Jump to content
chiliwili69

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

Recommended Posts

Thanks for the suggestion, but since most of the test (if not all) are CPU limited we preferred to keep the OC data of the CPU since OC the GPU will not affect the results in most of the cases. Even-though all data is valuable so, if people report GPU OC data , it will be collected in the Notes.   

This is the main issue at current with BOX, single threaded CPU limited. There is so much potential in rendering and ai objects that could greatly benefit from coding for multi threaded / core CPU's.

 

For VR and BOX in general especially for Multiplayer Servers, moving to take advantage of multiple CPU cores is a must.  Even with my aging AMD 8350 with 390X the difference in Balanced and Ultra at 1440p is not huge and even 1080p vs 1440p is little. The problem being single threated CPU limitations.

 

I sincerely hope that some stage in the next year, definitely by the time of Midway's release, they code for multi core CPU's. I do realise resource consrraints but, this is just things are.

Share this post


Link to post
Share on other sites

This also happened to me two or three times in the past. If the reported Max fps is higher than 91, the test is not valid. Having certain periods of fps above 91 is the reason for having higher avg.

I was mentioning this in this post:

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/page-3?do=findComment&comment=496714

 

Yeah - now another problem to troubleshoot.  For some reason, although OR is enabled in my startup.cfg and the display comes up split screen like usual on my monitor, no image is coming on in my Rift headset.  I do see an image in the headset when I just start Oculus Home app, but not with BoS.

 

Any ideas on what causes that?

Share this post


Link to post
Share on other sites

Yeah - now another problem to troubleshoot.  For some reason, although OR is enabled in my startup.cfg and the display comes up split screen like usual on my monitor, no image is coming on in my Rift headset.  I do see an image in the headset when I just start Oculus Home app, but not with BoS.

 

Any ideas on what causes that?

 

 

Is Steam and SteamVR running? If not, then you won't get any display in your Oculus when running Il-2 since it's a SteamVR title. 

 

Madmatt

Share this post


Link to post
Share on other sites

Is Steam and SteamVR running? If not, then you won't get any display in your Oculus when running Il-2 since it's a SteamVR title. 

 

Madmatt

 

Yeah - steamVR was running just like always.

 

I played a little Robo Recall just to make sure the headset really was working OK (it was), and then for a lark I decided to try starting BoS from my Oculus Home menu screen.  That worked, so I closed everything down and went back and tried starting BoS the usual way, and everything worked fine this time.  Not sure what that little hiccup was all about, but at least it was short-lived.

 

Ran a few more tests with the benchmark track, and this time the results with the new drivers came out essentially the same as with the old drivers.  So, no noticeable improvement with the new drivers (at least FPS-wise, with my maxed out settings), but at least the test and FRAPS seem to be working again.

 

On one of the tests I was letting Precision XOC run to see if I could gather info about my temps and stuff during the test.  According to that, it said that my max CPU temp peaked at about 82 right near the start of the track and had several other peaks into the mid-high 70s for about 10 seconds after that, and then stayed under 70 the rest of the way.  It looks like Precision XOC will keep records for about 5 minutes, so I may try running the test track several times in a row and see what it looks like.

 

[EDIT]  Ran the test track 3 times in a row, and saw the same pattern as I noted above repeat for each run.  Not sure what it means, but hopefully the big stress is from initially loading up the track/mission and not from flying around in it so maybe that's not too bad.  I may try a busy QMB for 5 minutes or so and see what happens.

 

[EDIT]  Ran about 10 minutes of a QMB (8v8 with a bunch of He-111s and Ju-52s as targets) - CPU temps pretty consistently in 60s and 70s the whole time (or at least the last 5  minutes that were tracked by Precision XOC) with one or two peaks around 78 or 79.  Not sure what that means, but I'm thinking maybe I should try something like that again with no OC and see how different it is.

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites
It looks like Precision XOC will keep records for about 5 minutes, so I may try running the test track several times in a row and see what it looks like

 

I have recently installed the free MSI AfterBurner tool which is something used to OC the GPU (something I still have not tried at the moment) but the nice thing is that it has very nice trends for CPU temperatures.

You can cofigure the variables you want the display (most of them GPU but also CPU), the sampling period, the number of columns graphs, etc.

 

If you use the deafult  sampling period and re-size the window to fill a full 4K monitor I can see about 40 minutes of history (more can be log in a file), which it is useful for longer tests. Here some samples:

 

post-18865-0-62292100-1506243204_thumb.png

post-18865-0-43361800-1506243226_thumb.png

Edited by chiliwili69

Share this post


Link to post
Share on other sites

I was experimenting a bit more with this nice AfterBurner trends and the Prime95 stress tool (Torture Test : Bench).

Firstly I put my two system fans at 100% and my single CPU fan at 50% with no OC (just 4.0 Ghz base clock).

 

Then I run first the 1 minute IL-2 VR test and then the Prime95 with 1, 2 and 4 Threads.  What I can conclude is that even the Prime95 with just one thread, it is more CPU demanding than IL-2 VR in terms of CPU heat generation and needed cooling.

Threfore if your machine supports well the stress with Prime95 with 1 thread, you are more than fine with IL-2 in VR. As you all know, stress tools go to stress as much as possible, but this kind of load is not normally demanded in IL-2.

So I can use Prime95 (1 thread) for all my future OC and cooling experiments.

 

This image show the 4 tests and then the temp with no load:

post-18865-0-08248200-1506249562_thumb.png

 

This another image shows all core temps and CPU temp which were all following same pattern:

post-18865-0-11403300-1506249580_thumb.png

Edited by chiliwili69

Share this post


Link to post
Share on other sites

What I can conclude is that even the Prime95 with just one thread, it is less CPU demanding than IL-2 VR in terms of CPU heat generation and needed cooling.

Threfore if your machine supports well the stress with Prime95 with 1 thread, you are more than fine with IL-2 in VR.

 

I assume you meant to say "more", given the results in the picture.  I've ordered some new fans (some of mine are making more noise than they should) and will run some more tests once I get them installed.

 

[EDIT]  I installed MSI Afterburner and ran a few test with my old fans and then installed my new Noctua fans (replaced my fans that came with my corsair H75 with a couple of NF-F12 PWM) and ran some more tests.

 

I'm really liking the noise reduction from the new fans a lot (no more jet whine that I was getting as the Corsairs ramped up), but I still haven't had any luck controlling the temp spike into the mid 80's at the beginning of the benchmark test track.  Overall, the CPU seems to be running close to 10 degrees cooler during the rest of the track (spikes into the mid 60's instead of the mid-70's), but I'm still getting that initial spike to about 85 right at the start.  Not sure if that's really a problem, but I was hoping to stay under 80 period.  Maybe I'll drop back to stock CPU clock and see how that affects the temps.

 

[EDIT]  Well, dropping back to stock CPU clocking dropped that spike down about 20 degrees, to 63, and about 10 degrees on average for rest of track (mid-50's instead of mid-60's).  Guess I need to read up on the effects of those higher temps and whether the increased performance is really worth it.

 

[EDIT]  I ran the RoG Realbench stress test for 30 minutes at stock CPU clock, and temps were consistently 75-77 with max = 79.  When I set the CPU OC to 49 and started to run the Realbench test again, the temps started getting up to 95 so I shut it down.  Leaning toward just going back to stock at this point.

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites

 

 

I assume you meant to say "more", given the results in the picture.

 

Yes, just corrected that in the post. I meant more. Thanks.

Share this post


Link to post
Share on other sites

 

 

but I'm still getting that initial spike to about 85 right at the start.

 

You can set the trend to sample every 100ms and monitor the Core temps, CPU temps and also CPU fan revolution/or water pump revolution (don´t know it it can be trended).

But the spike could be due the lag of the cooling system in reacting to a sudden load increase. Try to put a more aggressive CPU fan/pump profile or try to put it at manual to 100% before launching the test.

 

In any case, I think that if it is only a spike to 85 C it doesn´t matter. The important thing is the maintained temp along the game play.

 

Also, you have realise that some stress test, stress the CPU too much (after all this is the purpose), but sometimes much more than really needed to design your OC for normal gaming.

That´s why I said above that the simple Prime95 with 1 single thread is a good compromise.

Share this post


Link to post
Share on other sites

Well, I found another preset/tuning option in my BIOS, and selected the TPU2 water-cooled profile - it did its thing and wound up overclocking to 4.8.  I've run some tests with those settings - the RoG Realbench temps were hovering in the low 80s (peaked to 85) before I shut it down, and they peaked (in the initial spike) at 75 while running the VR benchmark track.  That sounds pretty promising, so I guess I'll stick with it for some more testing for a while.

Share this post


Link to post
Share on other sites

 

 

 

Thanks for this info. Those guys do a very nice test. I am amazed to see how much effort put Intel in the micro itself and how poor is their assembling process or thermal paste to produce those huge differences. Why Intel doesn´t do that at factory?

 

if this delidding thing is really true, it is more worth to invest in a proper deliding rather than a premium cooling system with many fans.

Share this post


Link to post
Share on other sites

Thanks for this info. Those guys do a very nice test. I am amazed to see how much effort put Intel in the micro itself and how poor is their assembling process or thermal paste to produce those huge differences. Why Intel doesn´t do that at factory?

 

if this delidding thing is really true, it is more worth to invest in a proper deliding rather than a premium cooling system with many fans.

 

I was intrigued by this too, but I have been unable to find the delidding tool they mentioned or even how much it costs.  Amazon reports them as being unavailable (and they don't know when they might be back in stock, and they don't say what the price was), and the maker's website has an explanation of how to use them but does not seem to have any for sale.  I suspect Alex Jones would say that Intel is somehow behind this disappearance.  Maybe eBay has some - off to check there.

 

[EDIT]  Maybe Alex Jones is right - none on eBay either.  You'd think there would be lots of people selling theirs after they've used them.  Hmmm.

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites

If you are going to delid your CPU yourself, I would recommend a tool like this: https://rockitcool.myshopify.com/

 

Do not attempt to delid without the proper tools designed specifically to delib CPU's. Also keep in mind that the delidding is just half the battle. You still need to clean up the old TIM and reapply with something that will provide a better thermal interface. Lots to choose from, please do the research and make an informed decision. 

 

If you want to pay to have someone do it for you, I would recommend using the services of http://www.siliconlottery.com They have been around for a while and are well respected in the overclocking community.

 

Delidding definetly CAN result in lower temps but just like that name in the website above says, its truly a lottery. You could pay to delid your chip and not notice much of a difference at all (if your chip is already a good performer stock) or a big difference (if it is not). 

 

It all comes down to how well the stock TIM was applied, how good the core silicon is in your chip and many many other factors.

 

As for the comment above about a delid replacing a good aftermarket cooling product. No! Never. Not going to happen. Delidding just gives you the potential for a little more thermal overhead but it is not a replacement for a good cooling solution.

 

Madmatt

Edited by Madmatt

Share this post


Link to post
Share on other sites

Of course :)

 

I just finished ordering the tool, paste and silicone rtv I'll need to delid my 7700X, and then ran a few more tests at my latest OC=4.8 settings, and now I'm getting FPS results more in line with my previous 4.9 tests but without the high temps.

 

It'll take a few days for all the packages to arrive, so I can run a few more tests just to be sure, but right now things are looking pretty good even before the delidding.  Still, it would be nice to gain maybe 10-20 degrees of cooling, if only to lessen the stress on the components a bit in the long run.

Share this post


Link to post
Share on other sites

 

If you want to pay to have someone do it for you, I would recommend using the services of http://www.siliconlottery.com They have been around for a while and are well respected in the overclocking community.

 

 

 

Hadn't heard of them before but interesting link. Seems they sell pre-delided CPU's with guaranteed overclocks at specified settings and their own 1 year warranty too. The prices are only marginally higher than normal retail chips too. I'm giving very serious consideration to the 8700k, might grab one from them, depending on price and what they bin them as once released.

Share this post


Link to post
Share on other sites

Indeed - checked out that link and it's an interesting concept.  If I decide to try it myself and wind up toasting my current 7700X, I might consider buying a new, already-delidded one from those guys instead of just getting another stock 7700X from my usual source.

Share this post


Link to post
Share on other sites

Last week I have upgraded to windows 10 and got 13% FPS drop. Defenders and such were disabled instantly after upgrade. I did not find the cause of FPS drop and returned back to windows 7 and FPS came back.

So strange

Share this post


Link to post
Share on other sites

Last week I have upgraded to windows 10 and got 13% FPS drop. Defenders and such were disabled instantly after upgrade. I did not find the cause of FPS drop and returned back to windows 7 and FPS came back.

So strange

 

Windows Game Mode may have been enabled.

 

I think it is ok for some windows games, not so much for what we do.

Share this post


Link to post
Share on other sites

Last week I have upgraded to windows 10 and got 13% FPS drop. Defenders and such were disabled instantly after upgrade. I did not find the cause of FPS drop and returned back to windows 7 and FPS came back.

So strange

 

After you upgraded to Windows 10, did you also check to make sure you were running the latest video drivers and put all their settings to what you had before? Just a thought because the drivers that MS installs by default for most video cards during a OS install are not optimized at all.

 

Madmatt

Share this post


Link to post
Share on other sites

Did not know about the Game Mode, may be that is the cause. But 13% is so much.

Also I did not notice any high loaded processes in task manager.

Drivers were the latest. Hardware is the same. CPU OC is the same. Passmark ST is the same. Only OS version and may be DirectX version is different.

 

Unfortunately I can only guess until the next time I decide to upgrade OS. But for those who have FPS less then expected (see chilliwilli's formula), you can try installing win7 as the second system and see if it helps.

Share this post


Link to post
Share on other sites

Hey guys,

i upgraded RAM today. 16gig DDR3 2666mhz. Increased from 1866mhz.

 

Updated Balapan test result on vive: 

 

Frames: 5054 - Time: 60000ms - Avg: 84.233 - Min: 46 - Max: 90

 

so a nice increase from 75avg that is in the current .xlxs sheet.

 

I'm still finding that i'm running the LOW settings option for my standard setup as it is providing a 89fps butter smooth during all game play. As soon as i select Ballanced or higher i get some stuttering which is just not acceptable after having it silky smooth on LOW.

Share this post


Link to post
Share on other sites

Made some measurements for 2.012d

 

i7 3770k @ 5.3GHz (1 or 2 cores 5.3, else 5.2)

DDR3 2400 10-12-13-30

MB: Gigabyte Z77-D3H

Windows 7 x64

 

Passmark 2875

FullHD monitor

Reshade ON   11210, 135, 276, 186 (average of 3 runs rounded to integer)

Reshade OFF 11448, 140, 288, 190 (average of 3 runs rounded to integer) soooo black and white =(

Edited by TUS_Samuel

Share this post


Link to post
Share on other sites

 

 

As soon as i select Ballanced or higher i get some stuttering which is just not acceptable after having it silky smooth on LOW.

 

Glad to see that your memory upgrade gave you a nice fps boost. Better than expected!

 

I don´t understand why you have stuttering at balanced. Based in your performance in the test, you should be at 90 almost all time with Balanced. (assuming same settings than test)

What settings do you use for normal flying at BOS? 

Share this post


Link to post
Share on other sites

 

 

i7 3770k @ 5.3GHz

 

Thanks for sharing the performance hit of Reshade.

 

BTW, you OC quite well to 5.3, being 3.5 your base clock. Any liquid nitrogen in place?  :good:  

Share this post


Link to post
Share on other sites

 

 

CptJackSparrow' timestamp='1508103498' post='521659'] Those frames are from vr that Samuel posted?

 

No, they are only in Monitor. Samuel is still not in the VR club. 

Share this post


Link to post
Share on other sites

Thanks for sharing the performance hit of Reshade.

 

BTW, you OC quite well to 5.3, being 3.5 your base clock. Any liquid nitrogen in place?  :good:  

 

Delidding + massive air cooler

 

 

Those frames are from vr that Samuel posted?

 

No, monitor only because I still have no VR. You can divide FPS by 2

Share this post


Link to post
Share on other sites

Delidding + massive air cooler

 

 

 

No, monitor only because I still have no VR. You can divide FPS by 2

Cool setup.

 

Cannot divide fps by 2 tho. VR in il2 calls damned AVX instructions. Your stuff will melt at that CPU clock before it can power off for safety }:-(

 

I want to delid as well. Did you use 8auer's tool?

Edited by 2./JG51_Fenris_Wolf

Share this post


Link to post
Share on other sites

CPU thermal throttling is very fast, so I don't worry about that.

 

My tool:

 

 

1005490471.jpg

 

 

 

But if I had 7700k, I don't know if I'd took the plunge

Share this post


Link to post
Share on other sites

Looking for CPUs with good Single Thread Passmark (STMark) performance should be the first thing to look at when upgrading or buying a new PC. (Based in the highly correlated values of fps in VR and monitor versus STMark).

Passmark continuously report the STMark performance of the new CPU, for example the new 8700K. As you will know, their top STMark CPUs are displayed here:

https://www.cpubenchmark.net/singleThread.html

and also here:

https://www.cpubenchmark.net/CPU_mega_page.html

(in this last link you can sort by STMark)

 

The problem with the STMark reported by Passmark is that it is an average of all the tests performed by the users (including the base clock and turbo, but not OC beyond turbo).

That´s why the 8700K shows a STMark of 2751, because its turbo speed is 4.7 Ghz.

 

In a previous post in this thread (https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/?p=514240) I showed how the STMark varies with the CPU clock speed. It is a pure linear relationship. I wanted to know if other CPUs (like 7700K or other) follow the same relationship, so I created a new tab in the online sheet called "STMark vs. OC"

There you will be able to compare the relationships. All CPUs show the same linear trend, but there are small differences for the same CPU speed. For example, the 4790K, for my surprise, is slightly over the 7700K. The STMark is not influenced by RAM speed, but the BOS VR it is.

In any case, the conclusion is that the important thing is to have a CPU with high OC capabilities, the more CPU OC ---> the more STMArk ---> the more fps in BOS.

For anyone wanting to run a series of STMark test at different CPU clock and report them here, I will be pleased to include it in the online sheet. Specially the 8700K at high clock speeds!!!

https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k

post-18865-0-27845200-1508578055_thumb.png

Edited by chiliwili69

Share this post


Link to post
Share on other sites

Inspired by this thread I am planning to upgrade my RAM from 8 GB 1600 MHz to 16 GB 2666 MHz (DDR3).

 

Does it make any difference if I get 2x8 GB or 4x4 GB?

 

I am planning to buy this one 2nd hand: http://www.geil.com.tw/products/showSpec/id/409 (4x4 GB, more precisely 2x4 + 2x4)

 

I have a 4790K and an ASUS z-97-A.

 

Does anyone have any info on this RAM or any other info that would prevent me pull the trigger?

 

Thanks!

Share this post


Link to post
Share on other sites

 

 

Black_Sab' timestamp='1508722570' post='524295'] Does it make any difference if I get 2x8 GB or 4x4 GB?

 

I was having exactly the same doubt when upgrading my RAM from 2x8 @1866 to 4x4Gb @2933. I read about it and for the performance it doesn´t matter if you have 2x8Gb or 4x4Gb.

In fact, with 4x4 Gb you have more surface area for heat dissipation.

 

In your BIOS, you will need to enable the XMP profile for your 2666 memory.

 

You can run the VR test (in VR or/and monitor) and report fps before/after upgrade.

Share this post


Link to post
Share on other sites

I was having exactly the same doubt when upgrading my RAM from 2x8 @1866 to 4x4Gb @2933. I read about it and for the performance it doesn´t matter if you have 2x8Gb or 4x4Gb.

In fact, with 4x4 Gb you have more surface area for heat dissipation.

 

In your BIOS, you will need to enable the XMP profile for your 2666 memory.

 

You can run the VR test (in VR or/and monitor) and report fps before/after upgrade.

 

Thanks chiliwili

Share this post


Link to post
Share on other sites

I thought the consensus was that the game didn't need more than 8GB.of ram...or have I missed something.I haven't seen anything over 4GB of usage in MSI afterburner. I accept that faster ram speeds make a difference however.

Share this post


Link to post
Share on other sites

Just one thing to keep in mind if you are OC your RAM. The more capacity you have, the more heat it will produce. 8 GB is what you should have these days, for less you need a very good reason. More, depends on the programs you‘re running. DCS wants 16 GB. BoX is more conservative.

 

1151 Socket CPUs are dual channel. You need two for full speed. Having two pairs gives in theory the same speed, but if you are pushing it, depending on how similarly binned those two pairs are, you might not be able to go as far as when your system just has to harmonize one pair.

 

2066 Socket CPU are sometimes 4 channel. You should need two pairs then.

Share this post


Link to post
Share on other sites

 

 

I thought the consensus was that the game didn't need more than 8GB.of ram...or have I missed something.I haven't seen anything over 4GB of usage in MSI afterburner. I accept that faster ram speeds make a difference however

 

Yes, 8Gb is enough for BOS.

When playing BOS in VR the RAM consumption is at 5Gb.

Wanting 16Gb I assume that there will be other purposes/games.

  • Upvote 1

Share this post


Link to post
Share on other sites

Glad to see that your memory upgrade gave you a nice fps boost. Better than expected!

 

I don´t understand why you have stuttering at balanced. Based in your performance in the test, you should be at 90 almost all time with Balanced. (assuming same settings than test)

What settings do you use for normal flying at BOS? 

 

800x600

sharpen

shadows - low

mirrors off

landscape x2

grass - normal

AA x2

 

SteamVR SS 2.0 (was 1.8 old ram)

Async repro - on

SteamVR beta - off

 

SteamVR SS being lowered doesnt seem to help the stutter difference between LOW and Balanced.

 

AceVenturi with a good I7 7700k rig also notices a big difference between LOW and Balanced.

Share this post


Link to post
Share on other sites

 

 

800x600 sharpen shadows - low mirrors off landscape x2 grass - normal AA x2

 

This are very similar to the VR test, and also very similar to the ones I use (except I use High preset and landscape normal).

 

I believe that if you run VR test at BALANCED you will get an avg very close to 90. Therefore, you should not have stuttering at all.

 

Is that in all scenes or in a particular one? is it something we can reproduce? (what mission, map, plane, time, season, weather conditions)

Share this post


Link to post
Share on other sites

Wraith, two things come to mind: What USB Hub PCI-e Card are you using, and have you disabled power savings in device manager for each USB Controller used? Are you using WiFi to connect your computer to your ISP/Home Router?

Edited by 2./JG51_Fenris_Wolf

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...