Jump to content
Sign in to follow this  
chiliwili69

How SS decreases your fps (testing results)

Recommended Posts

Most of us are using SuperSampling (SS) to increase the quality of the images in VR, as described here:

https://forum.il2sturmovik.com/topic/30771-how-much-ss-desirable-image-quality-samples/

 

But this nicer graphics has a price. The more SS, the more load to our GPU and CPU as well.

 

I have conducted three series of tests (LOW in green, BALANCED in orange, HIGH in red) using the IL-2 VR performance test described in the another post. For every test I changed the SS in SteamVR.

 

The used rig is i7 4790K 4.6 GHz; 16Gb 4x4 DDR3 2933 MHz; MSI Z97M-Gaming; Nvidia GTX 1070; Windows 10 64 bits; Oculus Rift.

 

Overall conclusion is that for my rig I could run at HIGH with SS=1.8, or at BALANCED with SS=2.8 or at LOW with SS=3.2

So, I will need to see what I am losing with BALANCED settings and what I am gaining with SS=2.8

 

The results are shown in this graph:

post-18865-0-61580600-1505689821_thumb.jpg

 

The text test data:

HIGH
SS=1.0 Avg: 84.950 - Min: 46 - Max: 91
SS=1.2 Avg: 83.467 - Min: 46 - Max: 91
SS=1.4 Avg: 82.533 - Min: 45 - Max: 91
SS=1.6 Avg: 80.483 - Min: 44 - Max: 91
SS=1.8 Avg: 76.550 - Min: 45 - Max: 91
SS=2.0 Avg: 74.100 - Min: 44 - Max: 91
SS=2.2 Avg: 68.117 - Min: 44 - Max: 84
SS=2.4 Avg: 60.717 - Min: 39 - Max: 74
SS=2.6 Avg: 61.100 - Min: 41 - Max: 75
SS=2.8 Avg: 52.750 - Min: 33 - Max: 68
SS=3.0 Avg: 51.467 - Min: 31 - Max: 67
SS=4.0 Avg: 43.983 - Min: 26 - Max: 47
 
BALANCED
SS=1.0 Avg: 88.667 - Min: 70 - Max: 91
SS=1.4 Avg: 87.617 - Min: 63 - Max: 91
SS=1.8 Avg: 86.483 - Min: 56 - Max: 91
SS=2.0 Avg: 86.100 - Min: 47 - Max: 91
SS=2.2 Avg: 85.767 - Min: 46 - Max: 91
SS=2.4 Avg: 79.217 - Min: 44 - Max: 91
SS=2.6 Avg: 75.417 - Min: 45 - Max: 90
SS=2.8 Avg: 79.667 - Min: 45 - Max: 91
SS=3.0 Avg: 75.817 - Min: 45 - Max: 91
SS=3.2 Avg: 64.367 - Min: 42 - Max: 75
SS=3.4 Avg: 59.333 - Min: 36 - Max: 71
SS=4.0 Avg: 56.700 - Min: 36 - Max: 71
 
LOW
SS=1.0 Avg: 89.383 - Min: 86 - Max: 91
SS=1.8 Avg: 89.117 - Min: 81 - Max: 91
SS=2.2 Avg: 88.883 - Min: 76 - Max: 91
SS=2.6 Avg: 86.733 - Min: 64 - Max: 91
SS=2.8 Avg: 87.933 - Min: 66 - Max: 91
SS=3.0 Avg: 83.617 - Min: 61 - Max: 91
SS=3.2 Avg: 73.300 - Min: 44 - Max: 81
SS=3.4 Avg: 69.850 - Min: 44 - Max: 79
SS=3.6 Avg: 69.000 - Min: 44 - Max: 78
SS=4.0 Avg: 58.083 - Min: 44 - Max: 70

 

 

 

Edited by chiliwili69
  • Upvote 2

Share this post


Link to post
Share on other sites

I haven't run a lot of rigorous tests with SS yet (other than that first set I ran a while back where on my rig I saw a big fall-off between 2.3 and 2.6 while running my maxed out settings).  I did try a few flights with Fenris' SS-heavy settings (basically using high and turning off in-game AA and sharpen and using SS = 2.9 or even 3.2), but I was taking a pretty hard FPS hit from the elevated SS and not getting graphics quality results I liked, so I've gone back to running Ultra with mostly maxed settings and SS=2.3.

 

If I have some time tonight I'll run some benchmark settings tests of SS (except on Ultra) and see how they compare.

 

[EDIT]

 

Specs:  Intel Core i7-7700K @ 4.6 GHz (Corsair liquid cooler); 16GB RAM (Corsair Vengeance DDR4 3200MHz);  EVGA GTX 1080 TI SC2, 11 GB; ASUS ROG Maximus IX Code MB; Realtek ROG SupremeFX audio;  Windows 10 Pro 64-bit; Oculus Rift (CV1)

 

The text test data:

ULTRA
SS=1.0  Avg: 73.467 - Min: 44 - Max: 91
SS=1.2  Avg: 69.050 - Min: 44 - Max: 91
SS=1.4  Avg: 69.133 - Min: 44 - Max: 91
SS=1.6  Avg: 68.483 - Min: 44 - Max: 91
SS=1.8  Avg: 66.867 - Min: 44 - Max: 91
SS=2.0  Avg: 64.883 - Min: 44 - Max: 91
SS=2.2  Avg: 66.717 - Min: 44 - Max: 91
SS=2.4  Avg: 64.967 - Min: 44 - Max: 91
SS=2.6  Avg: 64.383 - Min: 44 - Max: 91
SS=2.8  Avg: 63.367 - Min: 44 - Max: 91
SS=3.0  Avg: 63.333 - Min: 44 - Max: 91
SS=3.2  Avg: 61.650 - Min: 44 - Max: 85
SS=3.4  Avg: 61.133 - Min: 44 - Max: 81
SS=3.6  Avg: 57.617 - Min: 44 - Max: 79
SS=3.8  Avg: 57.233 - Min: 43 - Max: 76
SS=4.0  Avg: 57.450 - Min: 41 - Max: 74
Edited by TG-55Panthercules
  • Upvote 1

Share this post


Link to post
Share on other sites

Good to see clear and well presented tests.

 

My take away from all of those is that only a few data points in the two sets presented shows an average fps below 60 (and not much)  which is the limit of my TV used as monitor. My other hardware is very similar to the OP's.

 

I understand that a higher average may help to ensure that FPS does not sometimes spike below 60 fps - but it does seem that for the first time I can remember in my gaming history the hardware is actually winning the arms race against the software.

Edited by unreasonable

Share this post


Link to post
Share on other sites

 

 

Intel Core i7-7700K @ 4.6 GHz (Corsair liquid cooler)

 

In the past you were OCing to 4.9 or 5.0. You have a good liquid cooler, what is the reason to OC "only" to 4.6?

I was considering to put a better CPU cooling (liquid perhaps) but not sure.

Share this post


Link to post
Share on other sites

 

 

which is the limit of my TV used as monitor.

 

This test were run using Oculus Rift VR, not in monitor.

If we run these tests in monitor (FullHD 1080p) we achieve about double fps.

 

the test procedure and results are here:

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/


 

 

I understand that a higher average may help to ensure that FPS does not sometimes spike below 60 fps - but it does seem that for the first time I can remember in my gaming history the hardware is actually winning the arms race against the software

 

You have to look at the avg but also the min values. if you min value goes below 45 is not acceptable at all in VR.

In you case the min value should not go below 60 but in monitor is not as tragic as in VR.

Share this post


Link to post
Share on other sites

Appreciate the clarification.  My monitor is 4K, so potentially quite a lot slower than full a normal HD monitor, hence my interest in the VR figures.  For me going below 60 is not too bad, generally only happens with cloud about or in free camera view (grrrr).  Below about 40ish I start to notice when actually playing.   

 

I doubt that I will be into VR anytime soon, but monitoring developments with interest.

Share this post


Link to post
Share on other sites

In the past you were OCing to 4.9 or 5.0. You have a good liquid cooler, what is the reason to OC "only" to 4.6?

I was considering to put a better CPU cooling (liquid perhaps) but not sure.

If I remember correctly from another Thread, he mentioned he wanted to keep the game stable during longer play sessions. The AVX instructions basically fry the cpu and vrm, 4.6GHz is the maximum that has been cleared by Intel. You can go 4.7, and maybe even 4.8, but even at 4.8 the game might just come to a random halt after 2 hours. In MP, usually after you have just gotten shot at, as Murphy's Law dictates  :biggrin:

 

 

P.S. I don't know precisely what this is, but measurable Heat is not the deciding factor to stability in this regard anymore. On Z270 + Kaby Lake, a proper airflow over the CPU is necessary, because the VRM heatsinks are much more needed. Z97 (your chip) have them as well, but they aren't as bottleneck'y heatwise. It might be connected to the VRM and dynamic load-line-calibration, since its spikes and bad heat dissipation were what let systems crash of others in overclocker forums. After 4.6/4.7 the improvement for IL-2 isn't really noticable anymore anyway, certainly not enough to justify frying and torturing your CPU in its youth, so it is the smartest thing Pantherculus could do. For SS the bottleneck will be gpu anyway.

Edited by 2./JG51_Fenris_Wolf
  • Upvote 1

Share this post


Link to post
Share on other sites

In the past you were OCing to 4.9 or 5.0. You have a good liquid cooler, what is the reason to OC "only" to 4.6?

I was considering to put a better CPU cooling (liquid perhaps) but not sure.

 

At the higher overclocks I was getting an annoying frequency of short but loud "fan spikes", which seemed to be related to very short-term temp spikes. The temps never seemed to go over about 60, but the rapid rise and fall of the fan speed/noise was pretty annoying, especially since the rig runs so quietly the rest of the time.  I tried various combinations of fan speed profiles trying to smooth out the spikes, but no luck.  It still happens every once in a while now at 4.6 but not nearly as frequently as at the higher OCs. Also, I saw some posts somewhere about maybe Intel didn't design these CPUs to run AVX at higher than 4.6 for extended periods of time, such that while the short term benchmark tests results looked better at the higher OCs they wouldn't really hold up when flying BoS missions over extended periods of time.  That idea, coupled with the fan spikes I was getting, has led me to fall back to 4.6, at least for now.

Edited by TG-55Panthercules
  • Upvote 1

Share this post


Link to post
Share on other sites

 

 

After 4.6/4.7 the improvement for IL-2 isn't really noticable anymore anyway, certainly not enough to justify frying and torturing your CPU in its youth

 

I couldn´t do testing beyond that with the BOS benchmark, but I run until 4.9 the STMark as I said in another thread you were involved as well:

 

OC 4.7Ghz  STMark: 2826

OC 4.8Ghz  STMark: 2889

OC 4.9Ghz  STMark: 2945

 

So, for an STMArk of 2945 the VR expected with the test is almost 90 fps. 

When I have some time I will try to run this short tests at higher than 4.6.

Share this post


Link to post
Share on other sites

 

 

Also, I saw some posts somewhere about maybe Intel didn't design these CPUs to run AVX at higher than 4.6 for extended periods of time

 

That´s bad. It is amazing that a flagship CPU like the 7700K with base clock 4.2, can only OC 0.4 for large periods, even with a good liquid cooling like yours.

There should be a solution for this somewhere. I hope future 8700K will not suffer from this.

Share this post


Link to post
Share on other sites

That´s bad. It is amazing that a flagship CPU like the 7700K with base clock 4.2, can only OC 0.4 for large periods, even with a good liquid cooling like yours.

There should be a solution for this somewhere. I hope future 8700K will not suffer from this.

 

 

Also, I saw some posts somewhere about maybe Intel didn't design these CPUs to run AVX at higher than 4.6 for extended periods of time, such that while the short term benchmark tests results looked better at the higher OCs they wouldn't really hold up when flying BoS missions over extended periods of time.  That idea, coupled with the fan spikes I was getting, has led me to fall back to 4.6, at least for now.

 

I don't seem to suffer from this issue. Can run Avx all day long @4.8. Fan spikes, yes, but as I'm running two 140mm for the cpu it's all still very quiet. (actually my old monitor with no moving parts at all is louder than my computer, makes some electric buzzing sound it deserves a slow death for)

 

Out of curiosity what liquid cooler is it your using? Contrary to common belief the smaller Aio Liquid coolers are not necessarily very good.

Share this post


Link to post
Share on other sites

Yeah seems like I have seen many successful overclocking reports of the 7700k chip to 5 GHz over on other forums as well...

 

Granted you need a good cooler as it is a heat generator.

Share this post


Link to post
Share on other sites

I don't seem to suffer from this issue. Can run Avx all day long @4.8. Fan spikes, yes, but as I'm running two 140mm for the cpu it's all still very quiet. (actually my old monitor with no moving parts at all is louder than my computer, makes some electric buzzing sound it deserves a slow death for)

 

Out of curiosity what liquid cooler is it your using? Contrary to common belief the smaller Aio Liquid coolers are not necessarily very good.

 

I'm running a Corsair Hydro H75.  It seemed to handle OC'ing to 5.0 just fine for short periods like the benchmark testing, but the FPS improvement in the benchmark tests going from 4.6 to 5.0 were not that great (and at my maxed out settings, 5.0 was even a little worse than 4.6, as if some sort of down-clocking might be going on), and given the increase in fan spikes and the rumblings about AVX at >4.6 I just figured I might as well back it down to 4.6 and leave it there for a while.

Share this post


Link to post
Share on other sites

 

 

the FPS improvement in the benchmark tests going from 4.6 to 5.0 were not that great

 

in the VR test, using the test settings you achieved:

 

5.0 Ghz  88.2 avg fps  68 min fps

4.6 Ghz  84 avg fps  55 min fps

 

It only 4 fps for avg, but you should take into account that the closer you are to 90 the more difficult is to obtain a significant gain (90 is an asymptote).

In the min fps you get a substantial gain instead.

 

In order to get test the gained fps, you can run the same test, using just same settings but with Ultra graphics, so you will get results far from the 90 asymptote.

When you used your settings, If you obtained less fps at 5.0 than in 4.6, the reason would be that the 5.0 test was not really running at 5.0, so it is not a valid conclusion.

Share this post


Link to post
Share on other sites

Awesome test as always Chiliwili. Very good to have the numbers. In the end, what ss/graphics settings did you choose to play with?

Share this post


Link to post
Share on other sites

I'm running a Corsair Hydro H75.  It seemed to handle OC'ing to 5.0 just fine for short periods like the benchmark testing, but the FPS improvement in the benchmark tests going from 4.6 to 5.0 were not that great (and at my maxed out settings, 5.0 was even a little worse than 4.6, as if some sort of down-clocking might be going on), and given the increase in fan spikes and the rumblings about AVX at >4.6 I just figured I might as well back it down to 4.6 and leave it there for a while.

Checking comparison charts, the H75 seem to do very well. I have full understanding for your choice of going back to 4.6. Torturing a new computer for a few more fps never feels right.

Share this post


Link to post
Share on other sites

in the VR test, using the test settings you achieved:

 

5.0 Ghz  88.2 avg fps  68 min fps

4.6 Ghz  84 avg fps  55 min fps

 

It only 4 fps for avg, but you should take into account that the closer you are to 90 the more difficult is to obtain a significant gain (90 is an asymptote).

In the min fps you get a substantial gain instead.

 

In order to get test the gained fps, you can run the same test, using just same settings but with Ultra graphics, so you will get results far from the 90 asymptote.

When you used your settings, If you obtained less fps at 5.0 than in 4.6, the reason would be that the 5.0 test was not really running at 5.0, so it is not a valid conclusion.

Chili, as Pantherculus said before, this doesn't matter. He is having a valid conclusion. FPSmin is unimportant when considering it only appears for extremely short intervalls in time and stays above 45. Heat dissipation doesn't matter as well. A linear decreasing LLC mitigates the stability issues a bit. This all happens when AVX instructions are called. It might work for up to 5-10 minutes sorties, but after then, the stress in the CPU stacks up and hardlocks your system. I even had to reset CMOS twice. Sometimes at 4.9 it worked for 20 minutes, sometimes for 5. Both not enough to fly real sorties in IL-2, SP or MP.

 

On 5.0GHz overclock / 4.8GHz AVX I could run sometimes 1 hour, sometimes 30 minutes. 4.9GHz overclock / 4.7GHz AVX runs infinitely. Running infinitely without crashing is the major requirement.

 

 

What is not worth the higher FPSminimal +5 you get for a +100 or +200MHz AVX OC in an intervall of 3 seconds per 60 seconds:  2-3 game crashes with disconnects on Wings/TAW in one evening. The hassle to open your computer's case, reset CMOS, redo your OC afterwards. Then try to log back in to full servers while your comrades are waiting.

 

 

P.S. People say beheading an i7 will help with achieving a higher peak OC. I wonder if this holds true for a stable AVX OC as well. But it's a giant leap of faith. What will help are new technologies like VRworks, so the cpu load does not double for IL-2 in VR. 1 instance vs 2 instances. Or better multi-threading of IL-2's engine.

Edited by 2./JG51_Fenris_Wolf

Share this post


Link to post
Share on other sites

When you used your settings, If you obtained less fps at 5.0 than in 4.6, the reason would be that the 5.0 test was not really running at 5.0, so it is not a valid conclusion.

 

In the benchmark tests, the increased performance going from 4.6 to 5.0 was reasonably good as you noted, BUT I can't enjoy flying at those test settings (looks too crappy for my taste).  I can enjoy flying at my more maxed-out settings, so it's those results that really matter to me, and at those settings (for whatever reason, downclocking or otherwise) the results at a nominal OC of 5.0 weren't as good as at 4.6.  So, given that I'll be flying at my maxed settings, there's no reason for me to leave it at 5.0, regardless of what may be happening at 5.0 with the benchmark settings.

Share this post


Link to post
Share on other sites

 

 

results at a nominal OC of 5.0 weren't as good as at 4.6.

 

Yes, the benchmark settings are just to compare with peers but the importan thing is the fps you achieve with your personal settings.

The benchmark settings was just something not too demanding and not too light.

 

I have tried to capture in the online sheet most of your valuable tests, they are reflected in the different tab of the sheet. In the "Graph setting Test" I recorded a test at 4.6 GHz with 2.012c where you achieved 56.567 fps with your personal settings and shadows=high, but the tests at 5.0 were using other shadows=ultra or medium.

 

In summary, I don´t know where you reported the test at 4.6 and a 5.0 using your personal settings.

My only purpose here is to try to understand why you don´t gain fps when going from 4.6 to 5.0.

 

I run at 4.6 and I am thinking to go to higher freqs (with better cooling if needed) and experiment with OC only one core.

Share this post


Link to post
Share on other sites

In summary, I don´t know where you reported the test at 4.6 and a 5.0 using your personal settings.

My only purpose here is to try to understand why you don´t gain fps when going from 4.6 to 5.0.

 

 

There have been so many test iterations that I may not have gotten around to posting those specific results - they are below FWIW:

 

Game settings are as set forth in the attached startup.cfg file.  SS = 2.3 in SteamVR (both "enable" boxes below the SS slider are ticked).  Game version 2.012b for 5.0, 2.012b and 2.012c for 4.6 (I had already dropped back down to 4.6 by the time the 2.012c hotfix was released, and didn't re-run the 5.0 tests after the hotfix given that the difference between pre- and post-hotfix was pretty negligible at 4.6)

 

OC at 4.6 (v.2.012b):  Frames: 3357 - Time: 60000ms - Avg: 55.950 - Min: 44 - Max: 91

 

OC at 4.6 (v.2.012c):   Frames: 3394 - Time: 60000ms - Avg: 56.567 - Min: 44 - Max: 91

 

OC at 5.0 (v.2.012b):  Frames: 3155 - Time: 60000ms - Avg: 52.583 - Min: 44 - Max: 85

 

Panther's latest VR startup.zip

Share this post


Link to post
Share on other sites

 

 

OC at 4.6 (v.2.012b):  Frames: 3357 - Time: 60000ms - Avg: 55.950 - Min: 44 - Max: 91   OC at 4.6 (v.2.012c):   Frames: 3394 - Time: 60000ms - Avg: 56.567 - Min: 44 - Max: 91   OC at 5.0 (v.2.012b):  Frames: 3155 - Time: 60000ms - Avg: 52.583 - Min: 44 - Max: 85

 

Thanks for detailing the settings of this test. That´s weird you obtain less fps with more OC.

I would love to have a 7700K with cooling capacity and a 1080Ti to run the test with your settings and verify what is happening.

 

In any case, even if I reach 4.9Ghz with my CPU (just for the short IL-2 VR test) and current CPU cooler, my 1070 will not support SS=2.3 very well, so my results will not be comparable.

Maybe someone with similar rig to you could run the test with your setting to double check this strange lower fps at higher freq.

Share this post


Link to post
Share on other sites

I re-checked my test results and noticed that my results at 5.0 did take a bit of a nosedive between original 2.012 and the first hotfix 2.012b.  Maybe I'll try again at 5.0 later today and see if it's still lower than 4.6 or not with 2.012c.  I'll probably still run it at 4.6 long term because of the possible AVX/longevity issue, but I would like to understand why I'm not seeing better results at 5.0 for the tests at least.

Share this post


Link to post
Share on other sites

Thanks for detailing the settings of this test. That´s weird you obtain less fps with more OC.

I would love to have a 7700K with cooling capacity and a 1080Ti to run the test with your settings and verify what is happening.

 

In any case, even if I reach 4.9Ghz with my CPU (just for the short IL-2 VR test) and current CPU cooler, my 1070 will not support SS=2.3 very well, so my results will not be comparable.

Maybe someone with similar rig to you could run the test with your setting to double check this strange lower fps at higher freq.

If you plan on getting an i7 7700k, I recommend an air cooler. The high-end ones are just as effective as water coolers and cool the VRM rims on the board as well - some research on some pro OC forums indicated that they are the main issue for this lineup.

Share this post


Link to post
Share on other sites

 

 

If you plan on getting an i7 7700k

 

No plans, just love to have a rig like panther or yours.

My only short term plan is to study more the OC of the 4790K and perhaps acquire a better air cooler. And perhaps a proper ATX case (current one is micro-ATX and limit the since of CPU cooler) 

Share this post


Link to post
Share on other sites

OK - ran a couple more tests while OC'd to 5.0, to see what happens with 2.012c:

 

Benchmark test settings:

 

OC = 4.6:  Frames: 5040 - Time: 60000ms - Avg: 84.000 - Min: 55 - Max: 91

 

OC = 5.0:  Frames: 5210 - Time: 60000ms - Avg: 86.833 - Min: 56 - Max: 91

 

My "maxed out" settings:

 

OC = 4.6:  Frames: 3394 - Time: 60000ms - Avg: 56.567 - Min: 44 - Max: 91

 

OC = 5.0:  Frames: 3943 - Time: 60000ms - Avg: 65.717 - Min: 44 - Max: 91

 

That last result surprised me based on my tests with pre-hotfix 2.012, but I ran it again and got essentially the same result.  Given the performance hit that OVRDrop entails, those extra 9 avg FPS could come in handy - I may have to rethink leaving the OC at 5.0.  

Too bad we don't have a benchmark test track that lasts 30 minutes, to test over more typical mission time frames.

Share this post


Link to post
Share on other sites

I agree. On Sept 28th I will be back home and can fly again. I will record some sessions and pick a taxing one that lasts at least 30 minutes.

Share this post


Link to post
Share on other sites

I agree. On Sept 28th I will be back home and can fly again. I will record some sessions and pick a taxing one that lasts at least 30 minutes.

 

That would be great - I tried to make one last night in a QMB against 4 He-111s and 4 Ju-52s, but for some reason the recording I made only captured about 7 minutes of it instead of all 30-something. I may try again later today. 

Share this post


Link to post
Share on other sites

Just to add another data point to this discussion. My I7 7700K has been running without any stability issues at 4.9GHz with my AIO Corsair H115i water cooler since I got it last November and I have not had any issues with it. I also routinely do multi-hour sessions. Keep in mind, that I have also configured my bios so that my system does not downclock even during idle times. I do have really good airflow throughout my case though and also splurged on some really good Noctua fans designed to pull and push air through my 280mm radiator that the 115i came with. I ditched the stock Corsair fans the moment I unboxed it. Seems a lot of people are fixated on a fans RPM and not how much air it can actually move or if it was designed with radiators in mind. Anyway, that's a topic for a different day.

 

My point here is that while it may be true that some people and even Intel are warning against prolonged OC'ing, for me at least, 4.9Ghz has been stable for nearly a year now of sustained gaming and work. I also don't have any issue with the heating getting out of hand, even with CPU and GPU's maxed out. I have considered sending my 7700K off to Silicon Lottery to get delidded but it just hasn't been worth the hassle so far.

 

One last point, I was running at 5Ghz for a long time but I started running into some instability during benchmarking so rolled back to 4.9 which has been good ever since. Maybe I just got lucky with my chip. I did build this rig with overclocking in mind though so I also think that makes a big difference in results.

 

Madmatt

Share this post


Link to post
Share on other sites

Keep in mind, that I have also configured my bios so that my system does not downclock even during idle times. 

 

Any tips on exactly how this is done?  Although I would want to do this too, this latest BIOS on this MB is so complicated I'm not sure whether I've actually accomplished it or not.

 

Also, I still get these annoying "fan spikes" (don't know what else to call them) - where the fan(s) suddenly rev up very fast and noisy, only to cycle back down very quickly.  Then maybe 30 seconds to a minute later or so it happens again.  I haven't detected a specific pattern of occurence, and it doesn't only happen during high-stress gaming (sometimes happens while I'm just browsing these forums or other web pages).  They're short in duration, but very annoying, especially considering how quietly the box runs normally.  I've tried various fan speed profiles (trying to increase the idle/normal fan speed hoping to get in front of and prevent the underlying temp spikes), but haven't found anything that works yet (other than perhaps running fans at 100% all the time, which is too loud and doesn't seem good for fan life either).  It still happens at 4.6 but is much more frequent at 5.0, which is what led me to drop back down to 4.6 in the first place.

 

Any ideas what might cause this behavior and what might eliminate/minimize it?

Share this post


Link to post
Share on other sites

 

 

My "maxed out" settings:   OC = 4.6:  Frames: 3394 - Time: 60000ms - Avg: 56.567 - Min: 44 - Max: 91   OC = 5.0:  Frames: 3943 - Time: 60000ms - Avg: 65.717 - Min: 44 - Max: 91

 

These results are much more logical and aligned with the STMark correlations. 

So you gain about 2 fps for every 0.1 Ghz of OC. So OC over 4.6 also make sense.

Share this post


Link to post
Share on other sites

Panther,

 

The first thing to do to try and lock the CPU speed is to disable Intel SpeedStep in the bios. You have a very good motherboard, which should be similar to the bios mine runs. Do you know where to find that setting? 

 

Try that first but there are also some power options you may need to change as setting your voltage to "adaptive" in the bios can cause the CPU to downclock as well during idle or periods of low use, although that's not really a bad thing.

 

As for your fan spikes, this is a little more difficult to discern. Do you know for sure WHICH fan(s) it is that are/is speeding up? Your CPU is watercooled but you didnt say which type. Also are you running fans with your waterblock radiator? Could the noise be coming from the GPU fans or the case fans?  Are you running any monitoring utility to see what the RPM's and temp are on your CPU and GPU? How are you overclocking? Are you using a tool like MSI Afterburner or via bios or something like the Asus AI Suite? I think you mentioned setting fan curves before, what did you use to do that?

 

I used to get issues with my GPU fans doing the same thing, it turned out it was caused by running Gigabytes overclocking software that came with the card. Once I ditched it and settled on doing everything for my video card within MSI Afterburner, I was able to design a custom fan curve that doesn't spike and keeps my GPU nice and cool at all times. Since I play most games in VR now, I can be a bit more aggressive with my fan curve as I can't hear my PC with the headset on anyway. I think at idle temps (30C) I run at 40% of fan speed and it ramps up so that if my GPU his 60C or higher I am quickly in the 80-90's% of fan speed.

 

For my waterblock, i have it running the pump and fans at max speed at all times. The pump on the H115i is pretty quiet so the difference between max and quite mode was hard to really hear. I also use aftermarket Noctua fans which were designed for use with radiators and they also do a great job on keeping the noise (and temps) manageable. The rest of my case fans are lower rpm and low db types. My intake is at the front and bottom of the case with exhaust out the back and my radiator fans are in a push-pull configuration out the top.

 

Madmatt

Edited by Madmatt

Share this post


Link to post
Share on other sites

 

 

Too bad we don't have a benchmark test track that lasts 30 minutes, to test over more typical mission time frames.

 

I think I already commented this before with Fenris. This test is a IL-2 BOS test. It doesn´t matter if you can only keep the 5.0GHz during one minute or 1 hour.  The important thing is the results you get for those hardware/software settings.

 

If the problem is to maintain the OC at 5.0 during one hour or more, there are other simpler tools to stress your CPU (to stress your rig on single-thread mode you can use SuperPi with 32Million digits: http://www.superpi.net/Download/ ) and study how fans work and how your temperatures evolve.

 

Like the Passmark Test , our IL-2 VR is a performance test, not an stress test. These are two separate things.

Share this post


Link to post
Share on other sites

I think I already commented this before with Fenris. This test is a IL-2 BOS test. It doesn´t matter if you can only keep the 5.0GHz during one minute or 1 hour.  The important thing is the results you get for those hardware/software settings.

 

If the problem is to maintain the OC at 5.0 during one hour or more, there are other simpler tools to stress your CPU (to stress your rig on single-thread mode you can use SuperPi with 32Million digits: http://www.superpi.net/Download/ ) and study how fans work and how your temperatures evolve.

 

Like the Passmark Test , our IL-2 VR is a performance test, not an stress test. These are two separate things.

 

Chili - I don't disagree with you, but I do think there's a practical use for a longer BoS test (for me at least, though maybe not for the typical user or for normal benchmark purposes - your shorter benchmark track is better for that).  But for me, regardless if what some non-BoS-related stress test might show, if I can run BoS at my preferred settings and an OC level of 4.9 or 5.0 for 30 minutes or more without causing problems, then I'll be happy.  And the shorter benchmark test/track won't tell me that.

 

But you're correct that it may be more appropriate to try one of the real stress tests as part of my diagnostic efforts to figure out what my fans(s) are doing and why.

 

[EDIT]  Been messing around a little with my new fan profile settings and an OC of 4.9.  So far, so good - all anecdotal but no noticeable fan spikes so far.  Also confirmed your conclusion that each .1 of CPU OC should yield about 2 FPS (at the benchmark track settings at least - getting some odd FPS per .1 OC results at my "maxed out" settings - gotta do more testing). Results from benchmark settings test track:

 

OC at 4.9:  Frames: 5094 - Time: 60000ms - Avg: 84.900 - Min: 54 - Max: 91

 

OC at 5.0:  Frames: 5210 - Time: 60000ms - Avg: 86.833 - Min: 56 - Max: 91

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites

Panther,

 

The first thing to do to try and lock the CPU speed is to disable Intel SpeedStep in the bios. You have a very good motherboard, which should be similar to the bios mine runs. Do you know where to find that setting? 

 

Try that first but there are also some power options you may need to change as setting your voltage to "adaptive" in the bios can cause the CPU to downclock as well during idle or periods of low use, although that's not really a bad thing.

 

As for your fan spikes, this is a little more difficult to discern. Do you know for sure WHICH fan(s) it is that are/is speeding up? Your CPU is watercooled but you didnt say which type. Also are you running fans with your waterblock radiator? Could the noise be coming from the GPU fans or the case fans?  Are you running any monitoring utility to see what the RPM's and temp are on your CPU and GPU? How are you overclocking? Are you using a tool like MSI Afterburner or via bios or something like the Asus AI Suite? I think you mentioned setting fan curves before, what did you use to do that?

 

Thanks for all the tips.  I'll have to go into the BIOS again after my current scan ends to check out those things you mentioned.  I may have to open up my case and try to figure out exactly which fans are doing what - I'm not sure the ASUS utility (ASUS RoG AI Suite 3) I've been using to try to play with the fan profiles has them correctly labeled/identified.  My CPU cooler is a Corsair Hydro H75, running with the stock radiator fans that came with it.  I've been trying to monitor the temps with the ASUS utility above, but I haven't found any feature/capability to track it over time while BoS is running and retrieve the results in any sort of report/historical fashion (only seems to show it to you while you're actively looking at its graph/results in real time - I'm probably missing something about how it works because that doesn't seem very useful).  

 

I am trying a new fan profile I created (in AI Suite 3) and am currently running at an OC of 4.9, and so far the results have been much better from a "fan spike"/noise perspective, but I haven't had a chance to run any BoS tests at these settings yet.  I did the CPU overclock simply by typing 49 or 50 into the BIOS entry for the Core 1 ratio (it was already set to synch all cores, so they all changed to 4.9 or 5.0 depending on what I typed into the 1st core entry).  I had tried a couple of other methods (using the BIOS presets and also using AI Suite's overclocking applet), but just typing the number in that one BIOS entry seemed to work OK and was certainly easier.  But then again, maybe that's too simplistic.

Share this post


Link to post
Share on other sites

Panther, 

 

Setting it directly in BIOS is actually the best way to handle an overclock and you did it the way it should be done. If everything is stable at 4.9 by just typing in 49 into the ratio, then yes it is that simple and I would just leave it alone. I would be curious to see how your CPU voltage is set (if its set to a manual level or adaptive) but again, if its stable, then leave it alone.

 

As for running with ASUS Suite III, I too played around with that at first but eventually stopped using it and did all my tweaking in BIOS for my CPU and RAM timings. For my GPU, I do everything in MSI Afterburner.

 

For the fans, I control the fans attached to the watercooler via CorsairLink. For the 1080Ti, i control the fans with a custom fan curve in MSI Afterburner and for my case fans I don't their modulate speeds so I use their built in fan controller and run at peak speeds all the time.

 

You may want to run the app CPUID  which will report both your true clock speed and voltage on your CPU. I am curious as to what voltage its reporting. From what I have read, a 7700K has about a 80% chance on stock voltages of running at 4.9GHz so its sounds like you are right where you need to be.

 

Madmatt

 

p.s. By the way, their are lots of tools that allow you to look back at performance stats. MSI Afterburner is again my tool of choice for that. I can go back in time and look at what my CPU's were doing, what temps and speeds my graphics card was doing and more. If you dont have it, i would highly recommend it. It is manufacturer agnostic so it will work with a Aorus Gigabyte cards like mine, EVGA, ASUS and others.

Edited by Madmatt

Share this post


Link to post
Share on other sites

 

 

p.s. By the way, their are lots of tools that allow you to look back at performance stats. MSI Afterburner is again my tool of choice for that. I can go back in time and look at what my CPU's were doing, what temps and speeds my graphics card was doing and more. If you dont have it, i would highly recommend it. It is manufacturer agnostic so it will work with a Aorus Gigabyte cards like mine, EVGA, ASUS and others.

 

Just one caveat on EVGA Cards for those reading this that might have one.

If it is a newer ICX card with multiple fans, like my EVGA 1080 Ti FTW3 card, Afterburner will only control one fan, not the other two.

 

I have used Afterburner for many years and was the program of choice for me, but after getting this card I had to move to Precision XOC in order for it to have control over all three of my GPU fans.

 

If not for that I would have stayed with Afterburner. 

Share this post


Link to post
Share on other sites

@ Madmatt - Thanks.  I may have to look into that MSI Afterburned thing (I think I used it for something a couple of PC builds ago, but not lately).  Here is a screenshot of my CPUID screen at OC=4.9:

 

post-12147-0-91980000-1506098444_thumb.jpg

 

This was right after running another VR benchmark test in BoS, and the numbers were hovering right around 4900 consistently.  Before that test, CPUID was bouncing around between 800 and 4900, which made me wonder a bit about what would happen during the benchmark test, but the results came out right where it has been before at 4.9.

 

I had been using AI Suite 3 only to mess with my fan profiles and check my CPU temps, but last night I started messing with a few other settings in there and things got really weird.  I was on the verge of uninstalling it but today everything seems to be going normal again (and fans are behaving themselves) so I think I'll just declare victory and leave everything alone for a while, at least until something else screwy starts happening.

 

@ dburne - I've fired up Precision XOC to look at my GPU temps and stuff, but haven't changed any fan or other settings with it, so my GPU is (should be) running at stock SC2 settings.

Edited by TG-55Panthercules

Share this post


Link to post
Share on other sites

 

@ dburne - I've fired up Precision XOC to look at my GPU temps and stuff, but haven't changed any fan or other settings with it, so my GPU is (should be) running at stock SC2 settings.

 

I set a custom fairly aggressive fan curve in Precision XOC to keep my GPU nice and cool whilst gaming.

 

Mine runs full bore with each mission with GPU temps in the low to mid 50's. Of course having the Rift headset on I don't hear those fans running which helps. Nice thing about keeping it cooler is it not downclocking on me because of temps rising.

Edited by dburne

Share this post


Link to post
Share on other sites

I set a custom fairly aggressive fan curve in Precision XOC to keep my GPU nice and cool whilst gaming.

 

Mine runs full bore with each mission with GPU temps in the low to mid 50's. Of course having the Rift headset on I don't hear those fans running which helps. Nice thing about keeping it cooler is it not downclocking on me because of temps rising.

 

Hmm - it might be nice to be able to see what my GPU temps/fans have been doing during a BoS VR mission - does Precision XOC have a way to record that stuff during a mission so you can go back afterwards and see how things were going during that time?

Share this post


Link to post
Share on other sites

Hmm - it might be nice to be able to see what my GPU temps/fans have been doing during a BoS VR mission - does Precision XOC have a way to record that stuff during a mission so you can go back afterwards and see how things were going during that time?

 

That I am not sure, at least nothing has stuck out at me regarding that.  I myself have it displayed on my secondary monitor and can peek out at it easily enough.

 

Edit: It does have a hardware monitoring tab, I suspect that may do something similar to that, but I am not versed in it at all and looks a little foreign to me...

Edited by dburne

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
Sign in to follow this  

×
×
  • Create New...