GPU Upgrade advice for VR please.

[Stratman59]

Hi all, I recently took advantage of the Oculus Rift sale and purchased for £399, a realistic price IMO.

I absolutely love the immersion in IL-2 that VR gives so I really can't see me going back to a monitor. I bought the rift mainly for flight and racing sims although some of the VR titles are excellent too.  

 

My current setup is as follows:-

 

Windows 10 64 bit

MB - Gigabyte Z68AP-D3

CPU- I7 3770k ivy bridge

GPU- GTX 970 4 gig

8 gig of DDR 3

 

Now i know that this system is starting to be a little dated so I'm thinking where my best upgrade would be?

I'm feeling that a GPU upgrade to a 1080 would give me the best increase with the chance of supersampling.

 

Last night I tried to use a 1.5 supersample using my GTX 970 knowing it would be choppy and horrible but I was expecting to at least see a noticeable difference in the loading screens, to be honest I couldn't tell any difference and I'm not sure it was working at all?

 

As I'm flying with 45 fps on medium setting at the moment but so far, only with one other aircraft in the sky, will I see a noticeable difference using supersampling? I recall that on my old DK2 it made a big difference on Elite Dangerous even with the GTX 970.

 

A 1080 is a lot of money but if I'm going to buy I'd rather spend the extra and have a degree of future proofing over a 1070.

 

In a nutshell, will using supersampling improve the aircraft spotting and ground target identification? 

 

Many thanks 

 

 

[dburne]

A 1080 would absolutely be better, also a 1080 Ti would be the best thing for VR if you can swing that.

 

Also regarding your CPU, are you overclocking it or running it at default?

 

As far as Target ID, that is a drawback in this current generation of VR  due to the lower res, bout all you can do there is work on your aircraft identification skills best you can. Hopefully the next generation will be significantly better in that department.

[II/JG11_ATLAN_VR]

Also thinking about update choices:

1.i7 7700k New MB 16gb RAM DDR 4 or

2.amd ryzen 7 1700 new MB and ram DDR 4 16gb

I will keep my existing gtx 1060 6gb and Power supply be quiet 750w

Flying vr or cv1 Costs for update around 600€ is it Worth and improves fps now max 45

Asw off

Edited August 14, 2017 by II/JG11ATLAN

[coconut]

 

 

will I see a noticeable difference using supersampling

 

That depends a bit on you, I would say. Personally I find the benefits of supersampling are most noticeable on the gauges, not so much on the landscape and objects outside the plane.

I think it's more noticeable in ED because of the black/white contrasts you tend to see in space, and the prevalence of small text in the HUD.

[=EXPEND=Tripwire]

Testing super-sampling is what pushed me to replace my PC. The difference should be noticeable when you swap between the two settings but at the end of the day you are still constrained by the limitations of the headset. If you don't feel the difference is enough when you test it now, then the only advantage you will see once you upgrade is in smoothness.

Its up to you to decide from that point if the investment in the upgrade is worthwhile based on your financial situation and the time you invest using the VR headset.

[von-Luck]

Some great advice in here! GPU's these days are in an interesting place and there are several considerations. Do you play other games - do you own a high res monitor? Buying a 1080/1080ti just for one facet of your hobby might be cost inefficient. 1080/1080ti are top end beasts that are waaaaaay overkill for a simple 1920x1080 display. That said the easy adage is buy what you can afford.

 

For VR 1080ti is king but calculate the cost to benefit and if you come out ahead go for it.

 

von Luck

[SCG_Fenris_Wolf]

I came from a similar setup. I bought a GTX1080, but the CPU does not cut it. According to the tests from chiliwili's thread, you really ought to go with a CPU/MB upgrade too. Make sure to get some fast 3k MHz RAM as well. GL! 

Edited August 14, 2017 by 1CAG_Fenris_Wolf

[Stratman59]

Thanks for all your answers, I've decided to go with a MSI 1080 gaming x plus. GPU costs are very high at the moment and as I really only game with the VR a 1080 ti was just too expensive especially as Fenris_Wolf says I may need to do a CPU/MB/RAM upgrade depending on how the 1080 works out.

Looking at some benchmark comparisons though even a 5 years old the 3770k CPU seems to compare quite favourably with many of the recent i5 cpu's recommended for the Oculus Rift.   

[chiliwili69]

CPU- I7 3770k

 

PLEASE, PLEASE, don´t think that upgrade your rig for Il-2 VR is just go to a top GPU. It DOES NOT solve the problem.

 

In IL-2 VR, CPU/RAM is the determining factor for keeping fps at 90fps with SS=1.0

 

Then, depending on how good your GPU is you can go further with SS.

 

I think that the upgrade rig procedure to enjoy IL-2 VR would be:

 

1. Check the "Passmark single thread" number of your CPU at: https://www.cpubenchmark.net/cpu_list.php

If the number is below 2500 you would better consider to upgrade your CPU (and possibly Motherboard)

for example your i7-3770K has a passmark single thread of 2084. (below desirable)

 

 

2. If your CPU doesn´t allow OC, upgrade to a CPU which allow OC and have a nice CPU cooler. Then do OC in your CPU. (don´t tell me it is complex, it is not)

 

3. If your RAM speed is below 1500MHz, consider to upgrade your RAM to 3000MHz or above.

 

4. Once you have done the 3 steps above run the Il-2 benchmark in VR (https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/).

If you get less than 70fps avg you might then upgrade your GPU, but you should not spend more than a 1070.

 

5. Now, set the SS=1.0 and run Il-2 benchmark, anote the result. Then make increases of 0.1 until you see your performance decrease dramatically (is when you fully load GPU).

Edited August 15, 2017 by chiliwili69

[chiliwili69]

 

 

Thanks for all your answers, I've decided to go with a MSI 1080 gaming x plus.

 

Upps, I think I am a bit late here.

If you already make the order it doesn´t matter as far you still have some more budget for CPU/RAM.

 

So, we can test how your i7-3770K performs with a 1080.  

 

Do you OverClock CPU? What RAM speed do you have? 

[Stratman59]

No worries chilwili69, I wanted to upgrade the GPU regardless. I think you're right though, whole system needs an upgrade as it's getting on a bit in PC terms, nearly 5 years old! I'll give overclocking the CPU a go although I've never done it before. Also my Ram is only 1200 DD3 I think? Another few weekends overtime coming my way by the looks of things. 

 

P.S.

Are you sure I should look at Passmark single thread score? Mine at 2084 against an i9 which only has a score of 2504? There are almost no CPU's with scores close 2500? 

Edited August 15, 2017 by Stratman59

[SCG_Fenris_Wolf]

Aye, Stratman, you can play on low/medium quality settings, HDR/SSAO off, mirrors off, with this CPU/RAM combination. The latter is drastically too slow, I believe the top MHz for RAM you can have at this range was 1600MHz DDR3 @XMP1.3

 

I assume the GTX1080 will improve your maximal SS while sticking to low/medium settings, as chiliwili69 said. 

 

 

I also got a gtx1080 and a similar processor (i5 3570k OCed to 4.3GHz with 16GB 1600MHz DDR3), and IL2 sometimes drops frames, sometimes the camera jutters around. I then tested it on a buddy's with a Kabylake i5 + gtx1070, and it ran much better. While other games are better optimized, IL-2 is, as chiliwili said, incredibly CPU hungry. I am currently waiting for my i7 7700K & friends to arrive. Just for IL-2

[dburne]

 I am currently waiting for my i7 7700K & friends to arrive. Just for IL-2

 

I have been so tempted the last couple of months to go for a new i7 7700K build, but am doing my best to hold off until sometime next year and see what is out there then. Maybe latter part of 2018, it will be 5 years old then ( put it together in Dec 2013). My rig still seems to be running my games pretty darn good even in VR so I really want to put this off as long as I can. 

It has been tempting though...

[A_radek]

Dburne, to make things easier for you regarding the temptation: I'm running the 7700k and it's only an upgrade, not nirvana in il2 VR. To be honest I feel a 6-7ghz CPU would be more like it and would have provided il2 a proper performance headroom. But that won't happen any day soon.

 

Il2 will still go down to 45fps in heavy situations, even with hud off. Difference is that it won't do it for as long as an older CPU.

 

Han has said eventually il2 will be multithreaded. So while still a gamble, safest bet today would be a i7 7740x for it's single thread performance. And then drop in an i9 when more games go multithreaded. Both 7740x and this first gen i9 use the same socket/motherboard. The 7700k motherboard chipset is a little cheaper but it's also a dead end.

Edited August 16, 2017 by a_radek

[TUS_Samuel]

for example your i7-3770K has a passmark single thread of 2084. (below desirable)

 

My 3770K with air cooler gives score 2700 single threaded at 4.7-4.9GHz 1.350 mV. Later I will post benchmarks in neighbour thread

[A_radek]

My 3770K with air cooler gives score 2700 single threaded at 4.7-4.9GHz 1.350 mV. Later I will post benchmarks in neighbour thread

Looking very much forward to that!

[BeastyBaiter]

The single thread part is an issue for sure. I finally got all my issues with the some software conflicts resolved but the real limiting factor in BoS (and DCS) is single thread CPU speed. I'm running an R5 1600X at 4.0 GHz, it has a single thread passmark rating of about 2200 at that clock. Both DCS and BoS play 45 FPS most of the time regardless of detail level in VR. I can get them to hit 90 sometimes, but there isn't much headroom for other stuff. the real problem is simply the way the games are coded. They need to spread the graphics load across two CPU cores instead of putting it all on one. That would make pretty much any modern-ish CPU be able to run the games at 90 fps. Allegedly it is coming for both sims, but isn't here yet. I find 45 fps acceptable though, since it is 45 fps all the time with no dips and the details can be cranked way up. This is with a 1080 TI, which is fully utilized in DCS 2, barely touched in DCS 1.5 and reasonably well used in BoS (70% usage is typical, which means it would still be at 45 fps with my settings if the CPU didn't bottleneck).

Edited August 16, 2017 by BeastyBaiter

[TUS_Samuel]

They need to spread the graphics load across two CPU cores instead of putting it all on one. 

 

Agreed. Given that intel has 18-core CPUs single thread performance won't increase dramatically in the near future. Also big cache size may help because benchmarks show that memory clocks matter

[dburne]

Dburne, to make things easier for you regarding the temptation: I'm running the 7700k and it's only an upgrade, not nirvana in il2 VR. To be honest I feel a 6-7ghz CPU would be more like it and would have provided il2 a proper performance headroom. But that won't happen any day soon.

 

Il2 will still go down to 45fps in heavy situations, even with hud off. Difference is that it won't do it for as long as an older CPU.

 

Han has said eventually il2 will be multithreaded. So while still a gamble, safest bet today would be a i7 7740x for it's single thread performance. And then drop in an i9 when more games go multithreaded. Both 7740x and this first gen i9 use the same socket/motherboard. The 7700k motherboard chipset is a little cheaper but it's also a dead end.

 

Whew, good that helps me in holding off. I have spent enough money this year as it is!

 

The nice thing for me with IL-2, is it rarely goes down to 45 fps for very long, and even when it does it is still very smooth and stutter free. Never drops any lower than 45 fps ( ASW off ).

 

Thanks,

Edited August 16, 2017 by dburne

[chiliwili69]

 

 

Are you sure I should look at Passmark single thread score? Mine at 2084 against an i9 which only has a score of 2504? There are almost no CPU's with scores close 2500? 

 

The Passmark  Single Thread indicator reported by the web is just an average of all the passmark single thread tests performed by the users (the number of samples is indicated).

Those samples includes non-OCed and OCed systems. So this number should be read with caution, since it is a live number, always changing as more tests are reported by the users.

 

For example, the 4790K (392$) has about 14000 samples with ST index of 2530, it is an old CPU which has been OC'ed a lot and many test has been done. 

 

The 7700K (307$) has about 3000 samples with ST index of 2584, it is a newer CPU with less tests, and radek obtained a ST index of 2806 with OCing at 4.8GHz

 

The 7740X (349$) has only 17 samples with ST index of 2653, it is really new CPU with just few tests.

 

It is true  very very few CPUs has the ST index above 2500, but it really depends on the OCing used for the passmark test. As Samuel reported, he obtained a 2700 ST index with the 3770K.

[chiliwili69]

 

 

safest bet today would be a i7 7740x for it's single thread performance.

 

I think the same.

 

They are in the same price range and  the 7700K socket (LGA 1151) is a dead-end. As the dead-end of my socket (LGA 1150) three  years ago (but still surviving).

 

Rule: Just save 300$ from the graphics card an put them in the CPU/RAM and a good cooler to OC as much as possible.

[chiliwili69]

 

 

They need to spread the graphics load across two CPU cores instead of putting it all on one

 

On Dec-2016, Han said "One day when we will split simulation to several threads - cores will supreme"

 

So, we should assume that it will be something feasible to be do it.

 

But this "one day" doesn´t give a clue of timeframe. It could be this year or in 2020.

 

What it is clear that if they do it,  they will de-bottleneck the VR a lot. 

[SCG_Fenris_Wolf]

Well, I got an Asrock Z270 K6 in a sale for 99€ , quite the departure from a comparable board going at 300€ for the Socket 2066. With the Socket 1151, I can also use my 120€ watercooling set instead of buying a new one, hence I went with the i7 7700K. <4% stock performance gain for 320€ more costs - what choice is better here is obvious. 

Edited August 17, 2017 by 1CAG_Fenris_Wolf

[A_radek]

Well, I got an Asrock Z270 K6 in a sale for 99€ , quite the departure from a comparable board going at 300€ for the Socket 2066. With the Socket 1151, I can also use my 120€ watercooling set instead of buying a new one, hence I went with the i7 7700K. <4% stock performance gain for 320€ more costs - what choice is better here is obvious.

 

You got a good deal and I didn't mean to imply you made the wrong choice. I recently made the same after much hesitation. In fact even after reading about the i7 8400k I don't have much regret. In my case I needed the upgrade urgently (i7 920 previously) got a good deal and figured cpu market is so turbulent now who knows what will be the hot ticket in 3 years and gen2 VR makes an entry. Better not spend it all now.

[TG-55Panthercules]

The 7700K (307$) has about 3000 samples with ST index of 2584, it is a newer CPU with less tests, and radek obtained a ST index of 2806 with OCing at 4.8GHz

 

All this got me curious so I went in and ran another test on my 7700K - ST passmark result of 2647, running at 4.2 (I think - can't tell if it's running in turbo mode (4.5) or not - CPU-Z seems to think it is, but Passmark seems to think it's not) - I'm going to make one more try at OC'ing to 4.5 and see what happens.

 

Tried using the ASUS utility that came with my MB to OC my CPU - it got it up to 5.0 before I stopped it.  Ran a new Passmark test at 5.0, and the ST index/result jumped to 2937.  Gotta run another test with the VR benchmark to see if there's any practical benefit at this level.

 

[EDIT] - surprisingly (to me), the test on the VR benchmark (from this thread: https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/page-1 )with my CPU OC'd to 5.0 did not result in any improvement at all compared to running it at stock (either 4.2 or maybe 4.5 turbo - can't tell whether turbo was used or not) - actually a slight reduction at 5.0, but so close as to be within the realm of standard deviation.  But if it's correct, it would seem to indicate a point of significantly diminishing returns to overclocking CPU at this point.

Edited August 17, 2017 by TG-55Panthercules

[=ARTOA=Rauchenderkolben]

hey guys, im currently thinking of upgrading my 980ti for vr purposes. used 1080s are around 100€ more expansive the 1080ti is 300€ more expansive. ( got 2 offers on the table) 

I do own a i7 4790k with some 1800mhz 16 gig ram, ssd's, and I'm asking myself if the money is worth upgrading and if yes, wich one. I currently have 90 fps on suuuper low settings, on berloga they may drop down to 40 fps when encountered a battle with 15 airplanes. I wonder if a 1080 or 1080ti will make it possible to me to crank up the settings from low to medium towards high while keeping 90fps also in more crowded dogfights.

Il2 is my maingame, thats why i'd like to invest in it, make the pictures prettier while I spend my time flying. Also I'd hope to be able to increase the viewdistance as I mostly see dots when we are within a range of 3-4km, so quite close ! As well as I see a sharp boarder of the rendered trees, doesn't look very pleasing seeing a harsh line where trees stop beeing rendered. If I turn my head, that line moves accoardingly :S

Does anyone know if il2 currently even takes advantage of the new VR-technology-features of the 1080's vs the 980ti or does it simply come down to raw performance. How about the sli vr tech, if you own a sli setup, one renders the pic for the left eye and one for the right ?

Anyone with a 1080 ti running on high VR settings with buttersmooth results even in crowdy places ?
Anyone dissapointed by the VR performance in il2 with a 1080 ti and hoped for more fps ?

Edited August 18, 2017 by =ARTOA=Rauchenderkolben

[chiliwili69]

 

 

my CPU OC'd to 5.0 did not result in any improvement at all compared to running it at stock

 

I am glad to see that you can mamange OCing with an Asus tool. Which one do you use?

 

That´s is strange to me as well. My experience is that in VR I was getting about almost 1 fps with the test for every 0.1 GHz increase in the overclocking.

 

Your memory is also a top one, but please verify with CPU-Z that it is effectively running at 3200Mhz and also that you apply OC to all the cores, not only to one or two cores.

 

You can run a series of tests from 4.2 to 5.0 with increments of 0.1 and verify the fps impact. Radek has the same i7-7700K but at 4.8 and heis having a good results.

[chiliwili69]

 

 

the sli vr tech

 

VR doesn´t use yet SLI.

If you refer tot he VRWorks tech from pascal GPUs, IL-2 doesn´t use it. 

 

If you have an 4790K you could invest in good CPU cooler and fast speed (fastest supported by your MoBo), this will increase you fps in single player.

I have not experience in multiplayer.

[TG-55Panthercules]

I am glad to see that you can mamange OCing with an Asus tool. Which one do you use?

 

That´s is strange to me as well. My experience is that in VR I was getting about almost 1 fps with the test for every 0.1 GHz increase in the overclocking.

 

Your memory is also a top one, but please verify with CPU-Z that it is effectively running at 3200Mhz and also that you apply OC to all the cores, not only to one or two cores.

 

You can run a series of tests from 4.2 to 5.0 with increments of 0.1 and verify the fps impact. Radek has the same i7-7700K but at 4.8 and heis having a good results.

 

I used the ASUS AI Suite 3 tool that came with my mainboard support disk - it opens up in something called Dual Intelligent Processors 5, which has a 5-way optimization function that automatically steps through overclock settings in .1 increments until it reaches whatever target level you set it for (or you chicken out and stop it, which is what happened when it reached 5.0).  And yes, according to the ASUS tool I was overclocking all cores in synch, and CPU-Z indicated all 4 running at the targeted OC speeds.

 

I'm not sure how to tell from CPU-Z what it thinks the RAM is running at (the only numbers I see are confusing - under the memory tab it's showing only about 1505, and under the SPD tab its showing numbers between 1066 and 1600).  The BIOS shows it running at 3200 (instead of the lower number it showed before I changed it to 3200 in the BIOS).  Passmark shows "Clk: 1600.0MHz" under its system info tab.

 

I tried OC'ing to 4.6 (to eliminate the turbo/4.5 question), but although the BIOS and CPU-Z showed it running at 4.6 the VR benchmark results were no different from the ones I ran at 4.2 or 5.0.  There must be something wierd going on here, but so far I haven't been able to figure out what's happening.

Edited August 19, 2017 by TG-55Panthercules

[dburne]

 

Anyone with a 1080 ti running on high VR settings with buttersmooth results even in crowdy places ?

Anyone dissapointed by the VR performance in il2 with a 1080 ti and hoped for more fps ?

 

I am extremely pleased with my 1080 Ti and BoS in VR, and I came from a 1080.

Ultra settings, 2x AA, sharpen filter - landscape x4.

[chiliwili69]

I'm not sure how to tell from CPU-Z what it thinks the RAM is running at (the only numbers I see are confusing - under the memory tab it's showing only about 1505, and under the SPD tab its showing numbers between 1066 and 1600).  

 

In CPU-Z you should go to the memory tab, in the DRAM frequency.

For example, in the image below my laptop is showing 1197 for the I/O bus frequency. For all DDR memories you shold multiple this by 2, so it is 2400MHz.

In your case you are running at 3000MHz.

 

In the SPD tab, in the Timings table you will see all the profiles that you memory is able to be configured, including XMP profiles.

I think you have a 1600 profile (2x is 3200). To activate this profile you will need to enable XMP in the BIOS, so you will be able to run at your advertised speed.

Edited August 19, 2017 by chiliwili69

[Stratman59]

Seems I started an interesting debate! I'm not really into overclocking so I decided to do a rebuild of my rig as it was over 6 years old. 

I went for the i7 7700k on a Asus Z270 board with 16 gig 3200mhzx DDR4 ram and an MSI 1080 plus GPU. Was it worth it? 

Well I now get a solid 90FPS but TBH the difference really isn't that noticeable, having said that I've only ever flown with 1 opposing aircraft, maybe in a crowded sky it'll pay dividends. I needed to upgrade anyway as I do Cad work but for VR game wise; I don't think I've gained an awful lot?  

[TG-55Panthercules]

Seems I started an interesting debate! I'm not really into overclocking so I decided to do a rebuild of my rig as it was over 6 years old. 

I went for the i7 7700k on a Asus Z270 board with 16 gig 3200mhzx DDR4 ram and an MSI 1080 plus GPU. Was it worth it? 

Well I now get a solid 90FPS but TBH the difference really isn't that noticeable, having said that I've only ever flown with 1 opposing aircraft, maybe in a crowded sky it'll pay dividends. I needed to upgrade anyway as I do Cad work but for VR game wise; I don't think I've gained an awful lot?  

 

Did you happen to run the benchmark track from this thread: https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/ on your old PC or your new one?  I upgraded from a PC better than your old one to one very similar to your new one, and I've seen quite a significant improvement in my VR experience as reflected by the VR test in that thread (and visually as well), as noted below:

 

Old PC:  Avg: 56.567 - Min: 36 - Max: 62 

 

New PC (no overclock):  Avg: 84.417 - Min: 53 - Max: 91 

 

New PC (overclocked to 5.0):  Avg: 87.183 - Min: 64 - Max: 91 (simple one-click overclocking preset in my BIOS).

 

Seems surprising that you wouldn't have noticed an even more significant improvement from your old PC to your new one.

[Stratman59]

No unfortunately I didn't have time, too busy working to pay for the new set up (-: 

I had a few hours last night to really test out my new system, I can set all the bells and whistles on and still keep a solid 90fps so obviously that's a big improvement. What I really don't notice is much of an diference  when setting supersampling, I'm not even sure Its' working as even when setting it to 2.0 (in debug tool) my frame rates stay at a solid 90, surely there should be some hit?  

[TG-55Panthercules]

No unfortunately I didn't have time, too busy working to pay for the new set up (-: 

I had a few hours last night to really test out my new system, I can set all the bells and whistles on and still keep a solid 90fps so obviously that's a big improvement. What I really don't notice is much of an diference  when setting supersampling, I'm not even sure Its' working as even when setting it to 2.0 (in debug tool) my frame rates stay at a solid 90, surely there should be some hit?  

 

I would have thought so, but I just ran a quick test in my current PC config (in signature) using same/usual VR test settings except increased SS from 1.7 to 2.0 in Steam VR, and didn't see much of a hit either (haven't check the visual impact - just looking at the FRAPS results):

 

SS1.7:  Avg: 87.183 - Min: 64 - Max: 91

 

SS2.0:  Avg: 86.767 - Min: 62 - Max: 91

 

I think I'm going to do a few more tests and ratchet up the SS until I see if there's some sort of cliff/drop-off in performance, and then see if the visuals look much better quality at whatever higher SS results from those tests.

[chiliwili69]

 

 

SS1.7:  Avg: 87.183 - Min: 64 - Max: 91   SS2.0:  Avg: 86.767 - Min: 62 - Max: 91

 

Yes, this is exactly where a top GPU like yours will help.  When I will back to city-home I will also run a test to see where my fps drops with a 1070 (I guess it will be around 2.0 in SteamVR).

 

I will add this info to the spreadsheet since it is also useful information for other SS settings.

[SCG_Fenris_Wolf]

All this got me curious so I went in and ran another test on my 7700K - ST passmark result of 2647, running at 4.2 (I think - can't tell if it's running in turbo mode (4.5) or not - CPU-Z seems to think it is, but Passmark seems to think it's not) - I'm going to make one more try at OC'ing to 4.5 and see what happens.

 

Tried using the ASUS utility that came with my MB to OC my CPU - it got it up to 5.0 before I stopped it. Ran a new Passmark test at 5.0, and the ST index/result jumped to 2937. Gotta run another test with the VR benchmark to see if there's any practical benefit at this level.

 

[EDIT] - surprisingly (to me), the test on the VR benchmark (from this thread: https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/page-1 )with my CPU OC'd to 5.0 did not result in any improvement at all compared to running it at stock (either 4.2 or maybe 4.5 turbo - can't tell whether turbo was used or not) - actually a slight reduction at 5.0, but so close as to be within the realm of standard deviation. But if it's correct, it would seem to indicate a point of significantly diminishing returns to overclocking CPU at this point.

IL 2 in VR does not run at the overclock you have made in passmark, it throttles down to 4.6GHz due to AVX instructions in either SteamVR or IL2 itself once you start the testrun, I found out. So any test run of overclocked Intels that have not specifically removed the AVX throttling will be false. That's probably why I achieved best results in that performance comparison even though I didn't get the highest overclock. I upped AVX throttling to 4.7GHz.

 

By default, my board for example even throttles down due to AVX to just 4.2GHz!

Edited September 3, 2017 by 2./JG51_Fenris_Wolf

[A_radek]

IL 2 in VR does not run at the overclock you have made in passmark, it throttles down to 4.6GHz due to AVX instructions in either SteamVR or IL2 itself once you start the testrun, I found out. So any test run of overclocked Intels that have not specifically removed the AVX throttling will be false. That's probably why I achieved best results in that performance comparison even though I didn't get the highest overclock. I upped AVX throttling to 4.7GHz.

 

By default, my board for example even throttles down due to AVX to just 4.2GHz!

I could be mistaken. But as far as I know the Avx offset feature only first appeared with kabylake. I run 0 Avx offset. Would be interesting to test how much of a difference it really does. Did you notice any improvement while oc:ing before you changed your default Avx offset?

[SCG_Fenris_Wolf]

Default offset was in relation to actual OC and always kept the clock at 4.2 for AVX load.

 

I monitored clock while running test tracks, it stuck to 4.2, no thermal throttling, AVX was called

 

I removed AVX offset, clock 4.9, monitored, ran test, massive increase. Game would not run stable more than 5 minutes under load though.

 

Continued to lower it, AVX 0, adjusted CPU llc. Goal: Run infinitely without crash with highest clock.

 

Stable at 4.7, AVX 0, llc 1 (static), vcore 1.32 V.

 

Anything higher would give freezes whenever many aircraft entered the scene, like on typical WoL/TAW evening. People think you disconnect, not advisable, but only environment you can properly test continuously in! Aiming for stability!

 

Today I will run at 4.8, AVX 0, llc 2 (linear decrease of current to load), 1.33V. Maybe I can get better stability at load peaks with llc 2. IL 2 seems to be quite ruthless in its demand spikes. Simply speaking, you can spawn on an airfield with 12+ players or more in the 10km bubble and a town nearby (yesterday TAW, Staritsa airfield) and it holds the knife to your system's throat and screams with Arnie's voice "Load all that shit naow!!" And then your CPU goes apeshit and the VRM screws it and itself freezing the system, and Arnie slashes with the knife.

Edited September 3, 2017 by 2./JG51_Fenris_Wolf

[SCG_Fenris_Wolf]

That said, Intel cleared i7 7700K for 4.6 under AVX AFAIK (read this in some OC forums). It's not meant to go higher continuously.

You can get nice passmark tests, and good fps results short term, but try flying on a full server at max settings load for 2 hours at a clock higher than 4.7 with AVX offset 0 and automatic Llc (or static 1).

 

Interested to what you'll find, please report back.

 

Any ideas, criticism and suggestions on what I can improve are welcome. I had not OCed in 4 years and was kind of out of it. Did I miss anything? Oh, BTW, thermals are not an issue with my setup. Solely clock under load with AVX and VRM spikes are the ceiling here for continuous stability.

Edited September 3, 2017 by 2./JG51_Fenris_Wolf