Heavy Optimization Enhancements

While we (well... most of us!) are happy by how DCS improved visually in the last years, there are other things that would need a heavy hand on them: i.e. software optimization.

It's not possible to make a mission within a full-scale war: even if you just "script" and design the ingress-egress (as not fill the map with units that will not have any incidence on your mission because i.e. distance), hardware can hardly manage a regiment vs brigade engagement (soviet formations are of 95MBTs plus IFVs, supporting units, engineering, artillery bat, reconaissance bat. etc vs a NATO equivalent), because is literally unplayable. Don't get me wrong, I do like most of the campaigns available to us nowadays, but I do miss the feeling of "being there", in a proper war and not in some low intensity conflict.

The example stated above is actuallly, still, unrealistic. If you want to make a "realistic" Cold War era scenario, (or a fictional extended Cold War in the mid 90s), engagements in a FEBA would include a minimum of a Division equivalent unit on a width of 10kms.

So, I'd personally think that while future graphic and visual enhancements are welcome, a greatly improved optimization of hardware usage would make a very positive impact, not only in the Multiplayer server day-to-day basis environment, but in some of us that don't mind creating missions as realistic as possible and share them with our fellow pilots/users! :)

What do you guys think?

ED has been pulling the engine apart and starting to multi thread where possible and implementing Vulkan api technology instead of DX12 moving forward, they are making tweaks now as we post to the engine. Here are a few posts that I have come across over the years from one of ED's team members.

Moose is very good a creating a large auto environment that I'm trying to find the time to dig back into.

https://flightcontrol-master.github.io/MOOSE_DOCS/

Then there is also THIS that's on ED's to do list by the looks of it.

Thanks for the answer David.

My aim is not about the generation of the environment (I'm not sure how Moose works, what's its aim and if it can "generate" FEBAs realistically), but how the engine manages it. There is no way it can manage a regimental engagement, let alone a divisional one (which is what you would see if you engage with enemy forces).

The single-core/multi-core is 5 years old. I'm worried about them going to cope with a dynamic campaign if they don't solve the issue about two basic tactical units of an army engaging each other (plus the air units on CAP, CAS, etc).

However, I'm looking forward to what the so-called Vulkan API can do.

See DCS is a real time battle simulator, with real ballistics going on everywhere. So yes, a battle the size your talking about would be very difficult in DCS as there is no fake bubble things. The only way to do a battle that large with the foreseeable tech is to fake the large battle and you just fly the mission in it, that mission is updated on the fly with only the units need as you get to the area, then a battle takes place in real time. I don't see other way for this to work with today's tech and the eye candy we have. I still don't mind the good small scripted missions with the great voice overs, it also has it's benefits of being very different, it just needs to be easier for the builders to make the world more alive in general without crippling our systems.

"DCS is a real time battle simulator".

That's the thing, it can't even simulate a battle. The size I'm talking about it's just minuscule in a full scale war... by no means I want to have a sim who is simulating what's happening a logistics mission 400km away; I'd be happy to have action just around the FEBA.

I made missions where I activate/de-activate batallion size units in order not to have minus 45 FPS, but even in those cases is hard to have a proper fluent running simulator. I'm sure it can be optimized by some factors (like the ones you mentioned, multi-core usage and Vulkan API). I'm just afraid that optimization will be forgotten on the grounds of getting better visuals, therefore defeating the purpose of having a true simulation of a war environment.

Once DCS can tab many more cores and with the current CPU's going 6-32 cores already there will be options to do this.

I don't think the ballistics are the breaking point, throw a few cores at that and done. The more complex thing imho will be the Digital General behind the scene who pulls the right strings, so it is adaptive in it's given sandbox of variations.

I miss this too, the only reason I fly online is because there is no real offline play that catches me. I miss D.I.D.'s approach from decades ago with a full out war over scandinavia, kinda like that.

More cores, more RAM, better code and we get there

And Better AI BitMaster.

I sometimes find frustrating AI can't do a simple instruction...

I’m not a programmer, but would it be possible for DCS to separate a large battlefield* into, say, 6 (2x3) or 9 (3x3) squares** (“local zones”)? Each zone would be hosted by one computer, which would handle all computations within that zone. These zones would be touching each other, so if a unit crossed into an adjacent zone it would be transferred to that particular host.

I could imagine a small bubble-area (a”mini-zone”) which would contain all units close to each other, to prevent blinking out of existence due to immediate transfer to the other host.

*Something appropriate for battalion-sized forces + 200 miles one-way aircraft travel.

**If a Battalion combat radius is 200 miles, each square of a 3x3 mini-zone grid would be approximately 133 miles per side.

Batallion sized forces are insignificant on a large scale war FEBA.

Regiments cannot be managed currently with good performance, and regiments (in a Cold War environment) do not engage on independent operations. The smallest unit would be a Division.

And a Division would have a frontage of between 5 to 10km.

FUBAR situation I'd say.