Patch notes discussion thread 25th January 2023

[Thinder]

12 hours ago, Dragon1-1 said:

Well, I have a 1080ti, so I can't exactly run on 60% with those settings, but at 40% (basically Reverb's native resolution) and 25FPS it does seem somewhat nicer. Fortunately, I'm not sensitive to low framerates as long as they're consistent, however, I've noticed shimmering that appears much more severe than it used to be. I haven't been able to get it back to where it was on previous builds. I'm pretty sure I had more FPS as well. Oh, well, we'll see what the next patch does. 

The 30 series cards are out of my price range right now, but I've been considering upgrading to a 2080ti once DLSS drops, if it's really that great (I also saw a used Titan RTX at a surprisingly non-insane price, but I couldn't go for it right now and it'd inevitably require a new PSU, anyway). I'd really like to see how my current rig measures up once Vulkan drops, though. Given how old the 1080ti is at this point, it's holding up really well. 

The 1080Ti is a great card, it is limited by today's standards but was no doubt the best of its time, I used mine for two years and with some RAM/CPU optimization with my 5600X I was able to run DCS in VR at higher resolutions and still experience a smooth game.

It's not all in the GPU performance, there is a RAM/CPU bottleneck and under load it forces the CPU controller to throttle down, this limits all channels including GPU, at the end of the day, I gain more bounding my 5600X to a Cl14 B.die kit than any O.C I could have thrown at my 1080Ti.

I sold it to CEX for £150, after two years of service, it have been boosted with Afterburner but never O.Ced, I didn't need it since I had the performance I needed thanks to the RAM.

 

[YoYo]

21 hours ago, St4rgun said:

 

 

In MSFS I'm using DLSS because there it's the best overall AA technique speed / quality vise. In DCS as the current MSAA implementation leaves a lot of shimmering objects that's why I'm more than curious to see DLSS in VR in action here also. It can solve the temporal aliasing flickering and the transparent / semi transparent textures' aliasing as well. Let's see.

By the way as long as we speak I tried some new settings with my rig with really breathtaking results is DCS, even the clouds quality is highly improved on standard settings (SP only):

• set the OpenXR in Windows to 300% custom render scale (!!!) with motion reprojection disabled
• set OpenXR Toolkit as the following:
  • Performance tab:
    • upscaling to 60% with FSR -> this equals roughly 3200x3200 pixels resolution for my G2
    • frame rate throttling to 45
  • Inputs tab:
    • Shaking reduction to -70% to have smooth and stable IHADSS
  • System tab:
    • double check that motion reprojection set to Default (not forced on)
• And my DCS settings pages:

 

 

With these settings my 3090 can almost always produce 45 fps with all the airplanes I have (20) even on the ground and locked 45 fps in air.

The clouds seem much nicer despite being set to "standard" only and the overall cripsness of the image if phenomenal. The sweespot became much bigger as well.

Personally I don't like having 55-60 fps while that causes more microstutter on the G2 than 45 fps. As long as I can't have steady 90 its better to have stable 45. And I really hate motion reprojection's arifacts, so that's always set to off for me in DCS.

 

 

 

This is a little mess for me but if you like it, why not  . If you turn off MR and you limit FPS (without MR) you havent smooth picture but blurry affect with small stutter when you pan a head. This settings are good obly if you look directly straight ahead only.

[Dragon1-1]

2 hours ago, Thinder said:

It's not all in the GPU performance, there is a RAM/CPU bottleneck and under load it forces the CPU controller to throttle down, this limits all channels including GPU, at the end of the day, I gain more bounding my 5600X to a Cl14 B.die kit than any O.C I could have thrown at my 1080Ti.

Well, my RAM is CL16, so not that much slower. That said, the 5800X3D's design addresses this bottleneck specifically with its prodigious L3 cache. I'm GPU-limited pretty much all the time in DCS, which I suppose is a good sign, in a way. I'm looking forward to what multicore will change in this, I will certainly be waiting for at least the next patch before doing anything with my rig except for adding storage (hoping to clean up my utterly inane current setup a bit).

[Thinder]

1 hour ago, Dragon1-1 said:

Well, my RAM is CL16, so not that much slower. That said, the 5800X3D's design addresses this bottleneck specifically with its prodigious L3 cache.

It's not the speed of the Cl14 which makes the difference but the latency. The only Cl14 chips are made of B.die which is a much more stable material and allow for a much wider range of timings, it was designed first for O.C and used by both Intel and AMD players alike when their CPUs couldn't run at 3200MHz yet, the timings allowed them to do that.

The CPU controllers are limited in terms of frequency and number of ranks, if one CPU is used out of those limits it hits a bottleneck under load, and throttle down and this would happen with the 5800X 3D with a Cl16 kit as well, the cache didn't change the controller limits or remove the RAM/CPU bottleneck, it just allows for a much more efficient management of the data.

The Cl16 RAM kits doesn't remove this bottleneck, it doesn't have the right timings, the Cl14 does and under load, be it at 4K, the CPU controller can keep all channels open and bandwidth isn't lowered.

Lower latency is precisely what the Ryzen 3 and 3D were designed for, the difference between the two is that the 3D makes a much better use of Cl14 RAM kits thanks to its cache, and I tested both a 5600X and my actual 5800X 3D back to back with only one difference: The 5600X was boosted with Ryzen Master and was running on 3200Mhz Cl14 while my 5800X 3D was not boosted and ran on a 3600 Mhz kit, the 5600X also clocks faster at 3.7GHz/4.6GHz vs 3.4GHz/4.5GHz for the 5800X 3D.

Here is the result of a test with 3DMark Pro at 4K MSAA X 2, it is obvious that the GPU (same 1080Ti for both except boosted with Afterburner for the 5600X) gain a huge amount of speed, 18.91%, the CPU also ran faster at 14.34%, that's what you'll lose at 4K with a Cl16 kit.

The controller of both those CPUs wouldn't take 3600Mhz with a Cl16 kit, when I tried to O.C my Cl16 Crucial to 3600MHz, I instantaneously lost just above 1% running the exact same test, I never use another test or (Fire strike) or settings (4K MSAA X 2), simply because I don't play at lower resolutions.

Quote

I'm GPU-limited pretty much all the time in DCS, which I suppose is a good sign, in a way. I'm looking forward to what multicore will change in this, I will certainly be waiting for at least the next patch before doing anything with my rig except for adding storage (hoping to clean up my utterly inane current setup a bit).

Well, I'm not, the gain in CPU speed under load already make sure I don't even have a bottleneck with my RX 7900 XTX, in fact, I even gain a bit and the GPU itself does too, so giving credit for an increase in performances in gaming to the cache of the 5800X 3D alone is plain wrong, in fact, there are games where there is no gain at all or where it even loses to a 7 5800X.

In passing, the 7 5800X3D is 3.8% faster in multithreading than the 7 5800X, so I don't anticipate much trouble with my rig's performances, this is what optimization does for a PC.

[BIGNEWY]

removed the insults and off topic 

please stay on topic here

thank you

[Ajax_]

I stuck it out for about a week and have now reverted back to 2.8.1.34667

I would have liked to stuck it out till the next fix, but kept having DCS crash before I could complete a mission in my campaign.

@BIGNEWY Please don't take this post as an insult or off topic, I think letting others know some of us have had to rollback is fair for the thread...

(I really tried everything I could find to help out, learned some things about OpenXR and the toolkit, but in the end as everyone knows, crashes are frustrating).

Hoping the next patch resolves things!

Cheers

Win10-64bit/32GB Ram/i7-9700K/RTX2080Ti/Pimax8kx/CougarHOTAS

 

 

[BIGNEWY]

16 minutes ago, Ajax_ said:

I stuck it out for about a week and have now reverted back to 2.8.1.34667

I would have liked to stuck it out till the next fix, but kept having DCS crash before I could complete a mission in my campaign.

@BIGNEWY Please don't take this post as an insult or off topic, I think letting others know some of us have had to rollback is fair for the thread...

(I really tried everything I could find to help out, learned some things about OpenXR and the toolkit, but in the end as everyone knows, crashes are frustrating).

Hoping the next patch resolves things!

Cheers

Win10-64bit/32GB Ram/i7-9700K/RTX2080Ti/Pimax8kx/CougarHOTAS

 

 

No worries, 

open beta is a testing build and it will have its ups and downs, hopefully those who are having issues will have better results in the next build. 

thank you

[St4rgun]

On 2/1/2023 at 3:13 PM, BIGNEWY said:

We are very close with multithreading and testing continues, DLSS will be in a future build also, but it will all depend on test results.  

Thank you @BIGNEWY for the answer. May I ask if then current stall of cloud aliasing problem means that any progress on that field can be expected only with the new engine in the future (maybe DLSS as a new anti-aliasing technique)? Because the clouds in VR are still highly aliased without any change despite the last several patches.

[BIGNEWY]

5 minutes ago, St4rgun said:

Thank you @BIGNEWY for the answer. May I ask if then current stall of cloud aliasing problem means that any progress on that field can be expected only with the new engine in the future (maybe DLSS as a new anti-aliasing technique)? Because the clouds in VR are still highly aliased without any change despite the last several patches.

For the best clouds you need to use ultra, of course this comes with an FPS impact. Distant horizon clouds are the main issue with aliasing, the team will look to tweak it more and once DLSS is here the clouds will need reviewing again. 
 

[St4rgun]

4 hours ago, BIGNEWY said:

For the best clouds you need to use ultra, of course this comes with an FPS impact. Distant horizon clouds are the main issue with aliasing, the team will look to tweak it more and once DLSS is here the clouds will need reviewing again. 
 

As now I'm highly supersampling my G2 (300% in OpenXR and 60% in OpenXR Toolkit FSR) the clouds are really nice even at STANDARD despite of being on the horizon or being closer (in fact they're much nicer that using the G2's "native" 100% resolution of 3100x3100). It seems that there's an exact type of cloud layer which is still extremely aliased, but only that, not the other types. And the position of the sun can also influences the effect (worse aliasing at low sun angles).

After the last tweakings several patches ago the overall quality of the clouds really became much better, only that stubborn cloud type remained so ugly. Thanks for any help, becuse the ULRA setting is extremely power hungry in VR (when raining it literally becames a slideshow with a 3090).

[SPAS79]

7 hours ago, St4rgun said:

(when raining it literally becames a slideshow with a 3090).

Also your signature shows you have 32GB, which can be a limit sometimes. 

[giullep]

Dear ED, after You planned a future update on 1st march  (i hope a big update..) now we dont have any more infos, video or another update…could you kindly give any infos about this great work? Or you are to much hard working   on it? Thx 

[draconus]

8 hours ago, SPAS79 said:

you have 32GB, which can be a limit sometimes

Is it really your experience that turning on the bad weather makes 32GB RAM overload?

[St4rgun]

8 hours ago, SPAS79 said:

Also your signature shows you have 32GB, which can be a limit sometimes. 

Changing cloud preset from standard to ultra should (in theory) give the GPU more work, I have no idea how should my 32 GB be the bottleneck there. But using ultra is really useless for me, because with standard in 80% the clouds are pretty nice except some exact cloud type which is heavily aliased.

Not to speak of the lack of proper anti-aliasing at certain objects and edges, semi transparent surfaces etc. It's obvious that in VR even the MSAA 4X does work only partially, that's why I'm eagerly waiting for DLSS.

[SPAS79]

12 hours ago, draconus said:

Is it really your experience that turning on the bad weather makes 32GB RAM overload?

 

12 hours ago, St4rgun said:

Changing cloud preset from standard to ultra should (in theory) give the GPU more work, I have no idea how should my 32 GB be the bottleneck there. But using ultra is really useless for me, because with standard in 80% the clouds are pretty nice except some exact cloud type which is heavily aliased.

Not to speak of the lack of proper anti-aliasing at certain objects and edges, semi transparent surfaces etc. It's obvious that in VR even the MSAA 4X does work only partially, that's why I'm eagerly waiting for DLSS.

naw sorry I didn't mean 32GB will limit your options with weather/clouds... In general 32GB can be not enough for busy scenarios. 

[giullep]

Il 8/2/2023 at 01:58, giullep ha scritto:

Dear ED, after You planned a future update on 1st march  (i hope a big update..) now we dont have any more infos, video or another update…could you kindly give any infos about this great work? Or you are to much hard working   on it? Thx 

Ed dont reply…

[Lange_666]

20 days to 1st march, just sit it out and go flying...

[draconus]

2 hours ago, giullep said:

Ed dont reply…

Because they don't give out the patch notes/content before the patch day. About "...infos, video or another update" - it comes every friday.

Edited February 9, 2023 by draconus

[St4rgun]

On 2/9/2023 at 8:31 AM, draconus said:

Because they don't give out the patch notes/content before the patch day. About "...infos, video or another update" - it comes every friday.

 

We'll see in today's newsletter. I would be more than happy if the following weeks' newsletters will be about some BIG changes in details. I can easily wait until March to have them in OB 

[giullep]

Il 9/2/2023 at 08:31, draconus ha scritto:

Because they don't give out the patch notes/content before the patch day. About "...infos, video or another update" - it comes every friday.

 

ok..today...

[Lange_666]

8 hours ago, St4rgun said:

I would be more than happy if the following weeks' newsletters will be about some BIG changes in details.

It's BIG indeed, new pilot body for the F-18C.

[giullep]

Ahahah.. a very big info…a new (unknown and little used…) module and the pilot body of the hornet… very interesting and big infos about next update…ED you can do better…

[St4rgun]

3 hours ago, giullep said:

Ahahah.. a very big info…a new (unknown and little used…) module and the pilot body of the hornet… very interesting and big infos about next update…ED you can do better…

Don't be unfair. These are really HUGE - a module (!) OK, by a third party, but a module (!) a 3D model by ED's 3D modeling artist(s) which is near completion and a stick base which is really professional but of course made by a totally different company, not ED. As nowadays at least 98% of ED are working on the new pilot body's modeling it's more, than reasonable.

You can understand that as each newsletter length is quite limited it's impossible to write every itty-bitty tiny details every week which are really marginal and only a handful of hardcore DCS users are interested in like the new EDGE features, VR performance plans and preliminary results, experience with the new AA techniques like DLSS, MT testing progress and Vulkan implementation, challenges on the field of clouds in VR and new shadow techniques sneak peek. If ED would always write about those topic it would be like hourly adding a new screenshot about texturing a renewed AI bomber... It's indeed useless.

And I'm pretty sure if ED would dare to write even a line about those insignificant details then at least a dozen of angry users would comment here that they are absolutely not interested in any of those technical nonsense and by the way where is my X/Y module and Z map what ED promised!

Be fair and understanding: simply poor ED can't win, they'll be always the ones to blame. I really feel sorry for them.

Edited February 11, 2023 by St4rgun
typo

[giullep]

I'm sorry...you are absolutely right.we must be happy...now we know soon we will have a new flying jewel and also the new model of the f-18 pilot (but they already said it..)

[Baldrick33]

In my experience, developers write chapter and verse when they want to explain why things are taking longer than expected. So I look at the lack of info on the core stuff as a positive!

These things go to the wire, it is too early for a definitive list of what's coming in March, we just need to be patient (yeah right!)