released Multi-Threading Discussion

[dutchili]

Can somebody check in task manager if the thread 'DCS release process' is still running?

Edited March 7, 2023 by dutchili

[phant]

SOURCE
https://discord.com/channels/1071433028045377637/1071617175531552858/1080236158325366876
https://discord.com/channels/1071433028045377637/1071617175531552858/1080248120371326986


Bye
Phant

[okopanja]

2 minutes ago, Aghimien said:

200fps claim not valid.

ben_der talks about VR.

[Tom Kazansky]

7 minutes ago, Aghimien said:

I believe that claim by Ironmike on the performance in multithreading has been reviewed somewhat. That performance claim has been lowered down by Ironmike while answering questions here on thread of new stuff coming from Heatblur in the open beta update.

200fps claim not valid.
 

This.

See here plz:

 

[Eaglewings]

This.
See here plz:
 

Tom. That is the thread. Message No. 21 is where Ironmike addressed the earlier claim.

Ryzen 5 5600: RTX 3070: 32g Ram: Virpil Cm3 Throttle :Thrustmaster warthog flight stick: Thrustmaster TFRP rudder:Samsung Odyssey plus Vr headset :

[Tom Kazansky]

It is happening...

 

[phant]

Benchmark 1 - Empty mission with few AAAs and few planes

 

Benchmark 2 - Absolutly unplayable mission with 15 fps average at it's lowest point, now 62 average.

SOURCE https://www.reddit.com/r/hoggit/comments/11nsi40/quick_benchmark/

 

Bye
Phant

Edited March 10, 2023 by phant

[Tom Kazansky]

16 minutes ago, phant said:

Benchmark 1 - Empty mission with few AAAs and few planes

 

Benchmark 2 - Absolutly unplayable mission with 15 fps average at it's lowest point, now 62 average.

SOURCE https://www.reddit.com/r/hoggit/comments/11nsi40/quick_benchmark/

 

Bye
Phant

 

Thanks @phant ! This is exactly what I wanted to see. Good job ED so far!

[Bertuz]

Hmm, i'm a bit disappointment after all what I read. 

Tested flying straight ahead in F16 - Free Flight - Marianas. Resolution 3308 x 2208 

null

When I open up the FPS counter it states GPU bound. 

[Hiob]

3 minutes ago, Bertuz said:

Hmm, i'm a bit disappointment after all what I read. 

Tested flying straight ahead in F16 - Free Flight - Marianas. Resolution 3308 x 2208 

null

When I open up the FPS counter it states GPU bound. 

It really varies berween different scenarios. Gains range over a wide spectrum from none to ? (max 40% so far for me).

When you a really strictly 100% GPU bound, where should the gains come from? MT only affects CPU directly.

[Tom Kazansky]

3 minutes ago, Bertuz said:

Hmm, i'm a bit disappointment after all what I read. 

Tested flying straight ahead in F16 - Free Flight - Marianas. Resolution 3308 x 2208 

With that resolution it is quite obvious that you hit the GPU limit rather than the CPU limit especially in freeflight. You might wanna try some heavy CPU missions.

[Bertuz]

Yeah i think its a combination between high resolution and older GPU which makes it GPU bound and not gaining much. 

[Tom Kazansky]

Just now, Bertuz said:

Yeah i think its a combination between high resolution and older GPU which makes it GPU bound and not gaining much. 

Exactly.

[TheFreshPrince]

Did some benchmarks with undervolting/OC tests a week ago and just tried the same today after the patch, same settings:

Benchmark today after update: min 65 max 91 avg 75

Benchmark one week ago: min 61 max 85 avg 68

That's good already, but I definitely can feel it more in MP. On the full GS server it was very smooth. Like really, really smooth. Barely below 60 FPS, worst drop was to about 40 FPS. Nice!

(2D 1440p with RTX 2080 and Ryzen 5 5500, 32GB)

[Beirut]

Which settings are most affected by MT?

[Thinder]

On 3/6/2023 at 3:17 PM, SkateZilla said:

Its Valid in DCS.

B.Die Ram + SMA + VCACHE = Better Performance in DX9/11 Graphics Engines.

 

That's why B.Die is recommenced by G.Skill support, lower latency and Ryzen3/3D are designed for lower latency, it works with ANY application, now the latest I get from them is this:

I was worried about the CPU IMC support for 64GB because those sticks are two ranks instead of 4 and in theory, Ryzen  IMC are limited to 4 but after testing it doesn't seems to be the case with this kit and CPU, on the other hand I quote eurotech replying to my last email:

Quote

Dear Customer,

The main specification between F4-3600C14Q-32GTZRA and F4-3600C14Q-64GTZRA are the same, except that with dual-ranked modules, you should get a slightly better memory bandwidth performance. So overall, in most use-cases, there should be a small bump in performance if memory bandwidth matters to the application, but because they’re both DDR4-3600 CL14, it’s likely not very noticeable unless through benchmark numbers. The only major difference would be the larger memory capacity that your system can work with.

So considering the gain in performance, I  don't see where the "overpricing" is, people spend much more than this premium in CPU water cooling only to see a lower gain and lose their manufacturer warranty, from my PoV, it is optimization.

  

On 3/6/2023 at 9:58 PM, Raven (Elysian Angel) said:

Will this continue to be relevant once we have a fully fleshed out DCS 3.0 with MT + Vulkan, or will the gain from 3D cache be lower than it is now?
Or are flight sims a genre were 3d cache will always offer substantial benefits?

Latency related to cache or RAM is independent from the game engine, and in fact, the reason behind the Ryzen 7 5800X 3D cache design, it's all about data transfer times, it is only logical that a lower latency RAM will add to the gain obtained by the CPU cache.

In my case, going from 32GB to 64GB is motivated by fact that I do video editing and also play around with 3D Studio Max, I am planning to design some concepts ships based on Elite Dangerous, not that I hope they would use them but since this sort of activity was my job in the past it became a pleasant hobby.

Also on top of capacity, according to eurotech, there is a "a slightly better memory bandwidth performance".

 

Edited March 10, 2023 by Thinder

[BitMaster]

Very nice, it's so much smoother. MP with a full server was as smooth as never before imho.

[Aquorys]

Just flew in multiplayer on 4YA, the improvement seemed to be rather marginal. Running an i7 at ~4.7GHz, 80 GiB RAM, RTX2070S, NVMe storage, Oculus Rift S. I'd say the average is 30-40fps, but it still drops extremely every now and then, especially after using the map or external views of other aircraft, it would go to 10 fps on the ground (have seen as bad as 4 fps in the past). If I take the VR headset off for 30 seconds and put it on again, it's doing 30-40fps again. If I don't, then the frame rate remains bad forever. Not sure who's to blame, but I tend to suspect that this is either a programming error in some timing routines in DCS, or a fundamental design error. I mean, the quality of the Oculus drivers is really bad, but in this case, I don't think they are the source of the problem, because all other VR software, including other flight simulators, would also occasionally experience really bad frame rates temporarily for various reasons, but they all recover from it without me having to do anything.

[Fakum]

How will the implementation of MT impact settings for NVidia control panel? In other words, what would be the recommended changes if not default?

[Thinder]

I had ONE good run using the M-T release then it really got bad but I don't think it is coming from DCS, Windows 11 Pro for example doesn't do scheduled clean up despite me setting it for daily, I had to get rid of 240MB of junk files manually.

It went some way to recover the Visual quality and FPS I had in January but it's not totally there yet, I think the implementation of VR itself needs extra work, something changed since and it can be improved further, GPU usage was 100% but running cool, CPU usage was <> 16% at around 76°, I might be able to use a more demanding map for the CPU.

After that, the game started in a grey, unmapped mesh in a blue environment, it takes <> 10 sec before everything else falls into place, mapping, cockpit instruments, terrain and then it is real slow and jerky, I tested RAM sticks latency and CPU multithreading performances with CPU-ID CPU-Z and it's fine.

I tried testing with FPS VR  before that in the "normal" version, I had a low 40 FPS with lower settings than I used in January, Frametime was sky-high but I keep in mind that now I have to deal with bad AMD drivers too, so I'm reinstalling it and Drivers and if I can recover what I experienced during the first run it will be better.

 

Edited March 11, 2023 by Thinder

[Beirut]

1 hour ago, Fakum said:

How will the implementation of MT impact settings for NVidia control panel? In other words, what would be the recommended changes if not default?

 

That would be good.

 

I'm hoping to see what in-game settings are helped the most by MT. Does it help the clouds but not the water? Does it help MSAA but not SSAO? That kind of thing.

[Hoirtel]

Multi threading working very well for me, still CPU bound, average is up nicely, but... wow what great consistency now too. Graphs say it all. Track run in VR oculus quest 2. Max resolution, 80Hz, in normal oculus runtime (not openxr) recordings are Plasma Torture Map.

MT on top

Edited March 11, 2023 by Hoirtel

[Dragon1-1]

7 hours ago, Thinder said:

I tried testing with FPS VR  before that in the "normal" version, I had a low 40 FPS with lower settings than I used in January, Frametime was sky-high but I keep in mind that now I have to deal with bad AMD drivers too, so I'm reinstalling it and Drivers and if I can recover what I experienced during the first run it will be better.

You're not the first AMD user to report less than great results with MT. IMO, this needs to be dug into, I'm not an AMD user myself (not anymore, in part thanks to things like that), but getting in touch with other people with AMD GPUs and maybe starting a bug thread could be of use.

My 1080ti performs much better than it used to, I'm able to drive a Reverb G2 at 100% and a steady 30FPS, no other antialiasing but there's little need for it. Still tuning the settings, but the majority of people do seem to get results in that vein.

[Thinder]

Tested in 2D. High settings, it runs real smooth, I believe my issues are coming from the VR implementation, not Multithreading at all, and I'm sure it can run with the same high settings in VR once those issues and AMD drivers are sorted.

This is the first few seconds of the game, Mirage F1 EE, Free flight Caucasus, after that the FPS is real low, so there is a fundamental problem with the way VR runs VR with my Pico, I will submit a ticket to support to try to figure out what's wrong...

Edited March 12, 2023 by Thinder

[Thinder]

Corrected.

Thanks to support who responded quickly, I can play this release and it's great, I gained <> 15FPS flying the Mirage F1 at very low level Nellis AFB map in high settings, I'm sure I can improve on this with better drivers and a few tricks...