Jump to content

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


chiliwili69
 Share

Recommended Posts

thermoregulator

Ryzen 9 5900X, stock settings, FCLK 1900 Mhz

Arctic liquid freezer II 360, push-pull config

MSI PRESTIGE X570 CREATION

RAM G-SKILL F4-3200C14D-32GTZN, 32 GB dual rank, 3800 Mhz, 1T, CL 16-16-16-32, tRC 48, tRFC 304

EVGA RTX 3090 FTW3 Ultra Gaming

( AMD Ryzen 9 5900X @ 4423.97 MHz - CPU-Z VALIDATOR (x86.fr) )

 

CPU test:

 

2020-12-22 23:42:34 - Il-2
Frames: 7902 - Time: 60000ms - Avg: 131.700 - Min: 115 - Max: 174

 

2020-12-22 23:51:22 - Il-2
Frames: 8002 - Time: 60000ms - Avg: 133.367 - Min: 118 - Max: 179

 

GPU test:

 

2020-12-23 00:02:43 - Il-2
Frames: 10149 - Time: 60000ms - Avg: 169.150 - Min: 134 - Max: 207

 

2020-12-23 00:16:55 - Il-2
Frames: 10212 - Time: 60000ms - Avg: 170.200 - Min: 135 - Max: 206

Edited by thermoregulator
  • Like 1
Link to comment
Share on other sites

12 hours ago, thermoregulator said:

2020-12-23 00:16:55 - Il-2
Frames: 10212 - Time: 60000ms - Avg: 170.200 - Min: 135 - Max: 206

 

Many thanks. Those results are top!!, both in CPU and GPU.

It seems that your CPU and GPU is the perfect combination!. Is it just stock settings or did you overclock CPU? or GPU?

BTW, How did you solve the 60fps issue with the monitor?

Link to comment
Share on other sites

thermoregulator
1 hour ago, chiliwili69 said:

 

Many thanks. Those results are top!!, both in CPU and GPU.

It seems that your CPU and GPU is the perfect combination!. Is it just stock settings or did you overclock CPU? or GPU?

BTW, How did you solve the 60fps issue with the monitor?

 

Everything was stock during the test, except RAM tuning. According to the 60 fps issue, I started reinstalling stuff and it worked, I don't know what exactly was the cause. 

 

VR performance with G2, 100% SteamVR SS is still not great, though.  I am getting high GPU frametimes (12 ms, CPU frametimes usually bellow 5 ms, there are random frame drops and spikes, both CPU and GPU), fps is between 80 - 90 fps regardless of other settings and the performance is very inconsistent. With reprojection on, i am sometimes locked on 45 fps with ghosting, blurred image, and plenty of visual artifacts, and when I run the same mission again with the same settings, and never drop bellow 90 fps. I believe well known issue with G2 on X570 boards might be the cause. There are occasional problems with tracking and sound, followed with loss of performance, usually solved by rebooting the PC or replugging of the headset. For this reason i didn't do VR benchmark with G2 yet. I am waiting for updates, which will hopefully solve this. 

 

And thank you for you kind words. On that motherboard I have had 3900x before, which I have preordered before reviews went out. I wasn't very happy about its gaming performance and about the fact, that it wasn't able to reach its advertised clock. I have tried 3900xt after that, but even though It was able to boost 300 Mhz higher when idle (which is of course very useful), it's "gaming" clock and Il2 performance was actually even lower. So I am now very happy with the 5900x and wonder, what is the reason for the huge performance increase. One of the reasons is definitely higher sustained clock (3900x was somewhere between 4250-4450Mhz, depending on actual bios, changing up and down, 3900xt reached 4300Mhz typically, and my 5900x can sustain 4950 Mhz). I believe that in Il2, the max sustained frequency on zen 2 CPUs was limited by the frequency of its slowest core. With Zen 3, this is no longer the case. Other reasons might be the latencies, but I still don't fully undestand that 50 % (my guess) performance increase in this game...

Edited by thermoregulator
  • Like 1
Link to comment
Share on other sites

3 hours ago, thermoregulator said:

So I am now very happy with the 5900x and wonder, what is the reason for the huge performance increase. One of the reasons is definitely higher sustained clock (3900x was somewhere between 4250-4450Mhz, depending on actual bios, changing up and down, 3900xt reached 4300Mhz typically, and my 5900x can sustain 4950 Mhz). I believe that in Il2, the max sustained frequency on zen 2 CPUs was limited by the frequency of its slowest core. With Zen 3, this is no longer the case. Other reasons might be the latencies, but I still don't fully undestand that 50 % (my guess) performance increase in this game...

 

My experience is with the Intel processors. We found that IL2 CPU frame times were improved by a faster uncore (ring ratio) and by faster RAM (not lower latency, faster Mhz). This points, to my eye as someone who used to be a software guy, towards the hot spot being something to do with memory access or synchronization across cores. The hot spot is improved by faster synchronization that you get on Intel with a faster ring buffer through RAM and uncore frequency.

 

The Ryzen 5000-series chips have ditched a per-chiplet cache for one giant "infinity fabric" which all the cores use to talk to main memory. This is much faster than per-chiplet caches. Infinity fabric, plus the exaggerated effect of memory speed and tuning on the 5000-series chips (your result, Fenris' result, etc) lead me to the same conclusion -- the IL2 engine needs fast memory access and sync across CPU cores. It could be that cross-thread synchronization or communication is a bottleneck or hot spot of some sort.

 

That's just my conclusion from my semi-trained eyes.

  • Upvote 1
Link to comment
Share on other sites

Hi guys, I would like to benchmark for Il2 bos, in monitor mode no VR.  I understood the procedure but the demo files I downloaded don't work in my updated version of Il2 Bos.  How can I solve?

Link to comment
Share on other sites

Hi Guys I have just received my R5600x (upgraded from R3700x) and have been testing over the last couple of days. Apart from the new CPU my system hasn't changed since my previous benchmark test.

 

My system is a follows:

 

Asus Tuf Gaming X570 plus wifi

Level 3 cache = 32 mbytes

cores= 1 threads= 1

Memory = Ddr 4 16Gb 2x8

Nb Frequency 1600

Dram frequency = 3200

 

Test 1 = Default bios except Xmp profile enabled.

Test 2 = As above with pbo enabled and +200mhz boost overide

Test 3 = As test 2 with Memory timings 14-17-14-14-32-56

 

Test 1 frames 6672 avg 112.2 min 96 max 152

 

Test 2 Frames 7160 avg 119.33 min 104 max 165

 

Test 3 Frames 7360 avg 122.667 min 108 max 161

 

I have only carried out the cpu tests at the moment, the core speed was 4650 mhz default bios and easily hit 4850mhz with +200mmhz with boost overide enabled.

 

Cooling is a Noctua NH-D15s air cooling (max temp during tests 55 degrees.

 

Edited by shirazjohn
Link to comment
Share on other sites

23 hours ago, Startrek66 said:

Hi guys, I would like to benchmark for Il2 bos, in monitor mode no VR.  I understood the procedure but the demo files I downloaded don't work in my updated version of Il2 Bos.  How can I solve?

 

I have uploaded the mission after the last update. The mission is exactly the same. But I saw it has a different date since the game update. You can try to dowload it now and let me know if it works.

9 minutes ago, shirazjohn said:

cores= 1 threads= 1

 

Congrats for your upgrade and test. Just only changing the CPU from 3700X to 5600X gave you +40fps in the benchmark! These new Zen3 are amazing.

But, do you use only one core??

 

It is only impressive you CPU temp. I think that the 5600X with IL-2 would not need more that just the AMD default CPU cooler. Not too much heat generation.

 

 

Link to comment
Share on other sites

41 minutes ago, chiliwili69 said:

 

I have uploaded the mission after the last update. The mission is exactly the same. But I saw it has a different date since the game update. You can try to dowload it now and let me know if it works.

 

Congrats for your upgrade and test. Just only changing the CPU from 3700X to 5600X gave you +40fps in the benchmark! These new Zen3 are amazing.

But, do you use only one core??

 

It is only impressive you CPU temp. I think that the 5600X with IL-2 would not need more that just the AMD default CPU cooler. Not too much heat generation.

 

 

 

That should have been 6 cores 12 threads, i took the info from cpu-z which said 1 core 1 thread for some reason which is obviously not correct.

 

Just realized i have been using an old version of cpu-z , downloaded the latest version and it detected it correctly. 

Edited by shirazjohn
Link to comment
Share on other sites

2 hours ago, chiliwili69 said:

I have uploaded the mission after the last update. The mission is exactly the same. But I saw it has a different date since the game update. You can try to dowload it now and let me know if it works.

Thanks. Solved. 

Link to comment
Share on other sites

OK, I have now replaced my 3600 with a 5600X.

 

The rest of my system is unchanged, which means I'm using the B450 board with the new bios.

 

Motherboard: MSI B450 Gaming Pro Carbon AC

CPU: Ryzen 5 5600X  (stock, 4.6 no PBO. Wraith Prism cooler)

L3 Cache: 32

Threads: 12

Ram Type: DDR4 3200Mhz 14-14-14-34 - 32GB, 4x8GB B-die (which should run as dual, or so I belive.)

NB Freq: 1600

RAM latency: 14

GPU: Palit 3080

 

Vive Cosmos 2016x2380 per eye = 9.59m

 

CPU test: 

 

Frames: 7147 - Time: 70484ms - Avg: 101.399 - Min: 41 - Max: 147

 

VR Test 1: 

 

Frames: 4781 - Time: 75656ms - Avg: 63.194 - Min: 36 - Max: 91


 

It's a marked improvement from my previous averages (which were 70 and 44 repsectiveley) - but it falls short (probably by 12-15%) from what other users of the 5600X have been achieving.

 

Temps on the CPU are fine, always under 70. Boosts to 4650 no issue. My cinebench scores at 4432 and 603 (single core) are about what I would expect, and show no difference to B550/570 users.  So I wonder what could be the issue?

 

Roke

 

Edited by 56RAF_Roke
Link to comment
Share on other sites

On 12/28/2020 at 5:08 PM, 56RAF_Roke said:

Temps on the CPU are fine, always under 70. Boosts to 4650 no issue. My cinebench scores at 4432 and 603 (single core) are about what I would expect, and show no difference to B550/570 users.  So I wonder what could be the issue?

 

Thanks for your test. It is the first Vive Cosmos test we see here.

 

Regarding the performance you had, I have see that there is a new game update (4.505b). I thought that this could be the reason and run the CPU test myself, obtaining:

 

rames: 7108 - Time: 60000ms - Avg: 118.467 - Min: 103 - Max: 158
Frames: 7220 - Time: 60000ms - Avg: 120.333 - Min: 106 - Max: 166
Frames: 7051 - Time: 60000ms - Avg: 117.517 - Min: 103 - Max: 152

 

Which are the same than previous tests with 4.504. So the new game update should not affect performance.

 

I have updated the instructions, asking now for AMD Zen CPUs to report MCLK, FCLK and UCLK using ZenTimings free tool.

 

In the section 7 of the instructions there is a link to a post with a number of things to check when someone achieve a lower performance than expected.

You could also verify the performance of your RAM since it is also a big influencer.

We will need more test from people with B450 Mobos to see if they have same results in IL-2.

What is the resolution of your monitor?

On 12/28/2020 at 5:08 PM, 56RAF_Roke said:

Frames: 7147 - Time: 70484ms - Avg: 101.399 - Min: 41 - Max: 147

 

Hey, There is something wrong in the total time of the test. The Time should be 60000ms. That´s 60 seconds of the test measured by Fraps.

Please, double-check that you put 60s in the Fraps tool, and also the letter "P" to unpause the game and at the same time start Fraps. The same for the VR test.

Link to comment
Share on other sites

3 hours ago, chiliwili69 said:

 

Hey, There is something wrong in the total time of the test. The Time should be 60000ms. That´s 60 seconds of the test measured by Fraps.

Please, double-check that you put 60s in the Fraps tool, and also the letter "P" to unpause the game and at the same time start Fraps. The same for the VR test.

 

Um, thanks Chili. I'd been so busy benchmarking I forgot to reset fraps from 600 down to 60. 

 

Correcting this drags me up a little. For CPU I now have:

 

Frames: 6475 - Time: 60000ms - Avg: 107.917 - Min: 95 - Max: 146

 

And for VR1 its: 

 

Frames: 4220 - Time: 60000ms - Avg: 70.333 - Min: 45 - Max: 91

 

Still down, but not by quite as much. 

 

My MCLK, FCLK, and UCLK are all at 1600. 

 

I had a look at the other thread, but unfortunately I don't think the solution is there. I updated the chipset drivers but that had no effect.

 

Possibly it reflects the beta bios the B400 boards have to run on for now...or possibly I'll try reinstaling windows (which shouldn't have an effect I know), but I'm rather out of ideas for now.  

 

Roke

 

Link to comment
Share on other sites

10 hours ago, 56RAF_Roke said:

Possibly it reflects the beta bios the B400 boards have to run on for now...or possibly I'll try reinstaling windows (which shouldn't have an effect I know), but I'm rather out of ideas for now.

OK, the difference is smaller now. But still a bit low.

 

Is you monitor 1920x1080 native or 4K? I ask that because when used 150% Windows zoom there is an small performance penalty.

 

Do you use any other shader or Mod for IL-2?

 

Also, you can test AIDA64 Extreme memory test and report latency. Apart from Cinebench, try to run some benchmarks (CPU-Z, Userbenchmark, Passmark) and compare at the single-core performance of 5600X

Link to comment
Share on other sites

5 hours ago, Pinello said:

What settings are you guys using for this?:

Hardware accelerated GPU scheduling

 

..and this:

HPET

HPET, the tube..

 

There is always hundreds of miraculous tricks on the web to boost performance. Sometimes they work, sometimes they are fake or don´t impact in IL-2 VR.

In the past, I have tried many of them with zero net results.

 

Anyone who is proposing new tricks could first confirm that they work first for IL-2, running the benchmark five times without and five times with the trick. For the CPU test, GPU test or VR test. So if the gain is significant we can then take a closer look.

 

From here I encourague to anyone to try anything on his PC and share the experience.

Thanks.

Link to comment
Share on other sites

2 hours ago, chiliwili69 said:

 

From here I encourague to anyone to try anything on his PC and share the experience.

Thanks.

 

Tried to help 56RAF_Roke with possible reasons for lower performance. But I get your message. I'm out.

 

Link to comment
Share on other sites

SCG_Fenris_Wolf

Isn't Roke's board a daisy chain topology? I am on the road right now so cannot check myself, I have posted the link to the overview earlier.

 

If that is the case:

He's nerfing his board by using 4x8GB RAM, the performance penalty should be between 5-10%. With 2x16GB sticks, he should clock the 5600X's FCLK to 1800 (or 1900), so your RAM should be 2x16GB 3600MHz.

 

If you can still exchange it (I read your system is freshly bought), do it. IF it is a daisy chained memory topology.

 

Edited by SCG_Fenris_Wolf
Link to comment
Share on other sites

On 12/30/2020 at 2:20 PM, Pinello said:

 

What settings are you guys using for this?:

 

 

 

Thanks Pinello; tried your suggestion but unfortunately that wasn't the solution.

 

On 12/30/2020 at 9:12 AM, chiliwili69 said:

 

 

Is you monitor 1920x1080 native or 4K? I ask that because when used 150% Windows zoom there is an small performance penalty.

 

 

It's a 4k monitor Chili, but a 40' panel so scaling is at 100%. Running passmark single thread with CPU at stock gets me 3304, which is within 2% of the average for the 5600X. 

 

3 hours ago, SCG_Fenris_Wolf said:

Isn't Roke's board a daisy chain topology? I am on the road right now so cannot check myself, I have posted the link to the overview earlier.

 

If that is the case:

He's nerfing his board by using 4x8GB RAM, the performance penalty should be between 5-10%. With 2x16GB sticks, he should clock the 5600X's FCLK to 1800 (or 1900), so your RAM should be 2x16GB 3600MHz.

 

 

I'm sure it is daisy chain Fenris. I've read that running 4x8GB has the RAM behave as dual rank, which (so I was thinking) could offset any hit from the board topography. However, I'm easily confused by RAM on Ryzen so may have that completely wrong. I'll remove 2 of the DIMMS and see if there is any change. The only things new in my system are the CPU and GPU, so I can't exchange the memory (although there's always ebay).

 

If that doesn't work I may try a fresh install of Win10 and Il2 onto a spare hard drive, and run the benchmark on that (just to eliminate the windows factor).    

 

Thanks for the input.

 

Roke

 

*update: going from 4x8 gb to 2x8gb has not had any effect on the benchmark. 

*update 2: going to 3600 mem speed with 1800 FCLK also has not changed my results. 

Edited by 56RAF_Roke
Link to comment
Share on other sites

SCG_Fenris_Wolf

I didn't recommend to remove two DIMMs, it has no effect on the benchmark.

 

What I said was to replace them with 2x16GB and to ramp up FCLK together with memclock may help. 2x16 are dual ranked, 2x8 are not, so they can't take advantage of the topology.

 

It helped for me, the performance difference of 2x16 to 4x8 was considerably up towards 10%, and I put them to 3800. You have put 2x8 in which is something else.. However even if you would attempt to put in 2x16, if your bottleneck is anywhere else, or the fclk doesn't raise enough, it won't help that much. 

 

A clean, fresh system is a premise for any performance hunting anyway ;)

 

Anyway, it's complex as systems are, so good luck tinkering, and don't spend too much going through various options if there's no returning anything, let's be real, it's not worth it (unless it itself is part of the hobby).

Edited by SCG_Fenris_Wolf
Link to comment
Share on other sites

56RAF_Roke

It is part of the hobby, but preferably only a small part 😉

 

I'll try the windows thing, and if that doesn't work will think on before purchasing new RAM which might work, but isn't guaranteed. 

 

Truth is, Il2 vr runs pretty well at the moment, even if I'm leaving some performance on the table. 

 

Cheers 

Link to comment
Share on other sites

WallterScott

 Motherboard: asus rog crosshair viii dark hero
 CPU:                 ruzen 9 5950x
 CPU Freq:        Switch mode 4800 for all core +150 PBO
 L3 cache:        64  MB
 Cores:               16 (number of active physical Cores)
 Threads:           32 (SMT on)
 RAM type:        DDR4
 RAM size:        32Gb (16x16GB) dual
 NB Freq:          1900

 RAM Freq:        3800 MHz 
 RAM timings:  14-14-14-28-314
 GPU:                 3090 TI

Aida latency 54.6

GPU Asus 3090 Strix 

 

CPU Test IL2 4.505

2021-01-01 18:29:18 - Il-2
Frames: 8240 - Time: 60000ms - Avg: 137.333 - Min: 118 - Max: 186

2021-01-01 18:31:30 - Il-2
Frames: 8115 - Time: 60000ms - Avg: 135.250 - Min: 119 - Max: 179

 

VR test with Pimax 8kx will be done later

  • Like 1
Link to comment
Share on other sites

Jaws2002
2 hours ago, WallterScott said:

 Motherboard: asus rog crosshair viii dark hero
 CPU:                 ruzen 9 5950x
 CPU Freq:        Switch mode 4800 for all core +150 PBO
 L3 cache:        64  MB
 Cores:               16 (number of active physical Cores)
 Threads:           32 (SMT on)
 RAM type:        DDR4
 RAM size:        32Gb (16x16GB) dual
 NB Freq:          1900

 RAM Freq:        3800 MHz 
 RAM timings:  14-14-14-28-314
 GPU:                 3090 TI

Aida latency 54.6

GPU Asus 3090 Strix 

 

CPU Test IL2 4.505

2021-01-01 18:29:18 - Il-2
Frames: 8240 - Time: 60000ms - Avg: 137.333 - Min: 118 - Max: 186

2021-01-01 18:31:30 - Il-2
Frames: 8115 - Time: 60000ms - Avg: 135.250 - Min: 119 - Max: 179

 

VR test with Pimax 8kx will be done later

 

 

WoW. That's a monster.😍 :good:

 

You restored my confidence in 5950x. Looks like Robbiec's CPU is bottlenecked by something else.  Your results put the 5950x at the top of the pack in Il-2, just like other games.

  It may be a while until I could get one. I was actively looking for one for a while now, but looks like AMD forgot Canada exists.

BTW. What kind of score do you get in Cinebench R20? My 3950x is sitting comfortable at around 9800 points.

 

Again, that's a monster of a machine. Congrats! 

 

Edited by Jaws2002
Link to comment
Share on other sites

chiliwili69
2 hours ago, WallterScott said:

CPU Test IL2 4.505

2021-01-01 18:29:18 - Il-2
Frames: 8240 - Time: 60000ms - Avg: 137.333 - Min: 118 - Max: 186

 

Congrats for your new Rig, you moved to the AMD world.

 

The result is awesome!.

May I ask some details: Mobo model, RAM model, FCLK (I understand it was 1900), custom water loop?,

Did you overclock the RAM/timings or it was xmp profile?

Link to comment
Share on other sites

WallterScott
1 час назад, Jaws2002 сказал:

What kind of score do you get in Cinebench R20? My 3950x is sitting comfortable at around 9800 points.

On this motherboard, I get in c20 12100-12250 in a multithreaded test, and 650-660 in a single-threaded one. This variation is due to the fact that Dark Hero has Switch mode enabled, which switches between fixed frequency and PBO. Maybe if you use only PBO, the result in IL2 will be better.

 

1 час назад, Jaws2002 сказал:

Again, that's a monster of a machine. Congrats! 

Thanks. I'm happy with the processor 👌

38 минут назад, chiliwili69 сказал:

Congrats for your new Rig, you moved to the AMD world.

Thanks. This is my first AMD system. I had to wait a very long time for the motherboard.

 

40 минут назад, chiliwili69 сказал:

Did you overclock the RAM/timings or it was xmp profile?

No, it's not a profile. All timings are selected manually. But I didn't do it myself. I'm not good at overclocking memory.

 

41 минуту назад, chiliwili69 сказал:

RAM model, FCLK (I understand it was 1900), custom water loop

This is a stock samsun B-die. Yes, FCLK 1900. I haven't gone any further yet. Cooling-modified aquarium chiller.

But here it does not give such an advantage as on Intel. I think that 360-420mm water cooling is enough.

Link to comment
Share on other sites

Jaws2002

   

1 hour ago, WallterScott said:

On this motherboard, I get in c20 12100-12250 in a multithreaded test, and 650-660 in a single-threaded one

 

Very impressive!

 

I know about the hybrid mode Asus rog crosshair viii dark hero has. It can switch on the fly, from PBO, to all core overclock, based on workload. That way you have the high single core boost for gaming and the high multithreaded performance of the all core overclock, in one package.

 It's the first Motherboard to have this ability.  

I think this makes it the best Zen 3 motherboard right now. :good:

  

Link to comment
Share on other sites

Dumb question: Do we have enough data to see a price/performance curve for the AMD chips? It's usually the case that the "top" chip is lots of $$ for not much more performance, but I heard maybe AMD is binning the 5950X so it get better clock speeds in addition to more cores. Anything in our figures to confirm/deny that?

 

Alternate phrasing: For flying in VR should I be looking at a 5600X, 5800X, 5900X or 5950X ? Cost is a factor mainly because I think wasting money is dumb and if I waste it, it's longer until I can get the next upgrade (or if I waste money, I might not choose to get a 3080ti instead of a 3080, etc).

Link to comment
Share on other sites

Jaws2002

You are right. For Il-2, and gaming in general, the higher you go, the less you get for your money. But I think this release, the 5900x makes a lot more sense than the 5800x.

 5600x is the best six core CPU right now. Good Enough for gaming, but we don't know for how long six cores will be enough.

5800X is the best eight core CPU for gaming, but it's too expensive for an eight core CPU. AMD will release some sort of 5700X/5800, but we don't know when. 

   5900X is pretty much the top CPU you want for gaming and will last you a long time.

   For gaming only, I wouldn't recommend the 5950X. Too little performance advantage over 5900x for the huge price difference. Better put that extra money in the GPU.

Link to comment
Share on other sites

chiliwili69
On 1/1/2021 at 8:19 PM, WallterScott said:

All timings are selected manually. But I didn't do it myself

 

So what is your RAM manufacturer/model?

Who then did the manual timming?

Link to comment
Share on other sites

chiliwili69
20 hours ago, Alonzo said:

Alternate phrasing: For flying in VR should I be looking at a 5600X, 5800X, 5900X or 5950X ? Cost is a factor mainly because I think wasting money is dumb and if I waste it, it's longer until I can get the next upgrade (or if I waste money, I might not choose to get a 3080ti instead of a 3080, etc).

 

😁😁

That´s the question over the table now with just a dozen of tests done with different Zen3 chips.

As you know, the performance is not based on the number of cores of each model (I made the test to disable 2 of the 6 cores and was having exactly the same performance) but their ability to run at higher memory frequencies and achieve the lowest latency (I added AIDA64 latency test to the table).

 

This is the table right now:

Zen3.thumb.png.c9de114f3b6366428d2e23003129bac5.png

 

So you can see that in the top we have the 5900/5950, not because their # of cores but their ability to run at 1900:1900:1900 . (same MCLK:FCLK:UCLK is optimal for Zen3)

There is another 5900X running at 1800:1800:1800 and the results are the same than others 5600X/5800X. The difference in performance more or less is about 10% from 1800 to 1900Mhz.

There are RAM models with certified XMP profiles for certain Mobos which can run at 1900:1900:1900.

 

For my short experience with Zen3, maybe the most worth thing (putting $ in the equation) to do for just IL-2 VR is the 5600X, but choossing a Mobo model X570 which will be certified in the QVL of Gskill for memories of 3800-CL14 2x16Gb. So you will not need to tweak the timmings, just default XMP profile.

 

It is also said that future models of BIOS will allow 5600X/5800X to run at 1900 FCLK.

In this page you will see many people with 5600X reaching 1900 and 2000 FCLK.

https://docs.google.com/spreadsheets/d/1dsu9K1Nt_7apHBdiy0MWVPcYjf6nOlr9CtkkfN78tSo

In fact the top (lower) latencies are 5600Xs.

 

I am tempted to go deeper on this just to see how far I can go. But it soo time consuming and not exact science.

 

 

 

 

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

thermoregulator

For the sake of completeness, here are the CPU results of my other system:

 

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 unoptimized – left on XMP  - 16-19-19-39-89-1-666 (tCAS-tRC-tRP-tRAS-tCS-tCR-tRFC)

 

Cooler: NZXT Kraken X72

 

GPU: NVIDIA GeForce RTX 2080 Ti

Details here:

AMD Ryzen 5 5600X @ 4548.94 MHz - CPU-Z VALIDATOR (x86.fr)

 

2021-01-03 12:32:40 - Il-2

Frames: 7044 - Time: 60000ms - Avg: 117.400 - Min: 104 - Max: 157

 

 

Later I will try my 5900x with curve optimiser enabled – it seems to boost higher with it on.

Link to comment
Share on other sites

WallterScott
16 часов назад, chiliwili69 сказал:

So what is your RAM manufacturer/model?

Who then did the manual timming?

g.skill 3200 14-14-14-34  (XMP) 

Setting up secondary and tertiary timings was done by my friend. He is not a member of the forum. As far as I know, it does this based on the program "timings calculator".

Without fully configuring all timings, you can't get good results in the Aida benchmark.

  • Thanks 1
Link to comment
Share on other sites

R3animate

After the next round of BIOS updates next week / PBO2 features being added I'll likely do another round of tests (and provide the missing info from @chiliwili69's sheet such as my mem timings etc). Still no luck on a 3080 for me, at this point I'm waiting for the inevitable 3080TI ;).

  • Thanks 1
Link to comment
Share on other sites

thermoregulator

To all MSI - Zen 3 users: I get very good results with new beta BIOS (ComboAM4PIV2 1.1.9.0) on my system (x570, 5900x,rtx 3090, G2). It improved my overall experience with G2 - greatly improved FPS, better tracking, no sound issuses, no blue screens since BIOS update. PBO2 and curve optimizer improved performance too, CPU seems to boost to 5 GHz consistently during gameplay (I have just set "15" offset to all cores, everything else to auto. I believe you can get even better results setting per core curve.).  I am still unable to set FCLK to 2000 Mhz, though...

 

Edit: OK, I am sorry, Reverb G2 USB issues are still there and they are persisting on my board. Improved performance is still there too, though.

Edited by thermoregulator
Link to comment
Share on other sites

chiliwili69
21 hours ago, thermoregulator said:

For the sake of completeness, here are the CPU results of my other system:

 

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

 

Thank you so much for giving also test results of your "other" nice system (two homes?). It is very interesting that you can compare both CPUs.

It is the first 5600X tested here with 1900FCLK, please check that UCLK is also 1900.

If you could run AIDA64 and provide memory latency it would be nice for comparison purposes.

 

As you described, I understand that you only touched MCLK and FCLK in the BIOS, and all other settings were with XMP profile on. right?

Link to comment
Share on other sites

thermoregulator
39 minutes ago, chiliwili69 said:

 

Thank you so much for giving also test results of your "other" nice system (two homes?). It is very interesting that you can compare both CPUs.

It is the first 5600X tested here with 1900FCLK, please check that UCLK is also 1900.

If you could run AIDA64 and provide memory latency it would be nice for comparison purposes.

 

As you described, I understand that you only touched MCLK and FCLK in the BIOS, and all other settings were with XMP profile on. right?

Yes, you are right - two homes, and one office. I have I7 9700K 5.2 Ghz system too, but i guess the performance of Intel systems are well known now.  Just for fun I could do benchmark on my mining system - Threadripper 1920x, 3x Radeon RX 5700, but I doubt anyone would play on system like that 🙂

 

UCLK on 5600x is 1900 MHz too. You are right about timings and subtimigs. It is just XMP settings. Memory kits on this system ale Hynix DJR, so "stock" xmp timings are not so tight, and its tuning is not so easy as b-dies. With ryzen 3900 and 3900xt I was able to tune those kits beyond ram calculator fast timings easily. After new agesa for that MB is out, I will try to to that on 5600x too, but it seems little harder.

Link to comment
Share on other sites

chiliwili69
21 minutes ago, thermoregulator said:

Radeon RX 5700

Thanks. The interesting thing would be to see the GPU test and VR test with that 5700 GPU in your other 5900X or 5600X systems.

It is nice you are skilled with RAM timming, I will follow your progress with the 5600X.

Link to comment
Share on other sites

56RAF_Roke

Well, I've established that my issue was...um...something to do with my existing Windows install. Clean Win10 install on a different drive gave me the figures below, which I guess are about where they should be now.

 

I'm also now running 16GB RAM at 3600 with 1800IF, timings of 16-16-16-36 (single rank). The 5600X is still at stock

 

CPU: 

Frames: 6983 - Time: 60000ms - Avg: 116.383 - Min: 101 - Max: 152

 

VR1: (Cosmos, 100% SS)

Frames: 4622 - Time: 60000ms - Avg: 77.033 - Min: 54 - Max: 91

 

VR2: (Cosmos, 204%SS, 2876x3396 = 19.5)

Frames: 3823 - Time: 60000ms - Avg: 63.717 - Min: 44 - Max: 81

 

Perhaps the single rank memory is counting against me in VR? But  I can't help but notice that @hellfirelx 's system with the same CPU is out-performing my 3080 in VR with a 1080TI!

 

Thanks for the previous comments again.

 

Roke

 

 

 

 

Link to comment
Share on other sites

thermoregulator
On 1/4/2021 at 12:03 PM, chiliwili69 said:

Thanks. The interesting thing would be to see the GPU test and VR test with that 5700 GPU in your other 5900X or 5600X systems.

It is nice you are skilled with RAM timming, I will follow your progress with the 5600X.

I'd love to say I am skilled, but the truth is that it is purely hit or miss o my side. Trying dram calculator timings on my "secondary" system, 5600x, 3800 RAM, 1900 FCLK, MCLK brought no success yet.   On the other hand, I have been  more successful in buying stuff, and managed to find radeon 6800 XT  in stock for my secondary system. So I am looking forward to bench it when it is shipped.

  • Upvote 1
Link to comment
Share on other sites

chiliwili69
19 hours ago, 56RAF_Roke said:

Well, I've established that my issue was...um...something to do with my existing Windows install. Clean Win10 install on a different drive gave me the figures below, which I guess are about where they should be now.

:good:

Great!  This is one of the purposes of this benchmark. To detect if the PC is performing as expected  in monintor and VR.

 

The results of the 2 VRTests of hellfirelx are significantly different than other people with a 1080 or 1080Ti  (there are 3 including me). 

 

In the VRTest1 you had the same than other people with 5600X/5800X since in the VRTest1 we try to bottleneck only the CPU (assuming GPU is 1080 or above).

In the VRTest2 is a mix of both CPU and GPU. We will need more VRTest2 from people with 3080 cards and Zen3 CPUs.

  • Thanks 1
Link to comment
Share on other sites

Brandy_Lx

Hi,

 

I found time to do a benchmark session with the 3080. I also had a bios flash lately, decided to keep everything on stock PBO = Auto.  As long as I can't get a stable PBO advanced with curve optimizer setting,  I won't bother with that part. Still interested on trying to push the DRAM, but no time for that.

The card is a Gigabyte 3080 RTX Gaming OC 10GB, runs at a stable 1935 Mhz, Mem 9502 Mhz , no OC , its what it runs at on the power limit. Those cards really have no headroom, except you have one with a third power connector. There is some possible tweak with undervoltage on the voltage curve, but no time for that either.

 

New results :

 

CPU :

2021-01-07 23:17:21 - Il-2
Frames: 6890 - Time: 60000ms - Avg: 114.833 - Min: 101 - Max: 157

 

VRTest1 , Index SS 106%

2021-01-07 23:29:18 - Il-2
Frames: 5234 - Time: 60000ms - Avg: 87.233 - Min: 77 - Max: 91

 

VRTest2 , Index SS 216%

2021-01-07 23:39:11 - Il-2
Frames: 5094 - Time: 60000ms - Avg: 84.900 - Min: 72 - Max: 91

 

Exactly as expected, the CPU Test has not changed a lot.

The VRTest1 is the same, the 1080ti was really a beast.

The VRTest2 proves the 3080 is an improvement. That is basically the real win (besides RTX), you can push everything a bit higher and finish off by putting SS or AA as high as you can approaching 100% GPU usage.

 

In Il2 everyday usage I put SS 130% and 2x antialiasing MSAA and follow more or less the guidelines in the other VR threads. I love the immersion, but find spotting and identifying hard. I really dream of 4K resolution per eye in VR, hopefully they can soon drop the fresnel lenses for the next gen of headsets at an affordable price.

 

An additional remark to the VR Benchmark: You have to make sure any smoothing or reprojection is disabled. The best way to be sure of that, is when you use fpsVR, the reprojection % is greyed out and the only negative effect of latency is dropped frames. If reprojection is on, your fps are always dragged down towards the half fps. I see some have around 45 fps as a minimum. That could be a coincidence, but I had the same result when reprojection was on.

 

 

 

Edited by hellfirelx
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...