Jump to content
chiliwili69

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

Recommended Posts

so i running test in VR mod  and i have strange results )))

ASW OFF Settings High  all other settings maximum and no SS 

 

i7 7700k 4.8ghz + hyperthreading

 

Frames: 4992 - Time: 60000ms - Avg: 83.200 - Min: 46 - Max: 91

 

i7 7700k 5 ghz + hyperthreading

 

Frames: 5027 - Time: 60000ms - Avg: 83.783 - Min: 48 - Max: 91

 

and i was test new beta 2.011 

 

i7 7700k 5 ghz + hyperthreading

 

Frames: 5110 - Time: 60000ms - Avg: 85.167 - Min: 51 - Max: 91

Edited by 1/SG2_Ganeshka

Share this post


Link to post
Share on other sites

so i running test in VR mod  and i have strange results )))

 

 

They did not look strange to me...

 

Have you tried running this with HT disabled? Would be interesting to compare.

Share this post


Link to post
Share on other sites

They did not look strange to me...

 

Have you tried running this with HT disabled? Would be interesting to compare.

with 5.1ghz and HT off, I have worst results. Middle fps was 73...in first test with my old 6700k I was OC my 1070, at this moment I do not do this.Maybe whis higher frequensy it can use more gpu... I don't know

Share this post


Link to post
Share on other sites

with 5.1ghz and HT off, I have worst results. Middle fps was 73...in first test with my old 6700k I was OC my 1070, at this moment I do not do this.Maybe whis higher frequensy it can use more gpu... I don't know

 

Interesting, thanks for the info.

Share this post


Link to post
Share on other sites

 

and i was test new beta 2.011 

 

i7 7700k 5 ghz + hyperthreading

 

Frames: 5110 - Time: 60000ms - Avg: 85.167 - Min: 51 - Max: 91

 

Ganeshka, good to hear next patch is a step forward in performance. 

Share this post


Link to post
Share on other sites
So, my conclusion is that ram speed has more of an impact than I thought possible.

 

Many thanks for sharing your tests. The RAM speed was always a subjective matter for this game, but your tests put real numbers on the table.

I have compiled them on a table:

 

I have two comments about this table:

post-18865-0-00717000-1499157276_thumb.png

Marked in orange: The Balapan rig uses the same processors as you (7700K) at slightly higher freq (4.9) and slightly higher RAM speed (3200), but the important difference is that he is using the top graphic card 1080Ti and you are using the 1070. The difference avg fps difference is only 3.5 fps (most likely because higher OC and RAM). So for this Balapan test in just FullHD, a 1080Ti doesn´t add any value over a 1070, since none of them are never at full load (low render work). Probably at 4K, the 1080Ti will put some distance in between. I would like to see your machine with the Rift.

 

Marked in green: a RAM increase in 867 MHz is more os less equivalent to a OC increase in 0.6 GHz!!. This is a good discovery for this game. So now we know RAM speed is also an important factor.

 

The extra fps you get for +0.6 GHZ CPU OC is about 10 fps, so it is a gain 1.6 fps per 0.1 GHz step. Very similar to what I was obtaining with CPU OC (1.5 gain).

 

The extra fps you get for +867 MHz in RAM is about 10 fps, so it is a gain of almost 1 fps per 100 MHz. It is worth to invest in higher speed.

Edited by chiliwili69

Share this post


Link to post
Share on other sites

 

 

o i running test in VR mod  and i have strange results )))

 

Your results are very much aligned to what you should expect according to the table I was showing in my previous post #49.

 

Walterscott and Balapan are using the same CPU than you at 5.2 and 4.9 OC, and with their top graphics cards they obtain around 86.8 avg fps. You with the same CPU but a 1070 card you obtain 83.7 avg fps. Only 3 fps difference!!

 

At what RAM speed are you running your system?

 

After all, the 6700K and the 7700K have a small diference in Passmark performance. Probably the 7700K has more OC capabilities (off course with deliding)

Share this post


Link to post
Share on other sites

Your results are very much aligned to what you should expect according to the table I was showing in my previous post #49.

 

Walterscott and Balapan are using the same CPU than you at 5.2 and 4.9 OC, and with their top graphics cards they obtain around 86.8 avg fps. You with the same CPU but a 1070 card you obtain 83.7 avg fps. Only 3 fps difference!!

 

At what RAM speed are you running your system?

 

After all, the 6700K and the 7700K have a small diference in Passmark performance. Probably the 7700K has more OC capabilities (off course with deliding)

im running at stock 2666 i think 

i need to learn how to oc it ) i can try 

Share this post


Link to post
Share on other sites

according to the a_radek test, if you increase the RAM speed to 3000MHz you will increase your avg by 3 fps. So, same fps than Balapan and Walterscott and just with a 1070.

 

I am also thinking now to upgrade my RAM to a higher speed, my mobo supports speeds until 3300 in DDR3. So, I could gain more than 10 fps in avg!!

Share this post


Link to post
Share on other sites
@Ganeshka. If you bought 2666mhz ram, and haven't changed a thing in bios, chances are your running your 2666mhz ram at default 2133mhz. If that's the case, enabling your memory XMP profile in bios should do the trick. Overclocking ram beyond it's factory specification usually doesn't yield much gain but takes lot's of stability testing time. In CPU-Z under the "memory" tab you can check your current DRAM frequency. As it's more often than not dualchannel, the number it says x2 will be your current ram speed.

 

The only other games I heard of that really take advantage of ram speed are arma3 and fallout 4.

 

Arma ram speed comparison:


 

Fallout 4 ram speed comparison:


 

Large maps I suppose are the common denominator. In most games, that mostly lean on the gpu side of things anyway, it doesn't make much of a difference.

 

@Chiliwili. If you buy faster ram, make sure first your CPU multiplier and the Memory multiplier are independent. If it's baseclock overcklocking your doing your cpu max speed would probably not allow you to reach that 3300 ram speed. Edit: You could try downclocking your current ram to roughly estimate any performance increase faster would give.

Edited by a_radek

Share this post


Link to post
Share on other sites
My vr test:
Oculus version1.16.0.409144, i7 7700k with hyperthreading on at @4.8, ram @3000, 1070gpu and removed the terrain.ini for easier comparison. Still runnning cache at 4200 as I haven't had any go at fine tuning the overcklock. Graphics settings according to balapans post #20 but with mirrors off, landscapefilter off and gamma 0.9. Asw set to off in occulus debug tool and centered VR view once the headset was on the table, and before pressing "P" to start the fraps measurement and unpause the balapan track. Edit: Stuffed a shirt close to the sensor to activate the rift. View does not look centered on the monitor but guess this has to do with viewing only the right eye image there.
 
Frames, Time (ms), Min, Max, Avg
  5228,     60000,  56,  91, 87.133
 
Only ran this once so results could fluctuate a little both up and down. The high Avg number surprised me, not sure why. It is game vesion v2.010 I'm running. Perhaps Occulus drivers have improved lately.
Edited by a_radek

Share this post


Link to post
Share on other sites

i was OC my memory ... i don't  know if i do this correcte , just change value in bios from 2666 to 2933 , and everything work ))) Maybe i do something wrong ... ) I have asus z170i motherboard .

 

new results i7700k + hyperthreading + 2933 memory + ASW OFF + settings High (all maxed out) + no SS

 

Frames: 5064 - Time: 60000ms - Avg: 84.400 - Min: 50 - Max: 91 


45756362b04bt.jpg

Share this post


Link to post
Share on other sites

new results 2.011

 

new results i7700k + hyperthreading + 2933 memory + ASW OFF + settings High (all maxed out) + no SS

 

Frames: 5165 - Time: 60000ms - Avg: 86.083 - Min: 55 - Max: 91

Share this post


Link to post
Share on other sites

 

 

f you buy faster ram, make sure first your CPU multiplier and the Memory multiplier are independent. If it's baseclock overcklocking your doing your cpu max speed would probably not allow you to reach that 3300 ram speed. Edit: You could try downclocking your current ram to roughly estimate any performance increase faster would give

 

I think they are independent, since I have OC the CPU to 4.4 and changed the RAM freq from 1866 to 1066, I had a loss of 7fps on average, so almost 1 fps per 100MHz as per your test.

 

The fastest DDR3 is 3200MHz, but it is sold in very few places and it is very expensive.

I am considering more the 2600MHz which is more affordable.

 

I think I do base clock OC, I use the MSI Command Center to apply a ratio (currently 44 to obtain 4.4 GHz). I don´t know why you said this could limit the RAM speed I could reach.

Share this post


Link to post
Share on other sites

I don´t know why you said this could limit the RAM speed I could reach.

Chiliwili, while I sometimes sound as if know something about anything, that's rarely the case. Someone more knowledgeable correct me if the following is wrong.

 

Overclocking your cpu can be done with base clock or cpu multiplier. CPU multiplier affects CPU clockspeed only while a baseclock bump will increase not only CPU speed but also memory and cache speed. I was unnecessarily worried you might find a good clock speed for your CPU without ever reaching the full potential of that 3200 ram, if those two were connected for you. Clearly they are not and I suspect 44 is your CPU multiplier/ratio.

 

I agree the 2600mhz memory seems the more sensible choice and easier to sell once you upgrade it all.

 

Been wanting to compare the effects of cache/uncore frequency on il2 box but tend to spend my free time flying (getting shot by competent pilots) instead. Sooner or later..

 

As a side note: My VR benchmark above was with hud off. Thought this was part of the standard setting for the benchmarks.

Share this post


Link to post
Share on other sites

a_radek, thanks for the explanation. Since I checked that my there were an influence on the RAM speed, today I have ordered DRR3 16Gb of 2933MHz, it was a good deal (216$+shipping) I saw in pcpartpicker.com

Hopefully I will be able to sell my 1866 memory.

I will report the the test with the new memory to check if the fps increase are as good as in your case.

Share this post


Link to post
Share on other sites

 

 

As a side note: My VR benchmark above was with hud off. Thought this was part of the standard setting for the benchmarks.

 

I have done some testing in VR with the Balapan track, which by default is using icons for the planes (so HUD is ON but without maps).

Before pressing "P" (to unpause and activate Fraps record) I press H to disable the HUD, so we can record the track without the influence of icons which I normally don´t use (only some seconds to help spotting).

I found that only the icons on planes the fps decrease about 6fps the whole average.

 

I also notice that the Balapan track in VR is more subjective than in a monitor since the initial position of the Rift can cover more or less view of the cabin (the reset view only affect the positional but not the rotational). So when the fix the Rift to the table or to the seat to make the record the pitch angle of the Rift affects how much cabin you see. This can affect in about +-3fps on average.

 

Another weird thing to comment about all these tests is that in VR, you can run twice the same scenario with identical settings but the results can differ in +-3fps. So that´s why I run three or more tests to have an average. But I don´t know what could influence the PC since there are not other "important" applications running at the same time.

Share this post


Link to post
Share on other sites
Frames, Time (ms), Min, Max, Avg   5228,     60000,  56,  91, 87.133

 

You should be aware that with a 1070 you are having the same performance than Balapan who uses a 1080Ti. Look at the orange numbers:

post-18865-0-11264600-1500331021_thumb.png

 

Of course, we have not entered with the supersampling factor which is something that the Balapan test doesn´t take into account. This is something that people with a 1080ti could try to run the Balapan track at higher SS or PD.

 

Once I have my new memory and tweak around the max stable OC I can run my CPU, I will try with different SS values to analyze their impact on performance.

Edited by chiliwili69

Share this post


Link to post
Share on other sites

I run with icons on in Single Player, whilst the HUD is hidden. I have checked toggling them on and off throughout the flight and can not see any real performance hit from having them on. 

 

Of course that is just my observations from what I see. HUD though is another story.

Share this post


Link to post
Share on other sites

 

 

can not see any real performance hit from having them on
 

 

I didn´t try that while flying, just only run the test in VR. But I suppose it is something difficult to measure since the fps counter is up and down depending of the view of the planes around. The Fraps results with the Balapan track are:

 

ICONS ON
Frames: 3599 - Time: 60000ms - Avg: 59.983 - Min: 43 - Max: 91
Frames: 3801 - Time: 60000ms - Avg: 63.350 - Min: 43 - Max: 91
Frames: 3714 - Time: 60000ms - Avg: 61.900 - Min: 43 - Max: 91
 
ICONS OFF
Frames: 4163 - Time: 60000ms - Avg: 69.383 - Min: 44 - Max: 91
Frames: 4115 - Time: 60000ms - Avg: 68.583 - Min: 43 - Max: 91
Frames: 4050 - Time: 60000ms - Avg: 67.500 - Min: 44 - Max: 91
 
 
Pressing "H" the icons of the planes around are hidden and also the progress bar of the recorded flight.

Share this post


Link to post
Share on other sites

chiliwili,

Your right, didn't notice before but center VR-view will only center position not rotation. I guess best we can do is try to place the rift so that the reticle by scientific eyeballing is roughly in the middle of the screen and then center VR view. As position will be the same for us all and angle at least somewhat equal, it should be close.

About the icons though. Perhaps the balapan track has icons on by default, but as I have HUD off by default, I don't get neither hud nor icons unless I enable those manually before unpausing. So naturally It didn't cross my mind I was overriding the track default settings.

I think this thread is a very important one. Helped me decide what hardware to get and if I even should. I only wish more people with different setups would take some time to benchmark and post their results. Would not surprise me if even the devs would benefit from such info. However looking back at this thread it's quite messy, instructions for benchmarking split into different posts, some using their own settings and so on. You've spent a lot of time on this already Chiliwili. So won't ask you to, but as you have the first post and can edit it, perhaps you would allow me to gather all info in to step by step instructions on how to do a benchmark and we could then have that info in the first post. I'll try to find the time for such a task.

A suggestion of mine is to also at use some AA if not supersampling for the graphics settings. As I believe that would be more realistic for how people set things up and also the gpu would play a bigger part in these tests.

Share this post


Link to post
Share on other sites

 

 

perhaps you would allow me to gather all info in to step by step instructions on how to do a benchmark and we could then have that info in the first post. I'll try to find the time for such a task

 

Yes a_radek, the info in this thread is a bit messy and the steps to make the tests are not clear enough. I will be very please to put in the first post whatever procedure you suggest. We should also enable a way to report the results in an online spreadsheet (Google spreadsheet or MS Excel online), so the reported test can be used by anyone, of course, including developers.

 

There are many things that affects performance in VR, I would suggest that the table contains:

 

Usermane

Date

IL2 version

CPU

CPU GHz (no OC)

CPU GHz (of the test)

Passmark number

RAM type (DDR3, 4)

RAM size

RAM speed

GPU

VR device

PD/SS value

Test results in VR (frames, Min, Max, Avg)

Test results in Monitor (frames, Min, Max, Avg)

 

You can send me a private mail and we can define/refine the procedure.

Share this post


Link to post
Share on other sites

There are also a few other title's where the 1080 is tied in performance or even better than a 1080ti, which means only one thing, or a few..

Edited by icecream

Share this post


Link to post
Share on other sites

I've just got the rift and had trouble initially so found this thread. I've had a brief test using the track and get results as follows.

 

 Min  Max  Avg  

51 91 65.382 SS1 Low Gfx, HUD off AAx4

38 91 72.842 SS1 High Gfx, HUD off AAx0

44 91 59.877 SS2.0 Low Gfx, HUD off AAx4

 

I have I5 4670K not overclocked, GTX1070 and 12GB of DDR3-1600 RAM

 

I've found that anything over low settings has a big impact on performance but offers little to no improvement in quality. In contrast SuperSampling and to a much lesser extent AA vastly improves clarity with less impact on FPS.

 

Even though the AVERAGE framerate may be ok on Normal/High/Ultra I get a significant number of dropped Frames and Compositor dropped frames which for me makes the game unplayable.

 

CPU usage remains constant regardless of settings, about 40-60%.

 

Am getting a Ryzen 1700 build shortly so will post new results once that's setup. I don't expect any improvement as clock speeds will be similar but will be interesting to see if RAM speed (2933MHz) will help.

Edited by AceVenturi

Share this post


Link to post
Share on other sites

I've just got the rift and had trouble initially so found this thread. I've had a brief test using the track and get results as follows.

 

Min Max Avg

51 91 65.382 SS1 Low Gfx, HUD off AAx4

38 91 72.842 SS1 High Gfx, HUD off AAx0

44 91 59.877 SS2.0 Low Gfx, HUD off AAx4

 

I have I5 4670K not overclocked, GTX1070 and 12GB of DDR3-1600 RAM

 

I've found that anything over low settings has a big impact on performance but offers little to no improvement in quality. In contrast SuperSampling and to a much lesser extent AA vastly improves clarity with less impact on FPS.

 

Even though the AVERAGE framerate may be ok on Normal/High/Ultra I get a significant number of dropped Frames and Compositor dropped frames which for me makes the game unplayable.

 

CPU usage remains constant regardless of settings, about 40-60%.

 

Am getting a Ryzen 1700 build shortly so will post new results once that's setup. I don't expect any improvement as clock speeds will be similar but will be interesting to see if RAM speed (2933MHz) will help.

First ss benchmark comparison I see, and surprised it had such a low impact on avg framerate despite running at 2.0.

 

I also run a 1070, game settings at "high", only 2x aa and 1.3 ss set in steam. Not sure how the steam-VR ss value compares to setting the same value in oculus tray tool. But any higher than 1.3 and I feel a substantial lag/dropped frames when things start heating up around me online.

 

Looking forward to your ryzen benchmarks.

Share this post


Link to post
Share on other sites

Well, after the first failed trial (the first amazon-purchased 2933 memory cards were faulty, I returned and a new set was sent at no cost), my new DDR3 2933MHz is up and running. 

Let´s see what is the improvements in fps in monitor and VR.

As usual, I run three balapan flights per test and then average them. All test were running with 4790K at 4.6 GHz.

 

Results in Monitor:

 

RAM 1866MHz: Frames: 8587, Min:110, Max: 189 Avg: 143.1

RAM 2933MHz: Frames: 8888, Min:113, Max: 195 Avg:148.1

 

5 fps gain for 1067 MHz increase (I really don´t care about gain in monitor, just for reporting)

 

Results in VR:

 

RAM 1866MHz: Frames: 4417, Min:44, Max: 90 Avg: 73.6

RAM 2933MHz: Frames: 4917, Min:45, Max: 91 Avg: 81.9

 

8.3 fps gain for 1067 MHz increase, which is about 0.8 fps per 100MHz. That´s a bit less that your gain (it was about 1 fps per 100MHz), but overall very good.
 
I am happy with my memory upgrade, it has been a worth investment (let´s see if I sell my old memory) and it is clearly a factor to take into account when people are upgrading their rigs.
 
Conclusion, memory matters in VR.
 
Here the fps graph for both RAMs in VR:
post-18865-0-13051600-1501589687_thumb.png
Edited by chiliwili69

Share this post


Link to post
Share on other sites
Your right, didn't notice before but center VR-view will only center position not rotation. I guess best we can do is try to place the rift so that the reticle by scientific eyeballing is roughly in the middle of the screen and then center VR view. As position will be the same for us all and angle at least somewhat equal, it should be close.

 

Yes, having the reticle in the middle could be a good compromise. I have taken an screenshoot of the monitor view to let you know how we could find the same pitch rift angle to our tests. There are two red lights indicators (landing gear) which are almost at the border of the view. Then the reticle is almost at the border:

post-18865-0-71949100-1501597787_thumb.png

Edited by chiliwili69

Share this post


Link to post
Share on other sites
38 91 72.842 SS1 High Gfx, HUD off AAx0

 

 

I have been taken a closer look of your test apart from the SS factor. 

Your CPU is a I5 4670K not OC, that is you run at 3.4 GHz. That CPU has a passmark index of 7619, so it is a moderate CPU not OC'ed with 1600 MHz speed RAM.

 

What I still don't understand is how you can reach 72.84 avg fps in the Balapan test. According to other test your CPU-RAM-GPU should obtain something in the range of 50 fps avg. Look the table (in yellow your test):

post-18865-0-86712900-1501599868_thumb.png

 

So, I really interested to know how you obtain a better than expected performance with no SS.

 

Could you please try to run again the test being sure the setting are at HIGH, AAx0, HUD off, (as Balapan said)? or try just the test in monitor with Balapan monitor settings?

Could you also run the test with the reticle in the center (or red lights just at the bottom of monitor image)?

 

If you really have 72.84, then there is something else I am missing

Edited by chiliwili69

Share this post


Link to post
Share on other sites
Am getting a Ryzen 1700 build shortly so will post new results once that's setup. I don't expect any improvement as clock speeds will be similar but will be interesting to see if RAM speed (2933MHz) will help.

 

Ryzen 1700 has passmark of 14679 so you might get a surprise. If RAM speed is as sensible as it is in the Intel CPUs, it might be better to have a memory as fast as possible (not the fastest but price wise) 

Edited by chiliwili69

Share this post


Link to post
Share on other sites

Well, after upgrading my CPU, my RAM and OC´ed,  now it is the time to experiment with supersampling effect in VR performance.

I have done three SS tests (3 runs per test) with 4.6 GHz and all the Balapan settings for VR test and ASW Off by pressing Crtl+NumPad1.

I have changed the SS number in SteamVR  to 1.0, 1.5 and 2.0. (I don´t know how these numbers relate to PD of Oculus Tray tool, but I will learn that soon)

post-18865-0-31893700-1501604887_thumb.png

 

The Avg number are: (Edit: this is done with the old SteamVR where SS=OTT SS=Oculus PD. In new SteamVR, from Jul-2017 SS=(PD)^2)

SS 1.0: Frames 4914, Avg 81.9

SS 1.5: Frames 4243, Avg 70.7

SS 2.0: Frames 2872, Avg 47.9

 

The graph is:

post-18865-0-13560100-1501604774_thumb.png

 

So, SS affects severely the fps. I don´t know now if the bottleneck is the 1070 or other things.

 

I will try with other SS values in the 1.0-1.5 range and will experiment lower settings to see how much graphics quality I am loosing.

Edited by chiliwili69

Share this post


Link to post
Share on other sites

I believe 1.5SS in Steam VR equates to around 1.1-1.2 Pixel Density in the Rift.

 

Will be interesting to see what your further testing shows.

Share this post


Link to post
Share on other sites

First ss benchmark comparison I see, and surprised it had such a low impact on avg framerate despite running at 2.0.

 

I also run a 1070, game settings at "high", only 2x aa and 1.3 ss set in steam. Not sure how the steam-VR ss value compares to setting the same value in oculus tray tool. But any higher than 1.3 and I feel a substantial lag/dropped frames when things start heating up around me online.

 

Looking forward to your ryzen benchmarks.

I have been taken a closer look of your test apart from the SS factor. 

Your CPU is a I5 4670K not OC, that is you run at 3.4 GHz. That CPU has a passmark index of 7619, so it is a moderate CPU not OC'ed with 1600 MHz speed RAM.

 

What I still don't understand is how you can reach 72.84 avg fps in the Balapan test. According to other test your CPU-RAM-GPU should obtain something in the range of 50 fps avg. Look the table (in yellow your test):

attachicon.gif Balapan table-5.png

 

So, I really interested to know how you obtain a better than expected performance with no SS.

 

Could you please try to run again the test being sure the setting are at HIGH, AAx0, HUD off, (as Balapan said)? or try just the test in monitor with Balapan monitor settings?

Could you also run the test with the reticle in the center (or red lights just at the bottom of monitor image)?

 

If you really have 72.84, then there is something else I am missing

Thanks.

 

I must have grabbed the wrong record or something. I've run a few more tests. 

 

Maybe I'm not doing it right? The other entries seem to have a lot of frames or are they recording from game startup?. I only run FRAPS once the track has loaded and I've turned off HUD and Pause. I stop it as soon as replay finished message comes up.

 

Here's a HIGH, AAx0, HUD off, no SS. Screen res at 800x600

 

Frames 519 

Time (ms) 8562 

Min 33 

Max 91 

Avg 60.617

 

AVG 60 which seems more reasonable.

 

Here's same test with just screen at 1080p which I presume is the resolution the others are running at?

 

Frames 491

Time (ms) 6922 

Min 43 

Max 128 

Avg  70.933

 

Screen test seems to have some low frame rates somehow even though when I watched the FPS counter it never dropped below 90 so not sure how FRAPS reads Min FPS as 43! I've run it multiple times and same each time :(.

 

Should also say that my CPU almost never gets down to 3.4GHz unless it's idle. Normally sits about 3.7 and obviously boosts to 3.8 in game.

 

Can you explain exactly how the test is being performed I don't think I'm doing it right?

Edited by AceVenturi

Share this post


Link to post
Share on other sites

Aceventuri,

 

I think we all ran this test a little differently as the "how to" information is spread out in to different posts throughout this thread. Unless someone beats me to it I'll try to gather all info for benchmarking in to a single step by step list, tonight. But if you set fraps to start with the "p" key, and run for 60 seconds, You'll be unpausing the track and starting the fraps measurement at the same time with the same key.

 

Chiliwili,

 

Happy the memory upgrade worked out for you. And made me consider upgrading mine. 7-9 more frames would be very welcome. Your benchmarks are plenty more Interesting than the higher end rigs in that list and will help those considering VR but unwilling to invest in a new rig.

 

One last thing to try would be to increase cpu cache frequency. Called uncore in my asus bios. As I understand it, this like all other significant overclocking needs a voltage boost. But it uses the same increased voltage you already gave the cpu. Meaning It's a "free" upgrade. Usually cache frequency gives no significant fps increase in games, but usually neither should ram frequency.

Share this post


Link to post
Share on other sites

 

 

I don´t know how these numbers relate to PD of Oculus Tray tool, but I will learn that soon

 

Well, now I know. After some hours googling and some testing this is what I can say:

 

https://forum.il2sturmovik.com/topic/30436-pixel-density-supersampling-steamvr-oculustraytool-and-oculu/

 

The Oculus Mirror tool is really helpful to determine the actual render size used by the Rift. 

Share this post


Link to post
Share on other sites
System:
CPU: R5 1600X (stock clock for these tests)
GPU: RX 480 8GB @ 1380 MHz core, 2025 memory
RAM: 16GB DDR4 @ 2667 MHz
OS: Win 10
 
Balapan settings on 1440p monitor, HUD on
 
post-13947-0-55312700-1499988105_thumb.jpg
post-13947-0-36339500-1499987130_thumb.jpg
 
1440p monitor result:
AVG: 108
Min: 73
Max: 148
 
On Oculus Rift at stock settings (ASW on, Oculus home running in background (can't get rid of it), PD at default (1.0 in steamVR) and the hud on:
AVG FPS: 45
Min FPS: 45
Max FPS: 45

 

HUD OFF is the same.

 

Testing method: Reset track to very beginning to avoid the 3s offset that naturally occurs on first load, stare straight ahead at all times (TIR disabled for monitor tests), start recording and track at same time. Run 3 times, average results. I was unable to disable ASW for some reason, I had it off earlier but after rebooting I can't seem to get rid of it. I found it unpleasant to play with it off though, even if the frame rate tended to hover around 60.

 

DCS turned out to be completely unplayable in many areas of both 1.5 and NTTR (under 20 fps), so I grabbed a GTX 1080 TI a about an hour ago. Will install and add results for that once I have them, probably tomorrow.

  • Upvote 1

Share this post


Link to post
Share on other sites

Hey,

 

Thanks to a_radek we have described a new procedure for the test (using typical SS values) so people can follow easily the instructions.

 

I have conducted already my test:

 

Frames: 4867 - Time: 60000ms - Avg: 81.117 - Min: 45 - Max: 91

 

The rig spec at signature.

 

Editted 5th-Aug-17: I just followed procedure except for SSAO. I run it with SSAO uncheked

Edited by chiliwili69

Share this post


Link to post
Share on other sites

 

System:
CPU: R5 1600X (stock clock for these tests)
GPU: RX 480 8GB @ 1380 MHz core, 2025 memory
RAM: 16GB DDR4 @ 2667 MHz
OS: Win 10
 
Balapan settings on 1440p monitor, HUD on
 
 
 
 
1440p monitor result:
AVG: 108
Min: 73
Max: 148
 
On Oculus Rift at stock settings (ASW on, Oculus home running in background (can't get rid of it), PD at default (1.0 in steamVR) and the hud on:
AVG FPS: 45
Min FPS: 45
Max FPS: 45

 

HUD OFF is the same.

 

Testing method: Reset track to very beginning to avoid the 3s offset that naturally occurs on first load, stare straight ahead at all times (TIR disabled for monitor tests), start recording and track at same time. Run 3 times, average results. I was unable to disable ASW for some reason, I had it off earlier but after rebooting I can't seem to get rid of it. I found it unpleasant to play with it off though, even if the frame rate tended to hover around 60.

 

DCS turned out to be completely unplayable in many areas of both 1.5 and NTTR (under 20 fps), so I grabbed a GTX 1080 TI a about an hour ago. Will install and add results for that once I have them, probably tomorrow.

 

 

Thanks for this, Beasty - really appreciate it!

Share this post


Link to post
Share on other sites

GTX 1080 TI result, same method as above post, still unable to kill off ASW:

 

HUD on:

Max:45

Min: 45

AVG: 45

 

HUD off, same result. Also, I have some strange flickering around the prop now. Need to figure out a way to solve that as it makes the game unplayable. This is also a very unexpected result, I really was expecting a locked 90 fps going from an RX 480 to a GTX 1080 TI. It's probably running 89 fps or some nonsense, just below the cutoff for ASW while the RX 480 was just above the 45 fps mark. In any case, it made a bit more of a difference in DCS, so that is the important bit. It went from 20-ish fps to 30-ish fps around Batumi/Kobeleti (seems to be CPU bottleneck, too many trees and no clip mapping) at the exact same settings, other areas run much better. NTTR runs locked 45 fps all the time over the strip at virtually any graphics setting, probably cpu bottleneck at 89 fps :P. Away from Vegas or if getting off the deck, it goes up to 90.

 

I would like to test with ASW off, if someone can say how to do that, I'll redo the tests with the GTX 1080 TI. The previous way I did it was by completely blocking oculus home from running, but given that I do like some of the games there and unblocking it was far more challenging than blocking it, I'd rather not go that route again.

Share this post


Link to post
Share on other sites

@Beastybaiter, to switch off ASW:
 

"C:\Program Files\Oculus\Support\oculus-diagnostics\OculusDebugTool.exe"

 

Run that, switch "Asynchronous spacewarp" to disabled. Be aware though, whenever occulus auto updates to a newer version, your settings will be reset. So make a desktop shortcut for this tool or download and install "occulus traytool". A third party app. Pressing ctrl + Numpad 1 is a third way to do it but for me for some reason that does not work.

(Also make sure Pixels Per Display Pixel Override is at 0)

 

 

debugtool.jpg
 

You are the first one with AMD reporting performance. I'm sure whatever ways you come up with to improve things will greatly help other AMD users.

 

Following not directed at you in particular:

 

Chiliwili has updated the first post in this thread with full instructions on how to benchmark. And also what settings to use, in the hopes of us all running everything equally.

 

Graphics settings are higher now than before, and so the Gpu will play a bigger role.

Edited by a_radek

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...