SIM hyperthreading support up to 16 cores in DX10 is possible!

In my last thread "Intel or AMD CPU for DCS hyperthreading?" the majority declared Intel as the winner due to threading limitation. Now again I'm not sure any more...

Recently I bumped into a utility Add-On which made me exhume the old FSX dinosaur out of its grave for an unbelievable revival and gave it wings with such a performance, I couldn't believe it was possible. All my six cores were in full operation to the extreme limit and boosted my frames to a hitherto impossible height! (see attached pic) This utility offers also many configuration options in selecting Add-On priorities and affinities for further boosting.

So if hyperthreading support up to 16 cores is possible in DX9 + DX10 with such an old stiff engine like FSX, why is it not possible in a new flexible DX11 engine like DCS? Immagine the DCS performance with the AMD Ryzen Threadripper 1950X 16 cores all in use simultaneously!

Beyond that this greek team also offers an Add-On FSX Booster that can decrease / increase FSX frames WHILE playing, so that in a dog fight the Simmer can simultaneously increase the frames, or decrease them for a higher resolution to enjoy the beautiful environment sight.

This is indeed the unavoidable challenge to the ED programmers.

Red-Bull does not give wings to DCS. Those Add-Ons do

Not worth the implementation effort, if you have a decent setup you can run DCS full go. If you are poor, just tune down those settings and you'll run it just fine. After 2 minutes you won't notice the difference.

ED has more important things to do like...fix the guidance model for missiles (20 years and there is still no dynamic gain, something falcon 4 had out the box), and the drag model for missiles.

Not worth the implementation effort, if you have a decent setup you can run DCS full go. If you are poor, just tune down those settings and you'll run it just fine. After 2 minutes you won't notice the difference... ED has more important things to do like...fix the guidance model for missiles (20 years and there is still no dynamic gain, something falcon 4 had out the box), and the drag model for missiles.

If you took the trouble to do some research on this subject, you would have found out that there is no topic of higher importance than FPS performance. You have a lot of homework on DCS performance to catch with before diverting the topic out of its core, because the framework of this topic are the unemployed CPU cores turning thumbs and waiting to be engaged.

Just in case the thread title might cause any confusion between game performance, multithreading, hyperthreading and the role of DirectX in all of this, here more feedback:

This Multicore-App does feature Supporting Hyper-threading enabled systems. I am also aware that DCS uses a different engine than FSX. But the fact is: this App did break the two cores limitation barrier in FSX! (which uses DX9 or DX10)

Now I assume DCS uses DX11. Here is a summary from the Internet on DX:

1) DirectX is an API (Application Programming Interface) used for developing video games and other such media applications, much like OpenGL.

2) The differences between DX9 and DX10 is that Anti-Aliasing is supported in DX10 and DX10 supports shader v4.0 Anti-Aliasing is used to smooth the edges of objects in a video game and makes little difference in games playing in low resolutions. With it on, it will noticeably slow down performance in certain games while delivering an almost unnoticeable difference in visuals. The difference between DX10 and DX11 is that DX11 supports Tessellation. Hypothetically, DX11 is supposed to enhance performance with compatible systems as well as providing a graphics increase with a feature called tessellation. An example of the graphical improvements with tessellation on can be viewed here: http://www.youtube.com/watch?v=KtbUWiZ48a0 , tessellation is turned off at first. Too much tessellation can cause terrible lag just like too much anti-aliasing.

And a summary on Multithreading & Hyperthreading:

- Hyperthreading is a hardware thing and Intel branding. Most other people call it Simultaneous Multithreading (SMT). To the programmer, two hyperthreads look like two CPU cores. On the hardware side, multiple hyperthreads share a single core. (In the case of Intel, there are two hyperthreads per core).

- Multithreading (or multithreaded programming) is generally considered the concept of using more than one thread context (instruction pointer, registers, stack, etc.) in a single program. (Usually in the same process or virtual address space).

To frame the core of this topic, (in case I do decide to enhance my PC with a 16 cores AMD CPU) is the question if IT IS possible to engage the remaining unemployed CPU cores in a future version of DCS. In that case, I will decide for AMD and not Intel CPU.

But I thought it was changing to vulkan?

 

And a summary on Multithreading & Hyperthreading:

- Hyperthreading is a hardware thing and Intel branding. Most other people call it Simultaneous Multithreading (SMT). To the programmer, two hyperthreads look like two CPU cores. On the hardware side, multiple hyperthreads share a single core. (In the case of Intel, there are two hyperthreads per core).

- Multithreading (or multithreaded programming) is generally considered the concept of using more than one thread context (instruction pointer, registers, stack, etc.) in a single program. (Usually in the same process or virtual address space).

 

To frame the core of this topic, (in case I do decide to enhance my PC with a 16 cores AMD CPU) is the question if IT IS possible to engage the remaining unemployed CPU cores in a future version of DCS. In that case, I will decide for AMD and not Intel CPU.

Yes, we or ED just don't know by how much yet with Vulkan? DCS may use all the cores much much more. Unit ED gets this done and optimizes it, we just don't know how good and how much better it would be with say....amd threadripper 1950x (16-core) or my 7700k (4-core) and using all the 4 cores plus threads to the full could be enough to match my 1080ti or the new 2080ti? Or having an amd threadripper 1950x might even allow you to control 1000 AI unit battle in sim @ 70+ fps....maybe...:music_whistling:

The first chart is exactly DCS and is bottlenecked/limited by the one core feeding the GPU, like most games/sims. Vulkan will help a lot we hope:smilewink:

But I thought it was changing to vulkan?

Changing to Vulkan means more graphics performance for the GPU, but if there is no CPU core extension planned in the future DCS core engine, the performance will have to suffer from the bottleneck effect between the two.

Yes, we or ED just don't know by how much yet with Vulkan? DCS may use all the cores much much more. Unit ED gets this done and optimizes it, we just don't know how good and how much better it would be with say....amd threadripper 1950x (16-core) or my 7700k (4-core) and using all the 4 cores plus threads to the full could be enough to match my 1080ti or the new 2080ti?

Is this just an insider Info or an official DCS?

I'm just about to update to a High-End Computer starting with the Motherboard, which is dependent of the CPU type Intel or AMD, as you know.

I just want to know how long should I be patient with my PC purchase, because in this case, I'll definitely go for the 16 cores AMD CPU!

Changing to Vulkan means more graphics performance for the GPU, but if there is no CPU core extension planned in the future DCS core engine, the performance will have to suffer from the bottleneck effect between the two.

Offloading the graphics to the GPU and freeing up the core that bottlenecks and limits DCS fps and objects, see pic above.

Well that IS good news indeed!

Is this just an insider Info or an official DCS?

I'm just about to update to a High-End Computer starting with the Motherboard, which is dependent of the CPU type Intel or AMD, as you know.

I just want to know how long should I be patient with my PC purchase, because in this case, I'll definitely go for the 16 cores AMD CPU!

Not official at ALL and like I said it will be along time before we or ED know just how good at first and it could be years of optimizing to see it's full true potential.

I'm waiting myself to decide on my next build and perhaps go threadripper? or Intel's new CPU? I just don't know what will work best and if DCS will use all the cores....

...Not official at ALL and like I said it will be along time before we or ED know just how good at first and it could be years of optimizing to see it's full true potential.

Vulkan or DX11 apart, if DCS is flooded to the neck with many tasks at the same time, I wonder if it is not wise to contract FSPS with that 3rd Party Multicore-App. If they were able to crack that Microsoft FSX engine, I'm sure with a DCS collaboration of your Program engine source code they'll find something. In case of failure, they'll be liable for the time investment damage themselves, and DCS has nothing to lose!

It is a win-win situation.

I still do not understand why someone is still thinking that ED is not doing their best if it would be sooooo easy to implement a technique, what ever, which would improve the performance.

Do you think they do not care and do not want to sell the software anymore? Wake up, it is their work, so they will always try to be competitive.

Discussing a framework from the outside is always completely useless if you do not know the code base you want to improve, because you have to integrate them both.

Gesendet von meinem F5121 mit Tapatalk

I still do not understand why someone is still thinking that ED is not doing their best...

Do you think they do not care and do not want to sell the software anymore?

Wake up, it is their work, so they will always try to be competitive...

Discussing a framework from the outside is always completely useless if you do not know the code base you want to improve, because you have to integrate them both.

DCS Firewall must find a remedy against digital psycho crap :doh:

Whoppsi, hit a nerve?

Nothing personal intended, but these threads, especially the discussion about threading and how many cores are used and how it should be done etc open up many, many times - that's all what is funny in my opinion ;)

Gesendet von meinem F5121 mit Tapatalk

Edit: whatever - it was just a statement that, to me, it is not feasible how endless this discussion seems, for years now. Without deeper knowledge any proposal is not really useful using this and that technique - and in addition it is known that Vulcan is being chosen as the path to go.

Despite you started with insulting like psycho crap, not me :)

I did not mean especially you, I mean every thread about this topic for the last years, maybe an issue is that I also wrote "someone", better should have written "anyone"?

Whoppsi, hit a nerve? Nothing personal intended, but these threads, especially the discussion about threading and how many cores are used and how it should be done etc open up many, many times - that's all what is funny in my opinion ;)

Right now, it's just talk and marketing. If it delivers anything tangible the DCS "World" will rejoice and the negative people like I will be happily wrong.

I tend to make comparisons on real world examples. Why do other sims with a lot more money and sales still trip up here? Why is this such a common issue? Why can't one simply ... yadah yadah. So look at the other popular sim based on FPS and consider It's approaches. It takes the one big CPU hog that is not needing a strict time requirement (ie asynchronous AI calcs) and farms them off to another core for which it can create as many threads as the server admin has cores and deal with AI calcs. It doesn't do much else, yet it's still CPU bound.

Now I might be off. But if that's the best that a better funded company with a long standing mature product can do, then how is a small group of developers going to exceed that and pull up FPS bound by server CPU cycles enough to make a significant improvement?

I don't really need to know the code level, although I am familiar with what can trash a DCS server in quicktime. Unit count is the killer, but one step lower, each of the detection checks against each of the others, causes little dips of processing every second. Route calculations and AI calcs come in after that, heavy combat with lots of AI really saps the available FPS. Simple clients, do not. a Client on it's own on the server is just the export of it's data and report to the Server. Large servers can hold beyond 60 clients happily. Start adding AI to these and they quickly plummet.

Long story short, if ED can take the expensive non-timely async calcs to another core(s), the sim would scale much better. Nothing out-of-the-box enables that to happen without a large coding effort, I believe they attempted this in proof of concept internally and the results didn't justify the effort. A later story said they might. There's lots of posts on this, some ED even replied to.

Many server admins code AI to turn off and on dynamically to reduce the effect of the in line AI calcs. Not as well as say a "bubble" - since MP is where these issues mostly come out in the wash and bubbles are not predictable for many players who may roam anywhere they please.

End of the story will begin with what ED delivers. Since they don't ever get specific with what they are doing in front of the public, until they are sure it works and can be delivered, right now, it's not worth anyone postulating on the net effect of any change. Suffice to say, none of it is simple and definitely not managed with external software.

...End of the story will begin with what ED delivers. Since they don't ever get specific with what they are doing in front of the public, until they are sure it works and can be delivered, right now, it's not worth anyone postulating on the net effect of any change. Suffice to say, none of it is simple and definitely not managed with external software.