Jump to content

SYN_Vander BENCHMARK v6 to measure IL-2 performance in monitor & VR


chiliwili69
 Share

Recommended Posts

52 minutes ago, chiliwili69 said:

 

we only have 5 memory latencies data, but it correlate well with fps:

Untitled.png.f62b591cc13a9b0212116dd12cd41b79.png

 

But ATW is not old tech. It is something which is always enabled in the Oculus devices. The equivalent in SteamVR was Asynchronous Reprojection, and it is still used always in SteamVR.

 

 

Incorrect.

ATW preceded ASW for Oculus, and was replaced by ASW. Or you can maybe better say ASW swallowed up ATW.

Same way Steam Motion Smoothing replaced the Steam Legacy Reprojection.

 

https://developer.oculus.com/blog/asynchronous-spacewarp/

 

You won't see Oculus talking about ATW with their devices, they only refer to ASW currently. Specifically ASW2 nowadays.

 

Quote

 


Asynchronous Timewarp was introduced at Rift launch and provided an automatic, asynchronous version of time warp. It works by intervening every frame to time warp the last submitted eye buffers, even if the application is taking too long rendering the next frame. More about ATW can be found in these two blog posts:

Asynchronous Timewarp on Oculus Rift

Asynchronous Timewarp Examined

ASW builds on top of the virtual reality smoothing experience of ATW. ATW ensures that the experience tracks the user's head rotation. This means an image is always displayed in the correct location within the headset. Without ATW, when a VR application misses a frame, the whole world drags—much like slow-motion video playback. Encountering this while in VR is extremely jarring and generally breaks presence. ASW goes beyond this and tracks animation and movement within the scene to smooth over the whole experience.
 

 

Edited by dburne
Link to comment
Share on other sites

1 hour ago, chiliwili69 said:

That´s strange. My fps varies from 45 to 90 when running the benchmark with Legacy reprojection disabled.

 

I have been testing some QMB 8vs8 with not clouds and with Legacy reprojection enabled and it work quite OK. But I still really don´t know what legacy reprojection really is.

 

You get now a more consistent results. What did you do with PBO?

 

 

I tried multiple settings. In BIOS:  PBO off, On, On+200, On+100, auto. Then I reinstalled Ryzen Master and I think I'll keep it.  I clicked on PBO, 500 PPT, 200 TDC and 220EDC and it looks like it likes it here. I probably won't mess with it until the new version of Ryzen Master comes in February. 

    I watched my maximum boost in HWinfo and with this settings looks like i have seven cores that boost over 5GHz.:dance: That's pretty impressive for a CPU that advertises maximum single core boost 4.9GHz. 😁

 

774663811_singlecore.thumb.png.f6b565335f4d0383ceadb3d98a2d86bb.png

 

I also tinkered a bit with the memory and it now takes the "fast preset from Dram calculator and my latency dropped to 55.7ns. It didn't take all the values exactly like in the calculator, but I got no errors after two hours of stress test and that's good enough for now. There's new bios coming soon for Gigabyte motherboards, new Ryzen Master and a new clock tuner for ryzen, so I'll wait for those  before i tinker with it again.

 

 

cachemem3800fast.png.65d3ba9643d3f5824147543c3e8ee2cf.png

 

And here's a shot from Zen timings. maybe it helps others:

ZenTimings_Screenshot.png.b8541c26e7ea53f3f0bf197951ccb0e9.png

Edited by Jaws2002
  • Thanks 1
Link to comment
Share on other sites

chiliwili69
38 minutes ago, dburne said:

Incorrect.

ATW preceded ASW for Oculus, and was replaced by ASW. Or you can maybe better say ASW swallowed up ATW.

 

Wrong Don.

Read this:

Because of this, ASW doesn’t replace ATW. ATW is always active, ASW kicks in when needed.

 

from here: https://uploadvr.com/reprojection-explained/

  • Haha 1
Link to comment
Share on other sites

14 hours ago, Charlo-VR said:

"Open the settings file at :

<path-to-steam>\SteamLibrary\steamapps\common\MixedRealityVRDriver\resources\settings\default.vrsettings

Comment out these lines (add the // in front as here):
 // "motionReprojectionMode": "none",
 // "motionReprojectionIndicatorEnabled": false,"

 

I *think* that ensures I stay out of Asynchronous Reprojection mode, but all I know is that fpsVR tells me I remain in 80-90 fps in MP with rare stuttering and reasonably easy aircraft IDing with no ghosting. In MP I'd rather have occasionally minor stuttering close to the ground on full MP servers than the 45fps that doesn't allow me to spot and ID as easily.

 

I ran into this last night and can confirm it's good advice.

 

If you have any "motionReprojectionMode" setting in your default.vrsettings file then WMR ignores the SteamVR setting for motion reprojection, and just does what you wrote in the file. So if you want SteamVR to control motion reprojection on a WMR device like the G2, you need to comment this line out of the file. I did this last night and now I have a proper variable framerate from mostly-90 to mid-70s, rather than an instant drop to 45.

  • Thanks 1
Link to comment
Share on other sites

chiliwili69
On 1/15/2021 at 7:11 AM, Brandy_ said:

This article is older, but explains it:

https://uploadvr.com/reprojection-explained/

 

Especially this table is interesting:

ReprojectionExplainedDiagram.png

 

That link explain very well all the repro words used along the time. It could be seen as messy but it is well explained.

 

The current "Legacy Reprojection" option in the per Application Video settings of SteamVR is just the Interleaved Reprojection which was superseded by the cureent Motion Smoothing of SteamVR. Therefore the Legacy Reprojection is something that should be normally off. And in case someone has low fps then the recommended solution is Motion Smoothing.

 

https://www.reddit.com/r/Vive/comments/gnjnlh/legacy_reprojection_mode/

 

I don´t know why they kept the legacy Reprojection option, perhaps for older apps or who knows.

 

The fact is that running the VRTEst1 of SYN_Vander I get about 77fps avg with Legacy-Repro OFF and 88 with legacy-Repro ON. This is the fps in each moment of the track.

Untitled.png.404e720a51a9328e6a73cfff9805a5b3.png

 

In theory, the green line (Legacy Repro ON) should be just 90 or 45, but nothing in between. In theory, the Legacy Repro is switching to 45 after being under 90 for few seconds. But here it doesn´t go to after being several seconds below 90.

I also don´t know why the fps are higher when it is swichted ON.

 

I have tried the Legacy repro ON in QMB 8vs8 with 130%SS with no clouds at 80HZ mode (and High settings and almost everything maxed out), and it run OK. Basically because I am always at 80.

But if I run the same QMB 8vs8 but with cloudy sky, then the legacy repro enter in action and produce some sttuters which I dont like it.

 

Independently if the Legacy Repro improves (or not) the experience I think that the SYN_vander benchmark should be run with Legacy Repro OFF. Since the Legacy Repro is a kind of rotational motion smoothing, and we want to measure performance just without any kind of ASW or Moting Smootthing.

 

I will update intructions in the OP.

 

@Brandy_ I will not take into account your VR TEsts with the Legacy Repro ON. You can retest them with your 3080 and Legacy Repro OFF.

For your normal setting your are of course free to choose legacy ON or OFF.  😉

Many thanks for giving us some light about this high fps you had.

 

 

Link to comment
Share on other sites

5 minutes ago, chiliwili69 said:

In theory, the green line (Legacy Repro ON) should be just 90 or 45, but nothing in between. In theory, the Legacy Repro is switching to 45 after being under 90 for few seconds. But here it doesn´t go to after being several seconds below 90.

I also don´t know why the fps are higher when it is swichted ON.

 

A while ago I was reading some documentation on Valve's variable resolution for Half Life Alyx (and other games) maybe from a games developer conference or something. They have an algorithm that varies the sample rate (= GPU load) for each frame by looking at the frame time of the current frame. If they think they have enough headroom, they increase the supersample. If they missed the frame time target, they reduce supersample. There are some numbers and ratios in the algorithm, like how aggressively they increase/decrease GPU load to maintain target frame time, and how much headroom they want to keep under ideal circumstances.

 

I expect that all the reprojection algorithms have similar numbers and ratios, that change how fast they detect a "we should reproject" situation vs "we should render native frames". Maybe these numbers change the overall behavior of the algorithm.

 

An aside: I think IL2 would be absolutely amazing with an Alyx-style variable sample rate. Alyx mostly manages to keep your GPU at ~90% load while cranking the visuals, and when it gets to a difficult scene you keep max-FPS but lose a bit of fine detail. Since I usually lose framerate in a dogfight, I wouldn't mind losing some fine detail at those times, and then getting the detail back when I'm high in the sky and searching for enemies below.

  • Upvote 1
Link to comment
Share on other sites

31 minutes ago, Alonzo said:

 

A while ago I was reading some documentation on Valve's variable resolution for Half Life Alyx (and other games) maybe from a games developer conference or something. They have an algorithm that varies the sample rate (= GPU load) for each frame by looking at the frame time of the current frame. If they think they have enough headroom, they increase the supersample. If they missed the frame time target, they reduce supersample. There are some numbers and ratios in the algorithm, like how aggressively they increase/decrease GPU load to maintain target frame time, and how much headroom they want to keep under ideal circumstances.

 

I expect that all the reprojection algorithms have similar numbers and ratios, that change how fast they detect a "we should reproject" situation vs "we should render native frames". Maybe these numbers change the overall behavior of the algorithm.

 

An aside: I think IL2 would be absolutely amazing with an Alyx-style variable sample rate. Alyx mostly manages to keep your GPU at ~90% load while cranking the visuals, and when it gets to a difficult scene you keep max-FPS but lose a bit of fine detail. Since I usually lose framerate in a dogfight, I wouldn't mind losing some fine detail at those times, and then getting the detail back when I'm high in the sky and searching for enemies below.

 

Yes Alyx does a great job in this.

 

 

1 hour ago, Alonzo said:

 

I ran into this last night and can confirm it's good advice.

 

If you have any "motionReprojectionMode" setting in your default.vrsettings file then WMR ignores the SteamVR setting for motion reprojection, and just does what you wrote in the file. So if you want SteamVR to control motion reprojection on a WMR device like the G2, you need to comment this line out of the file. I did this last night and now I have a proper variable framerate from mostly-90 to mid-70s, rather than an instant drop to 45.

 

On my G2 Motion Smoothing in Steam VR appears to work appropriately.

I set it to on and force it to 45 fps, and it stays at 45 and is very smooth. 

Edited by dburne
Link to comment
Share on other sites

Brandy_Lx
2 hours ago, chiliwili69 said:

Brandy_ I will not take into account your VR TEsts with the Legacy Repro ON. You can retest them with your 3080 and Legacy Repro OFF.

For your normal setting your are of course free to choose legacy ON or OFF.  😉

Many thanks for giving us some light about this high fps you had.

 

No problem, I'll post them as soon as I get time to go through a benchmarking session again.

Glad to help, I prefer to be transparent, as this (+10fps avg) looked exceptional after all.

 

 

Link to comment
Share on other sites

I tried an early F-32a beta Bios and I was able to get the infinity fabric to 2000Mhz without problems.:good:

 

Now, if I knew how to get my memory to 4000Mhz...... :scratch_one-s_head:

 

2000flk.png.8a976487120d652aeb2943f5c75a8f5f.png

 

 

 

 

Link to comment
Share on other sites

chiliwili69

I have just got a 3080, replacing my old 1080Ti.

 

The new tests:

 

with new Gigabyte Gaming OC 3080 (out of box )
CPU Test
Frames: 6967 - Time: 60000ms - Avg: 116.117 - Min: 101 - Max: 163

 

GPU Test
Frames: 8414 - Time: 60000ms - Avg: 140.233 - Min: 109 - Max: 170
Frames: 8474 - Time: 60000ms - Avg: 141.233 - Min: 113 - Max: 165
Frames: 8459 - Time: 60000ms - Avg: 140.983 - Min: 111 - Max: 164

 

VRTest1
Frames: 4580 - Time: 60000ms - Avg: 76.333 - Min: 44 - Max: 91
Frames: 4712 - Time: 60000ms - Avg: 78.533 - Min: 47 - Max: 91
Frames: 4647 - Time: 60000ms - Avg: 77.450 - Min: 43 - Max: 91

 

VRTest2
Frames: 4818 - Time: 60000ms - Avg: 80.300 - Min: 50 - Max: 91
Frames: 4341 - Time: 60000ms - Avg: 72.350 - Min: 44 - Max: 91
Frames: 4458 - Time: 60000ms - Avg: 74.300 - Min: 44 - Max: 91
 

Summary:

On CPU Test, same results. As expected.

On GPU Test in 4K I went from 95 to 140, so +45fps in 4K!

On VRTest1, same results. As expected. Here bottleneck is CPU in both.

On VRTest2 with 216%SS (19.5 Million pixels) I went from 45 to 75, so this is +30fps in VR. Probably here I was also a bit bottlenecked also by CPU.

 

The best of all is that the GPU was much more silent than my previous 1080Ti. Specially in the 4K and VRTest2.

 

I will try now how I play with fpsVR with the new card with the Index in QMB 8vs8 in Kuban with clouds. And will see if I increase my usual 130% SS in the Index.

 

  • Like 1
  • Upvote 2
Link to comment
Share on other sites

chiliwili69

Just a Kuban novorosyk QMB 8vs8 with High Settings, and High shadows, high clouds and 150Km horizon, and FXAAx2 with 130% SS with Index at 80Hz.

 

I am very pleased with the performance of the 3080. Very smooth with clouds and near cities. Now I can really put clouds in my flights without being contrained.  There is room for more since temperature is quite good (since no OC, just out of box.).

 

This is fpsVR during no clouds mission low flight over novorosyk.

3080.png.abc8585db414e683c0cf56c04021884d.png

 

Here the GPU load and other variables with the sme QMB with clouds and no clouds.

 

3080_2.thumb.png.89b16d08dd4e493cbef029ac8de125f5.png

 

 

 

Edited by chiliwili69
  • Like 3
Link to comment
Share on other sites

Congrats :salute:  

 

Looks great!  This new cards are powerful. I need to replace mine too. I just got hung because of the 10GB of ram. 

 

Congrats!

Edited by Jaws2002
Link to comment
Share on other sites

chiliwili69
9 hours ago, TheSNAFU said:

Where on earth did you get it?

 

From time to time I was just checking this stock page https://geizhals.eu/, and also sometimes ebay, but the prices were a bit crazy.

 

I was not really in a hurry, I knew I had to upgrade my GPU this year (I passed the Nvidia 2000 series). Then,  I just saw a guy (in https://www.milanuncios.com/) from my own city who was selling  a brand new 3080 for 999€, with bill bought in Dec-2020 form a known online shop at the same price. I just went to meet him and everything was OK.

 

My other option was to buy a 6800XT or 6900XT from PCcomponentes.

Link to comment
Share on other sites

I finally got around to doing mine. I only have a 1440p ultrawide, so I couldn't do the 4K test. I also ran out of time to do VR test 2.

Motherboard:  MSI Z370 SLI PLUS
CPU:          8086K (delidded)
CPU Freq:     5.1 Ghz
L3 cache:     8 MB
Cores:        6
Threads:      6 (HT = off)
RAM type:     DDR4
RAM size:     16GB (2x8GB)
NB Freq:      4800 MHz (or Uncore Frequency or UCLK)
RAM Freq:     3800 MHz (this is 2xDRAM freq with Dual channel)
RAM timings:  16-16-16-36-666
GPU:          RTX 3080

Optional:

CPU Cooler: Noctua NH-U14S, delid, Thermal Grizzly Conductonaut + Kryonaut
RAM Model:  G.Skill F4-3600C16-8GVK (Samsung b-die)
GPU Model:  EVGA 3080 FTW3 Ultra, stock clocks, silent profile (factory overclocked compared to non-OC models)

1080p
-----
2021-01-18 06:34:41 - Il-2
Frames: 5976 - Time: 60000ms - Avg: 99.600 - Min: 88 - Max: 131

2021-01-18 06:37:41 - Il-2
Frames: 6134 - Time: 60000ms - Avg: 102.233 - Min: 92 - Max: 135

2021-01-18 06:41:04 - Il-2
Frames: 6129 - Time: 60000ms - Avg: 102.150 - Min: 89 - Max: 136

VR Test1 Reverb G2 9.5million 50% SteamVR SS
--------------------------------------------
2021-01-18 06:50:48 - Il-2
Frames: 4135 - Time: 60000ms - Avg: 68.917 - Min: 52 - Max: 90

2021-01-18 06:55:29 - Il-2
Frames: 4204 - Time: 60000ms - Avg: 70.067 - Min: 56 - Max: 91

2021-01-18 06:58:30 - Il-2
Frames: 3992 - Time: 60000ms - Avg: 66.533 - Min: 51 - Max: 87

 

 

  • Upvote 1
Link to comment
Share on other sites

chiliwili69
2 hours ago, Alonzo said:

finally got around to doing mine

Thank yuo Alonzo for your tests. They are well aligned with other Intel CPU at that freqs.

Link to comment
Share on other sites

1 hour ago, chiliwili69 said:

Thank yuo Alonzo for your tests. They are well aligned with other Intel CPU at that freqs.

 

That's what I was afraid of.... 😉   Now I have no excuse, the chip is clearly holding back my GPU. Or, I guess I could just get used to 45 FPS mode and crank the eye candy, that's the other option.

Link to comment
Share on other sites

46 minutes ago, Alonzo said:

 

That's what I was afraid of.... 😉   Now I have no excuse, the chip is clearly holding back my GPU. Or, I guess I could just get used to 45 FPS mode and crank the eye candy, that's the other option.

 

It is what I do.

In my G2 I run 100% resolution - 3172x3100 per eye in my case, Motion Smoothing on forced at 45 fps.

Graphics all cranked up, except no AA - not needed at this resolution.

 

Game is gorgeous.

 

Link to comment
Share on other sites

SCG_Fenris_Wolf
10 hours ago, dburne said:

 

It is what I do.

In my G2 I run 100% resolution - 3172x3100 per eye in my case, Motion Smoothing on forced at 45 fps.

Graphics all cranked up, except no AA - not needed at this resolution.

 

Game is gorgeous.

 

 

I have an RTX 3090 as well. If I do that, as soon as I hit heavy clouds or overcast online - my FPS goes below 45.

 

Could you please see how it performs in: Quick Mission, Rhineland, 11:00 o'clock, heavy clouds / overcast, Artillery & tanks & AAA, 8 vs 8, altitude 1500m. It's my stresstest.

 

I am also interested in how your GPU is used under these circumstances. It seems like mine isn't capped out but can't reach 45FPS anymore.

 

It would also be great to get a spoiler of your graphics section in startup.cfg then.:good:

Edited by SCG_Fenris_Wolf
Link to comment
Share on other sites

=VARP=Ribbon

VR tests:

Reverb G2

i7 10700k 5.0GHz (no xmp profile)

32GB 3200MHz corsair lpx

msi meg z490 unify

evga rtx3080 ftw3 ultra

 

VR test 1- 50% resolution per eye  on G2

Frames, Time (ms), Min, Max, Avg
  3599,     60000,  49,  86, 59.983

 

VR test 2 - 100% resolution per eye on G2

Frames, Time (ms), Min, Max, Avg
  3389,     60000,  48,  77, 56.483

 

I thought that going from 50% to 100% per eye resolution would affect performance a bit more 🤔

Edited by =VARP=Ribbon
Link to comment
Share on other sites

5 hours ago, SCG_Fenris_Wolf said:

 

I have an RTX 3090 as well. If I do that, as soon as I hit heavy clouds or overcast online - my FPS goes below 45.

 

Could you please see how it performs in: Quick Mission, Rhineland, 11:00 o'clock, heavy clouds / overcast, Artillery & tanks & AAA, 8 vs 8, altitude 1500m. It's my stresstest.

 

I am also interested in how your GPU is used under these circumstances. It seems like mine isn't capped out but can't reach 45FPS anymore.

 

It would also be great to get a spoiler of your graphics section in startup.cfg then.:good:

 

Yeah may be another day or so but I will check it out and advise , kinda knees deep in something else right now.

Link to comment
Share on other sites

SCG_Fenris_Wolf
1 hour ago, =VARP=Ribbon said:

VR tests:

Reverb G2

i7 10700k 5.0GHz (no xmp profile)

32GB 3200MHz corsair lpx

msi meg z490 unify

evga rtx3080 ftw3 ultra

 

VR test 1- 50% resolution per eye  on G2

Frames, Time (ms), Min, Max, Avg
  3599,     60000,  49,  86, 59.983

 

VR test 2 - 100% resolution per eye on G2

Frames, Time (ms), Min, Max, Avg
  3389,     60000,  48,  77, 56.483

 

I thought that going from 50% to 100% per eye resolution would affect performance a bit more 🤔

It does, but it depends on clouds mostly, sun and shadows in combination.

 

It doesn't scale evenly with all settings. Some make the resolution increase hit incredibly hard but only in heavy clouds / overcast environment with the sun too.

 

Edited by SCG_Fenris_Wolf
Link to comment
Share on other sites

thermoregulator

OK, so I have just tried to benchmark 6800xt on:

 

CPU: AMD Ryzen 5 5600X, 6 Core - 12 Threads,

 

         PBO auto, auto OC +200MHz, 4850 MHz during benchmark, FCLK 1900 MHz

 

MOBO: Asus TUF GAMING B550-PLUS

 

Memory:

 

2x 16Gb dual rank, F4-3600C16D-32GTZNC, 3800 MHz, timings/subtimings not optimized – left on XMP  - 16-19-19-39-89-1-666 (tCAS-tRC-tRP-tRAS-tCS-tCR-tRFC)

 

Resizable BAR on

 

Something is obviously wrong here:

 

CPU test:

 

2021-01-19 18:41:18 - Il-2
Frames: 5508 - Time: 60000ms - Avg: 91.800 - Min: 82 - Max: 129

 

GPU test:

 

2021-01-19 18:59:23 - Il-2
Frames: 3103 - Time: 60000ms - Avg: 51.717 - Min: 44 - Max: 58

 

So terrible results... much worse than with 2080ti with the same CPU, and absolutely not comparable to my other system with 5900x and RTX 3090. Maybe its some settings in Radeon software, which i am not familiar with, I just set wrong, maybe this game runs terrible on AMD GPU. Or resizable BAR? I don't know. Will try other games, but I am a little busy ATM...

 

Edit: Unable to do VR test. WMR does not work on AMD currently. Well.... that's typical for AMD. They HW is usually usable after a year after release, if you are lucky...

Edited by thermoregulator
Link to comment
Share on other sites

@thermoregulator It also appears worse than the 5700 performance someone tested earlier. 

 

There is probably some sort of driver issue going on here. Especially since your CPU test is significantly under performing for that CPU as well. 

 

Hopefully AMD is able to get it sorted out, but looks like for now, the 3080 is the gold standard for Il-2.

Link to comment
Share on other sites

thermoregulator
38 minutes ago, Voyager said:

@thermoregulator It also appears worse than the 5700 performance someone tested earlier. 

 

There is probably some sort of driver issue going on here. Especially since your CPU test is significantly under performing for that CPU as well. 

 

Hopefully AMD is able to get it sorted out, but looks like for now, the 3080 is the gold standard for Il-2.

Yes, previous CPU test with 2080ti was as expected, so i could be driver issue, problem with resisable bar, or something. Can not test more - I am on my primary system now. Also, I get black screen on my reverb with 6800xt - it does work at all. This Nitro+ 6800xt SE will be sentenced to one year of cryptomining, it seems. 

Link to comment
Share on other sites

RAAF492SQNOz_Steve
16 hours ago, =VARP=Ribbon said:

VR tests:

Reverb G2

i7 10700k 5.0GHz (no xmp profile)

32GB 3200MHz corsair lpx

msi meg z490 unify

evga rtx3080 ftw3 ultra

 

VR test 1- 50% resolution per eye  on G2

Frames, Time (ms), Min, Max, Avg
  3599,     60000,  49,  86, 59.983

 

VR test 2 - 100% resolution per eye on G2

Frames, Time (ms), Min, Max, Avg
  3389,     60000,  48,  77, 56.483

 

I thought that going from 50% to 100% per eye resolution would affect performance a bit more 🤔

 

Alonzo made a comment in another post about IL2 benefitting from faster RAM frequency over latency, with Intel CPU's, so I thought I would try it.

 

I have the same CPU as you and swapped my 3200 MHz G.Skill CL16 Ram for some Patriot Viper Steel Series DDR4 4400 MHz 16GB Dual Memory Kit (CL19) that was quite cheap.

As a result of this change I gained a consistent extra 3 frames Avg ( i.e. 5% improvement) with VR test 2 settings and a HP G2.

 

I was somewhat surprised by this result as conventional  web reviews conclusions suggests that intel CPU's do not benefit much from faster RAM over 3200 MHz.

 

It is up to individuals to decide if spending 100 Euro or so to gain a 5% improvement is useful.

 

I would note that Patriot Viper Steel Series DDR4 4400 MHz uses Samsung B Die modules and I was able to install it and set the XMP profile for 4400 MHz without any issues. Was also able to drop it's frequency to 3600 MHz and set it to CL15 but lost 1.5 fps Avg in the process and have now reset it to 4400 MHz.

 

NOTE: Some reviewers/ testers of the Patriot RAM had trouble getting 4400 MHz and had to set it to 4266 or 4000 MHz but I did not personally have this issue.

 

Added comment below

WARNING/ CORRECTION   Not long after posting the above the Patriot RAM started to cause computer crashes and now reports failures with Memtest 86 using both XMP profiles. So........ after giving it a good wrap I am now having to return it. :( 

Gunna miss those extra 3 fps and what a waste of Samsung B die modules.

Looks like I am going to have to find another supplier if I want those 3 fps back. :(  :( 😪

 

 

 

Edited by RAAF492SQNOz_Steve
Reported failure of RAM
  • Like 1
Link to comment
Share on other sites

29 minutes ago, RAAF492SQNOz_Steve said:

 

Alonzo made a comment in another post about IL2 benefitting from faster RAM frequency over latency, with Intel CPU's, so I thought I would try it.

 

I have the same CPU as you and swapped my 3200 MHz G.Skill CL16 Ram for some Patriot Viper Steel Series DDR4 4400 MHz 16GB Dual Memory Kit (CL19) that was quite cheap.

As a result of this change I gained a consistent extra 3 frames Avg ( i.e. 5% improvement) with VR test 2 settings and a HP G2.

 

I was somewhat surprised by this result as conventional  web reviews conclusions suggests that intel CPU's do not benefit much from faster RAM over 3200 MHz.

 

It is up to individuals to decide if spending 100 Euro or so to gain a 5% improvement is useful.

 

I would note that Patriot Viper Steel Series DDR4 4400 MHz uses Samsung B Die modules and I was able to install it and set the XMP profile for 4400 MHz without any issues. Was also able to drop it's frequency to 3600 MHz and set it to CL15 but lost 1.5 fps Avg in the process and have now reset it to 4400 MHz.

 

NOTE: Some reviewers/ testers of the Patriot RAM had trouble getting 4400 MHz and had to set it to 4266 or 4000 MHz but I did not personally have this issue.

 

 

 

Very interesting thanks for reporting the results.

I am running 3200 MHz CL14 ram and have thought about replacing it with some maybe 4000 MHz.

Just not sure the gain would be worth spending the money though. 

I fly with Motion Smoothing forced at 45 fps anyway so probably I would not benefit so much.

Link to comment
Share on other sites

35 minutes ago, dburne said:

I am running 3200 MHz CL14 ram and have thought about replacing it with some maybe 4000 MHz.

Just not sure the gain would be worth spending the money though. 

 

 

 I don't know if it's an AMD issue or not, but I got no gains going from 3800Mhz to  3866MHz and very similar latency. This may not be apples to apples comparation, because I swapped to a beta bios between the test runs. 

   Intel based systems may be different. Going from 3200Mhz to 4000Mhz you should notice a solid difference. I remember last year's il-2 Benchmark results were very dependent on memory and the top few places were separated because of memory speed. 

 With the Beta Bios I got my infinity fabric to 2000Mhz, but was not able to get my 3600Mhz Gskill kit to 4000Mhz.  I think that's mostly because I'm clueless. 😆

   I'll try again, once new bios is released. I had USB issues with that beta bios and had to flash it back.

 

3800Mhz specs and latency:

ZenTimings_Screenshot.png.b87c8cb9f2b77c92cda5c2fb224788df.pngcachemem3800fast.png.57dd951660e67268fc0985345f5abf12.png

 

3866Mhz specs and latency:

 3866zentime.png.2f781a825ece53e8f26f391f9e2b9ef6.pnglatency3866.png.5c8d41fb4bb5aaa37d23c015ac194201.png

 

Edited by Jaws2002
Link to comment
Share on other sites

RAAF492SQNOz_Steve
10 hours ago, Jaws2002 said:

 

 

 I don't know if it's an AMD issue or not, but I got no gains going from 3800Mhz to  3866MHz and very similar latency. This may not be apples to apples comparation, because I swapped to a beta bios between the test runs. 

   Intel based systems may be different. Going from 3200Mhz to 4000Mhz you should notice a solid difference. I remember last year's il-2 Benchmark results were very dependent on memory and the top few places were separated because of memory speed. 

 With the Beta Bios I got my infinity fabric to 2000Mhz, but was not able to get my 3600Mhz Gskill kit to 4000Mhz.  I think that's mostly because I'm clueless. 😆

   I'll try again, once new bios is released. I had USB issues with that beta bios and had to flash it back.

 

From memory, Alonso indicated, when running IL2, that Intel works better with faster RAM and that AMD works better with lower latency RAM but I cannot remember at what resolutions he was referring to

The Patriot 4400 MHz RAM I used has a "out of the box" memory latency of 45.2 ns so appears to also be a good choice for AMD systems. 

 

Did see a video of the Patriot DDR4 4400 MHz  RAM being overclocked and the guy changed a gazillion settings and ended up running at 4600 MHz with a latency of 36 ns.

Achieving that type of result is well above my skill level. :(

 

EDIT After posting this the RAM started to fail and will be returned to the supplier as faulty

 

Link to OC video for those who like to live on the OC edge......................

 

Quote

 

 

Edited by RAAF492SQNOz_Steve
added section on RAM being faulty :(
Link to comment
Share on other sites

thermoregulator
1 hour ago, dburne said:

 

Very interesting thanks for reporting the results.

I am running 3200 MHz CL14 ram and have thought about replacing it with some maybe 4000 MHz.

Just not sure the gain would be worth spending the money though. 

I fly with Motion Smoothing forced at 45 fps anyway so probably I would not benefit so much.

I believe you can OC your ram to 4000 Mhz easily. 3200 Mhz CL 14 ram are usually good kits...

Link to comment
Share on other sites

26 minutes ago, thermoregulator said:

I believe you can OC your ram to 4000 Mhz easily. 3200 Mhz CL 14 ram are usually good kits...

 

That would be interesting, I think I did once try running it at 3600 without much luck, but I really did not do much like messing with timings.

I love overclocking CPU's just never got much into ram overclocking.

Link to comment
Share on other sites

43 minutes ago, RAAF492SQNOz_Steve said:

From memory, Alonso indicated, when running IL2, that Intel works better with faster RAM and that AMD works better with lower latency RAM but I cannot remember at what resolutions he was referring to.

 

I only know about Intel, really. I'm not sure of the situation on AMD. I'm looking at Chili's spreadsheet right now and I'm probably going to just copy one of the top-performing configurations (well, not the 5950X chip, that's +40% more expensive than the 5900X in Canada!!!).

Link to comment
Share on other sites

thermoregulator
13 minutes ago, dburne said:

 

That would be interesting, I think I did once try running it at 3600 without much luck, but I really did not do much like messing with timings.

I love overclocking CPU's just never got much into ram overclocking.

I am not expert myself. What model of ram do you have? 3200 Mhz cl 14 kits are usually samsung b-dies. I have similar kits on my ryzen system, and achieving 3800 Mhz, 16, 16, 16, 32 was really easy with 1.390 V.  With some more voltage it could definitely do 4000 Mhz and more, but on Ryzen, I am limited by infinity fabric speed. Memory overclocking can quite time consuming though.

Link to comment
Share on other sites

23 minutes ago, Alonzo said:

I'm looking at Chili's spreadsheet right now and I'm probably going to just copy one of the top-performing configurations

 

 

Ryzen Dram calculator tool has good specs up to, and including 3866MHz. For Ryzen, you can just install that and Typhoon burner and you can get very close. No guess work, or too much knowledge needed. All you need is to know where everything goes in BIOS.

Link to comment
Share on other sites

15 minutes ago, thermoregulator said:

I am not expert myself. What model of ram do you have? 3200 Mhz cl 14 kits are usually samsung b-dies. I have similar kits on my ryzen system, and achieving 3800 Mhz, 16, 16, 16, 32 was really easy with 1.390 V.  With some more voltage it could definitely do 4000 Mhz and more, but on Ryzen, I am limited by infinity fabric speed. Memory overclocking can quite time consuming though.

 

G Skill Trident Z RGB 3200 MHz CL14.

Link to comment
Share on other sites

thermoregulator
1 minute ago, dburne said:

 

G Skill Trident Z RGB 3200 MHz CL14.

Those should be samsung b-dies, and 4000 Mhz should be definitely possible with them. I have them too, just NEO version optimized for Ryzen. 

Link to comment
Share on other sites

1 minute ago, thermoregulator said:

Those should be samsung b-dies, and 4000 Mhz should be definitely possible with them. I have them too, just NEO version optimized for Ryzen. 

 

Ok thanks, I may take a look at that. I know they were expensive when I bought them.

Link to comment
Share on other sites

RAAF492SQNOz_Steve
7 hours ago, Alonzo said:

 

I only know about Intel, really. I'm not sure of the situation on AMD. I'm looking at Chili's spreadsheet right now and I'm probably going to just copy one of the top-performing configurations (well, not the 5950X chip, that's +40% more expensive than the 5900X in Canada!!!).

I actually thought your existing PC results were vey good for an Intel system.

 

It is pretty obvious from the feedback that SCG_Fenris_Wolf has left on the forum that even with a 5950X, top notch RAM and RTX 3090, that graphically setting compromises still have to be made when encountering heavy cloud or lots of target scenario's.

 

Upgrading to 5900X is still going to expose you to a GPU bottleneck after a bit (or useful?) amount of a performance gain, if you are trying to drive a G2 at 90 Hz.

 

I have no idea if swapping to 5900x is going to get a 5 or 50 percent performance gain at the resolution you need to run so would be interested in hearing the results if you proceed with swapping CPU's.

 

I have my fingers and toes crossed that Nvidia's (eventual) implementation of Resizable Bar gives us recent Intel build users enough additional improvement to allow us to run some decent cloud settings.

 

 

Edited by RAAF492SQNOz_Steve
Yet another typo
  • Upvote 1
Link to comment
Share on other sites

thermoregulator

I managed to get reverb g1 running on radeon 6800 xt. Firmware update of the headset is necessary - you have to manually find the file on HP site. There is no info on the issue whatsoever, so this might be helpful to somebody.

 

I will try to run Il2 with this combination, but according to poor 2d performance, my expectations aren't really high.

 

Some other games run just fine, so this  low performance issue seems to be il2 related.

 

Link to comment
Share on other sites

@dburne You can check what chips are on the memories module using CPU-Z. What I would was both my Ripjaws V 3600 16-19-19-38 kit, and the TridentZ Neo 3600 16-16-16-32 kit I got were Hynix CJK dies. Apparently G.Skill switched some of the TridentZ Neo kits over to Hynix parts a while ago, so it's hit or miss. 

 

I returned the Trident memory and did a bunch of tweaking on the memory timings, but even getting it up to 3900-16-16 did not reduce my memory latency below about 63ns and didn't have any reliable impact on my frame rate, so I just resent everything back to the XMP profile. No point upvolting the ram for nothing really. 

Link to comment
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
 Share

×
×
  • Create New...