was incorrect.   I did the Ryzen test first before I transferred everything to the Intel computer, so it's hard to transfer back to do more testing since I'm sticking with the i5.  I may do some more tests on RAM speed effects by lowering my speed from 3200 down to 2133.

 

Also to complete the missing entries in the spreadsheet see info below, thanks for putting them in.

i5 MB:  Asus Z-170 AR

R5 MB:  MSI B350 Tomahawk

R5 OC: 3.6

Frames (VR/Monitor):

i5@4.4 (row 8): 4065/8647

i5@4.2 (row 9): 3657/8384

R5@3.6 (row 12):2764/6620

 

I noticed the increased clock speed on the i5 was more important for VR (~400 frames) than the monitor (~260 frames), which says again VR is more CPU dependent.

Share this post


Link to post
Share on other sites

Thank you Wraithzit, added to the sheet. Your avg fps is very much aligned to the one I had when I used 1866MHz memory. As you might have read form this thread, I also had a 4790K@4.7GHz with 1866MHz RAM, then upgraded the RAM to 2933MHz and I increases 8 fps my average. a_radek and Ganeshka were also reporting nice increments in fps when going to higher RAM speeds.

First of all, thanks for the help with dealing with my grey screen issues!

 

I'm looking to max out my settings and get a few more food to help me avoid motion sickness in VR. Based on this thread and your findings, with my current setup I figure the cheapest way to do that is by upgrading my RAM as it seems to be the weaker part of my setup. Looks like my motherboard doesn't support ddr4, so I'm looking at ddr3 RAM. Is size a consideration? I figure speed is the most important, when I look at both Speed ie upgrading to something in the mid to high 2000s MHz, should I worry about getting more than 16GBs of RAM? Would I be better to get 4x4GB sticks? 2x8GB sticks? 4x8GB sticks?

Or should I just be trying to get a couple of sticks of the fastest RAM I can find..

 

I'm thinking about getting these as it seems like others with my motherboard have given them good reviews:

G.SKILL TridentX Series 16GB (2 x 8GB) 240-Pin DDR3 SDRAM DDR3 2400 (PC3 19200) Desktop Memory Model F3-2400C10D-16GTX

 

I want to start saving up for either the next gen of Nvidea GPUs or wait until they are released and the prices drop of the current 1080s/ti so I figure the RAM update is the best way to improve my setup without spending a lot.

 

Would be interested to hear anyone's thoughts!

Thanks in advance!

Share this post


Link to post
Share on other sites

Wow! this is a very nice performance. Take into account that the ideal avg is 90fps, so going from 84 to 87 is a big improvement since the closer you are to 90 the more difficult is to increase the avg. The effort to get closer to 90 increases exponentially.

 

If you have a chance to run it at monitor this will set the reference bar for all as well.

 

Enjoy now you new toy!

 

 

Ran the benchmark on my monitor with the VR test settings from the first post (as of today), while still OC'd to 5.0 with RAM at 3200, and FRAPS reported the following:

 

Frames: 10706 - Time: 60000ms - Avg: 178.433 - Min: 141 - Max: 214

 

Had a chance to introduce one of my sons (visiting for the eclipse) to BOS in VR today.  First I let him fly a couple of quick QMB missions on the monitor just to get used to how it looked on my system (and how to fly with my control setup), and then let him try it with my Rift.  He was gobsmacked - I made the "mistake" of turning on unlimited ammo for him during his monitor test missions, and had a real effort to drag him out from under the VR headset.  Finally he toasted his engine (or maybe took some damage from some flak - couldn't tell for sure what happened trying to watch along on the monitor) and had to attempt a deadstick landing in the snow.  It ended predictably (nosed over and upside down), but he had a blast.  His comment afterwards - "I've never been all that interested in flying this game, but now see how cool it is" - it may cost me another $400 for a Rift bundle for his birthday (I'm already planning to give him my old PC so he'll have something to run it reasonably well), but I may have made another convert for BoS :)  

Share this post


Link to post
Share on other sites

Ran the benchmark on my monitor with the VR test settings from the first post (as of today), while still OC'd to 5.0 with RAM at 3200, and FRAPS reported the following:

 

Frames: 10706 - Time: 60000ms - Avg: 178.433 - Min: 141 - Max: 214

 

Had a chance to introduce one of my sons (visiting for the eclipse) to BOS in VR today.  First I let him fly a couple of quick QMB missions on the monitor just to get used to how it looked on my system (and how to fly with my control setup), and then let him try it with my Rift.  He was gobsmacked - I made the "mistake" of turning on unlimited ammo for him during his monitor test missions, and had a real effort to drag him out from under the VR headset.  Finally he toasted his engine (or maybe took some damage from some flak - couldn't tell for sure what happened trying to watch along on the monitor) and had to attempt a deadstick landing in the snow.  It ended predictably (nosed over and upside down), but he had a blast.  His comment afterwards - "I've never been all that interested in flying this game, but now see how cool it is" - it may cost me another $400 for a Rift bundle for his birthday (I'm already planning to give him my old PC so he'll have something to run it reasonably well), but I may have made another convert for BoS :)

 

:good: 

 

Nice!! Lucky son!

Share this post


Link to post
Share on other sites

Went ahead and tested the effect of RAM speed, lowered it from 3200 to 2133 with my i5@4.2 Ghz:

 

For VR, I lost about 6 FPS, for the monitor I lost 9 FPS.  So about a 8%-10% increase going from 2133 to 3200.

To update the spreadsheet, here's the actual data:

 

Copy row 9 (4.2 GHz) and change the RAM to 2133.

VR: 3284/44/90/54.7

Monitor:  7828/103/167/130.5

 

Worth changing RAM, maybe not.  But for a new build definitely go for the faster RAM (mine was 16GB G.Skill Ripjaws (VKB))

Share this post


Link to post
Share on other sites

 

 

should I worry about getting more than 16GBs of RAM? Would I be better to get 4x4GB sticks? 2x8GB sticks? 4x8GB sticks?

 

With the same 4790K CPU, I was having the same kind of questions you have now.

 

I think that BOS doesn´t gain anything going beyond 16Gb, the proof is that only 5Gb are used when you play. So even 8Gb could be OK as well.

 

For the speed point of view it doesn´t matter 4x4GB than 2x8Gb (putting them in the right slots indicated by your MoBo).

 

The benefit of 4 over 2 is that, for same memory size, you have more surface for heat dissipation than with 2. Also, if you plan to put 4x4Gb, you can first buy 2x4Gb and verify if this is enough (most likely it will be), and them buy another two stick if needed.

 

Your MoBo supports until 3300MHz (https://www.gigabyte.com/Motherboard/GA-Z97X-SOC-rev-10#sp), so I will go to the fastest one you can get. But the price increases a lot when you go to the top one. This is up to your budget.

 

In DDR3 there is less options in the market. I went to 2933MHz and I don´t regret it. I think it is worth. In your case you would gain even more than me since you came from 1333MHz.

 

I would do first the RAM upgrade (please run the test before/after) and then, if needed, the GPU.

 

Remember that a 4790K with good CPU cooling (your is liquid) and good RAM could have a longer life than you should expect. (Latest 7740X don´t go too much beyond).

So, you will have time to upgrade your GPU one or two times along the next year.

Share this post


Link to post
Share on other sites

 

 

His comment afterwards - "I've never been all that interested in flying this game, but now see how cool it is"

 

Yes! new generations will be more interested in flight sim than before because the existence of VR.

I did the same with my 10 year old soon with the Spitfire. Now I have to dose those sessions to one or two 30min session per month. He is still too young for VR.

These new generations are very lucky, and we also as well. :dance:

Share this post


Link to post
Share on other sites

With the very few test we have conducted (many thanks to contributors) I have started to make some correlations. You will see that I added a X-Y plot in the same spreadsheet:

post-18865-0-69110200-1503311540_thumb.png

 

https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k/edit#gid=1870565807

 

One of them is to see the correlation between Passmark singlethread with the AVG fps in VR (blue dots) and in monitor (red dots). It is still early since the number of samples is very low, but they seems to start to correlate well both.

 

Also if you see slope of the lineal correlations, you will see that the monitor slope (0.071) is the double of VR slope ( 0.036). So our guess ballpark number of halving the fps of the monitor was not far from reality.

Edited by chiliwili69
  • Upvote 1

Share this post


Link to post
Share on other sites

Yes, that's a pretty strong correlation between ST CPU and FPS.  Also looks somewhat independent of the GPU, which means get a 1070 and invest the rest in CPU?  Of course with the miners  driving up the price of the 1070 you might as well get a 1080.

Share this post


Link to post
Share on other sites

Hey guys did, anyone try to oc vram, I think it gave me a significant increase in performance. I haven't measured anything yet but, earlier I played on multiplayer and I was having an average of 70 fps, later overclocked my vram by 400mhz on my get 1070 (using 7700k at 4,6Ghz, 2666 ram) and hopped back in the same match the amount of players had increased and I was getting 85 avg fps (even on takeoff). My current settings ultra ,aa x4, full dynamic resolion factor ,distant landscape x4 ,mirrors off ,lanscape filter sharp ,screen res 720x576 rest of the settings on. (no supersampling)

Share this post


Link to post
Share on other sites

Ran this test on my rift using i7 3930K no oc, gtx 1080ti, 16gb 3333mhz and the settings outlined on the first page.

 

 Avg: 48.368 - Min: 45 - Max: 76

 

My fps is well below that of others with a 1080ti, I think I've got something setup wrong.   Anytime an airplane appeared on the screen the fps seemed to stay locked in at 45fps.  ctrl num1 did not appear to do anything but I do have asw turned off in steamvr and debug conole of vive.

Edited by AcesHigh

Share this post


Link to post
Share on other sites

Finally managed to get this test going today.

 

I thought it would be interesting for you to have a GTX 970 result up in the board, as its recommended as the required GPU for a good experience in VR. This will hopefully help others out there decide if they have the rig to run VR well enough in IL2 before they make the investment.

 

Current setup:

Gigabyte Z97X SOC Force Motherboard

CPU: i7 4790k set to run at default of 4.0ghz

GPU: Gigabyte GEForce Windforce 970 gtx gaming G4

RAM: GSkill Ares 2x8GB DDR3 1300mhz

Oculus Rift with touch (all other is in signature)

Windows 10 64 Pro

 

 

 

My settings are default in BIOS (after the reccomended settings I had found online somewhere had a lower performance in VR and downright abysmal in monitor in this test with hyperthreading disabled etc. *see earlier post with link to be settings in bios and Nvidea control panel) conclusion, don't disable hyper threading for this game (unless someone has a post that disputes this, I would be happy to fire it up again) NVidea control panel settings have only "set to maximum performance" wherever it was an option in global settings.

 

No overclocks of CPU, GPU or RAM

 

Passmark Rating

CPU mark: 11462

CPU Single Thread: 2651

 

VR Test Results Balapan Track: (this explains my nausea in VR and screen freezes)

Frames: 2647 - Avg 44.117 - Min: 26 - Max: 46

 

Monitor Test Results:

Frames: 8215 - Avg: 136.917 - Min: 72 - Max: 182

 

I have new DDR3 2400 RAM on the way and will try the test after installing that to compare and see if RAM speed makes a noticeable difference. Then I will try overclocking CPU, GPU, new RAM and post those results as well to see what I can get from what I have.

Edited by SimpleThe1st

Share this post


Link to post
Share on other sites

 

 

Also looks somewhat independent of the GPU, which means get a 1070 and invest the rest in CPU?

 

Somewhat yes. Going beyond a 1070 is justified if you still want to increase the SS beyond 2.0 (in SteamVR). 

Share this post


Link to post
Share on other sites

 

 

My fps is well below that of others with a 1080ti, I think I've got something setup wrong.

 

Thanks for running the test.

 

I think that your CPU is clearly the bottleneck, since your RAM and GPU are in the Top of the Top range.

 

Try to run the Passmark benchmark (https://www.passmark.com/download/pt_download.htm ) and see what CPU mark and Singlethread numbers do you obtain.

 

The 3930K is a very good CPU but with low speed clocks. You will need to do overcloking if you want to increase your performance.

 

There is some guidance here:

https://forum.il2sturmovik.com/topic/29881-overclocking-4790k-better-bos-performance

  • Upvote 1

Share this post


Link to post
Share on other sites
Passmark Rating (wasn't sure what to write here, so I just copied down all the data on the screen)

 

Many thanks for running the benchmark with the 970, it will be important to many who are considering upgrading their GPU.

 

You only need to run the CPU test, and the two values to report are the CPU Mark ( 11472.6 in your case) and the CPU Single-Threaded. In the new v9 Passmark Performance Test they are reported where I indicated in red:

post-18865-0-64935000-1503419372_thumb.png

Please, run just the passmark CPU test to know your CPU Single Threaded score

Edited by chiliwili69

Share this post


Link to post
Share on other sites

I made more measurements with 3770k 1080Ti 

 

First I overclocked CPU a bit

Before: 4.9GHz RAM1600 (this info is already in the spreadsheet)

Passmark 2696

8969 116 212 149 (frames, min, max, avg)

 

After: 5.1GHz RAM1600 

Passmark 2830

9550 120 226 159

 

Then I tried to overclock RAM

5.1GHz RAM2400

Passmark 2830

10224 133 238 170

 

And finally I played with graphic settings and realized that switching the fullscreen on gives some more FPS. However in bench instructions fullscreen should be off.

 

5.1GHz RAM2400 fullscreen on

10842 134 239 180

 

Overclocking to 5.2GHz did not improve FPS which is strange, so I stopped at 5.1 for now. Besides air cooler limits OC capabilities especially when testing at full load of all cores

 

Edit: sorry, mistaken 2400 results with 2000. Now numbers are fixed

Edited by TUS_Samuel

Share this post


Link to post
Share on other sites

Results of the test in VR with the Vive

 

Frames, Time (ms), Min, Max, Avg
  5188,     60000,  57,  90, 86.467

 

Running i7 7700K @ 4.8, 3000Mhz Ram, Settings all as per the first post.

 

 

 

 

Results of the same test with only SS @ 2.3 in steamVR as the change

 

Frames, Time (ms), Min, Max, Avg
  5094,     60000,  54,  90, 84.900

Edited by =TBAS=Tripwire

Share this post


Link to post
Share on other sites
VR Test Results Balapan Track: (this explains my nausea in VR and screen freezes) Frames: 2647 - Avg 44.117 - Min: 26 - Max: 46 Monitor Test Results: Frames: 8215 - Avg: 136.917 - Min: 72 - Max: 182

 

Something strange is happening here in your VR test. Your monitor results are relatively good, looking at those you should expect something more in VR, at least an avg of 60-65. As far as we have seen, VR Avg is about half of Monitor Avg, but we should take into account that all VR&monitor tests has been done with 10XX Pascal cards.

 

I don´t know if the RAM upgrade, the CPU overclock, or the GPU upgrade will correct this anomaly.

Edited by chiliwili69

Share this post


Link to post
Share on other sites

 

 

Overclocking to 5.2GHz did not improve FPS which is strange, so I stopped at 5.1 for now. Besides air cooler limits OC capabilities especially when testing at full load of all cores

 

Thanks for reporting this test in monitor, it is impressive that you OC your CPU from 3.5 baseclock to 5.2 with just air cooler. 

Based in your performance in monitor you have demonstrated that a 3770K would be in theory more than valid for VR. But a good OCing is needed.

 

Any plans for VR?

Share this post


Link to post
Share on other sites

Something strange is happening here in your VR test. Your monitor results are relatively good, looking at those you should expect something more in VR, at least an avg of 60-65. As far as we have seen, VR Avg is about half of Monitor Avg, but we should take into account that all VR&monitor tests has been done with 10XX Pascal cards.

 

I don´t know if the RAM upgrade, the CPU overclock, or the GPU upgrade will correct this anomaly.

Possible that reprojection forced on in that test? Seems odd to not go over 45 by much. Perhaps a check of the steam VR Settings? Edited by =TBAS=Tripwire

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...