any news on sim thread multi-threading?

[hanab]

7 hours ago, BIGNEWY said:

As we have mentioned many times work is going well on multithreading, testing is in progress. When we are ready to share more details we will. 

thanks

As many things are still in testing since many years or month :

- vulkan

- dynamic campaign

- dynamic weather

For all of that it will be cool to have News….

(and i don’t speak about all EA modules that are not finished since many years)

[dutchili]

Multi threading implies multi core. Normally the operating system decides which core executes a thread. If you have a hundred cores, the OS may decide to use them all.

For performance expectations: running a mission from a local server (same pc as you play on) gives you a nice preview for just the ai being run in a different thread. The impact is huge. Just try it. 

When more things are run in parallel the game might show the same performance as Solitaire.

I am not involved in ED, nor DCS development, just an ex-programmer who loves the performance effects of multi threading.

[xfirf]

vor 3 Stunden schrieb hanab:

As many things are still in testing since many years or month :

- vulkan

- dynamic campaign

- dynamic weather

For all of that it will be cool to have News….

(and i don’t speak about all EA modules that are not finished since many years)

Sure. This sucks kinda. But to be realistic: Modules pay their salary, not multicore or vulcan implementation. Never forget this.

Assuming those would be real numbers (it´s an example):
Implementing multicore would take 100 developers 1000 days to do this. But ED has 2 devs working on it it would take 50.000 days to do the same thing as the 100. There is progress everyday but the finish line is still far away. And if something goes wrong it will become even worse. Unfortenetly this happens in every software-project.

 

I would love to have multicore and vulcan available NOW. But I have to be patient too. 

[Tom Kazansky]

38 minutes ago, xfirf said:

But to be realistic: Modules pay their salary, not multicore or vulcan implementation. Never forget this.

I disagree. In the future with upcoming new simulation projects, Vulkan and above all multicore could be the main reasons for further income.

But: ED do not need us to tell them. They don't need to be pushed by the community to walk that path (anymore) imho. I bet they want this future proof (VR-)performance at least as much as we do.

[Dragon1-1]

1 hour ago, xfirf said:

Assuming those would be real numbers (it´s an example):
Implementing multicore would take 100 developers 1000 days to do this. But ED has 2 devs working on it it would take 50.000 days to do the same thing as the 100. There is progress everyday but the finish line is still far away. And if something goes wrong it will become even worse. Unfortenetly this happens in every software-project.

This is very wrong, programming doesn't work like that. ED likely has optimum staffing on the multicore project. Let's say it takes 10 devs 1000 days to do it. In best case scenario, how much time would 100 devs take? Well, also 1000 days. Probably more, because the "best case scenario" is 10 devs working, and 90 twiddling their thumbs and being paid for not bothering those 10. OTOH, if you put 2 devs on a project suitable for 100, it'll probably never get done at all, because if there are this many tasks that can be worked on at once, you're probably dealing with a project that no single dev can do. Ultimately, programming is more often limited by having too many contributors, rather than too few. A specific piece of code can be actively worked on by a single dev at a time. You also have to do some tasks before the others can be even started. Depending on the task, a single programmer can hold up the whole thing. This is why in software dev, having a small number of highly skilled staff is better than having a lot of newbies (something management often forgets, thankfully not at ED).

Indeed, this is a very similar conundrum to multicore computing itself. There's only so many things you can split processing into, and if you try to use too many cores for one thing, coordinating them will eat more performance than you gain. In this case, each programmer is a "core" that has a specific task that has to be completed before the program can advance. Just like you can't infinitely get more performance by adding more cores, you can't indefinitely speed things up by throwing more manpower at it. There are exceptions (on both sides of the analogy) where massive parallelization is possible, but rendering engine development isn't one of them.

[cfrag]

2 hours ago, dutchili said:

Multi threading implies multi core.

No, it does not. You can (and we did this back when we had only single-core CPUs) multi-thread on a single core. If you have multiple cores, you simply can execute more than one thread simultaneously. If you don't, the threads are serialized and run one after the other. That's why multi-threaded systems scale to a degree when you add cores. "To a degree" is important because many threads, when they are scheduled have pre-conditions that require other threads to complete before they can start, and any amount of additional cores can't speed up your system if you have a single, long-running thread that all other threads are waiting for. So adding 9 cores to a single-core system will usually not make it run 10x faster.

And that is the art in parallel processing / coding. It's incredibly difficult to correctly parcel out the things that can truly run in parallel. A big, monolithic application that was not designed from ground up to parallel-process can expect a modest boost of 10-20% performance gain with multiple threads, depending on the underlying architecture. 

For example, many missions in DCS rely on (quite extensive) Lua code libraries. You can't simply spin off each script into their own thread, because you'd immediately get into race conditions. So you'll probably run all Lua code in a single thread, gaining pretty much nothing (I'd be surprised if all mission-specific Lua processing wasn't already spun out to their own thread)

2 hours ago, dutchili said:

If you have a hundred cores, the OS may decide to use them all.

If the app can really start 100 independent threads, yes. Usually, you'd see 6 cores maxing out, and 94 idling because there aren't enough independent threads available to start, all queued threads require some running thread to complete before they can start (they require a result, or contest a resource). To repeat: coding for parallel-processing is incredible difficult even with the compiler tools available, and talent is rare. Upgrading an existing app to a parallel architecture is even more difficult and very easy to screw up. So playing it safe, the usual approach is conservative and to go for the quick wins: spin out the obvious unrelated tasks to a different thread (e.g. server), and try to identify the candidates for most gain using profiler tools. 

Having said that, I know that ED has some fine talent at their disposal, so I'm hoping for the best, and keenly await to see what they manage. Even a 20% jump in performance is welcome, and in my eyes a great achievement.

36 minutes ago, Tom Kazansky said:

I disagree. In the future with upcoming new simulation projects, Vulkan and above all multicore could be the main reasons for further income.

That seems disingenuous. ED make money only via model sales, and the models that they already sold can rarely monetize again (except through upgrades). ED don't sell Vulcan, they don't sell the base game. People don't buy anything because of Vulcan, and I'd wager that (unless you mean that graceful plane) most don't even know what Vulcan is. I believe you mean Vulcan's potential uptick in performance could contribute to better sales. But most customers would be hard-pressed to attribute that to Vulcan. They'd simply say 'good, fluid graphics, great model. That's why I bought'. They do buy an aircraft model; they don't buy a graphics API. Vulcan may be an enabler, the model is the product. Put differently: if ED produce a bad model, even the best Vulcan technology won't help them, while a good module usually sells even if performance is marginal. 

Edited September 19, 2022 by cfrag

[BIGNEWY]

5 hours ago, hanab said:

As many things are still in testing since many years or month :

- vulkan

- dynamic campaign

- dynamic weather

For all of that it will be cool to have News….

(and i don’t speak about all EA modules that are not finished since many years)

You have been around a long time hanab, you should understand by now everything takes a long time, nothing is going to change that.
 

1 hour ago, Tom Kazansky said:

But: ED do not need us to tell them. They don't need to be pushed by the community to walk that path (anymore) imho. I bet they want this future proof (VR-)performance at least as much as we do.

We understand the community want more information, but you are correct, we will only pass the information when we are ready to, we all have to remain patient. 

thanks

 

[Tom Kazansky]

17 minutes ago, cfrag said:

hat seems disingenuous. ED make money only via model sales, and the models that they already sold can rarely monetize again (except through upgrades). ED don't sell Vulcan, they don't sell the base game. People don't buy anything because of Vulcan, and I'd wager that (unless you mean that graceful plane) most don't even know what Vulcan is. I believe you mean Vulcan's potential uptick in performance could contribute to better sales. But most customers would be hard-pressed to attribute that to Vulcan. They'd simply say 'good, fluid graphics, great model. That's why I bought'. They do buy an aircraft model; they don't buy a graphics API. Vulcan may be an enabler, the model is the product. Put differently: if ED produce a bad model, even the best Vulcan technology won't help them, while a good module usually sells even if performance is marginal. 

I still disagree. Apart from the point that I don't get why you reduce my statement just to Vulkan, although I said "..and above all multicore could be the main reasons for further income", I for one have already problems with buying new maps and modules, because I exclusively play DCS in VR, and with all those more and more demanding products, my fun using DCS in VR gets more and more reduced. (And it is not that I couldn't affort modern hardware, there is just none that does the job like multicore would do.)

Maybe not all people buy because of APIs but all of us know what feels better as soon as we use it.

So my statement is: without Vulkan and above all without multicore DCS would get a hard time with selling anything (to people like me) as soon, as new contenders come up. And I say this havy hartedly as a DCS fan boy.

And I repeat: I bet ED knows that better than we do. So I'm not too much concerned about the future of DCS.

[cfrag]

Just now, Tom Kazansky said:

I still disagree. [...]  I said "..and above all multicore could be the main reasons for further income"

I guess It's mostly a matter of semantics. I believe what you are saying is 'good performance enables better sales, and without Vulcan/multithreading there won't be sufficient performance". So while we may argue the semantics, I believe we agree on the underlying principle: good performance is one of the preconditions for good sales (as a fellow VR-only user I feel your pain). Whether this is delivered by Vulcan or some other key technology like multi-threading may be beside the point - keeping game performance up is important. And one path to better performance may be Vulcan, another perhaps broader multi-threading. On this I think we both agree.

The second part of my point is that if from now on ED only produces bad models, no amount of performance can save them. People don't buy the Hornet only because it performs well, but because it's an incredibly well done simulation that runs well enough on their system. Performance helps sales, and lowers the entry bar for people with marginal systems (and when you do VR, your system is always marginal  - both Hind and Apache really hit my system hard). But no-one buys the Hornet model exclusively because of good performance. They might decide against buying the model if performance on their system is too low, and having Vulcan and multithreading in place may have saved that particular sale.

Put differently: Vulcan/Multithreading itself is not a reason to buy a module, and it can be what saves a sale on a marginal system. 

[xfirf]

vor einer Stunde schrieb Dragon1-1:

This is very wrong, programming doesn't work like that. ED likely has optimum staffing on the multicore project. Let's say it takes 10 devs 1000 days to do it. In best case scenario, how much time would 100 devs take? Well, also 1000 days. Probably more, because the "best case scenario" is 10 devs working, and 90 twiddling their thumbs and being paid for not bothering those 10. OTOH, if you put 2 devs on a project suitable for 100, it'll probably never get done at all, because if there are this many tasks that can be worked on at once, you're probably dealing with a project that no single dev can do. Ultimately, programming is more often limited by having too many contributors, rather than too few. A specific piece of code can be actively worked on by a single dev at a time. You also have to do some tasks before the others can be even started. Depending on the task, a single programmer can hold up the whole thing. This is why in software dev, having a small number of highly skilled staff is better than having a lot of newbies (something management often forgets, thankfully not at ED).

Indeed, this is a very similar conundrum to multicore computing itself. There's only so many things you can split processing into, and if you try to use too many cores for one thing, coordinating them will eat more performance than you gain. In this case, each programmer is a "core" that has a specific task that has to be completed before the program can advance. Just like you can't infinitely get more performance by adding more cores, you can't indefinitely speed things up by throwing more manpower at it. There are exceptions (on both sides of the analogy) where massive parallelization is possible, but rendering engine development isn't one of them.

It was more like an easy numbers example to show that they can´t set all developers on it and how this affects the speed of the developement.

[Dragon1-1]

Well, it didn't show it. It was also wrong, since the real reason they can't set all their devs on it is that it does not affect the speed of development the way you describe, the relationship is nonlinear and beyond a certain point it outright inverts. This is a common mistake, but "too many cooks spoil the soup" is very much a thing in software dev.

[Tom Kazansky]

1 hour ago, cfrag said:

and when you do VR, your system is always marginal

but here lies the rabbit in the pepper (like we Germans say ), it is only DCS that puts problems to my hardware, even in VR. I (hope I) would not need a "science fiction system", as soon as the DCS engine is up to date. What it will be some day.

 

1 hour ago, cfrag said:

Vulcan/Multithreading itself is not a reason to buy a module, and it can be what saves a sale on a marginal system. 

the lack of those for me is getting more and more a reason not to buy demanding modules (yet!). (Don't worry ED, I'll still buy a lot of them anyways, but just because of hope.)

But I agree, it's a lot just semantics and we don't disagree in general. I just want to make it clear, that the bread would not be sold without a competitive oven

Edited September 19, 2022 by Tom Kazansky

[_UnknownCheater_]

Always,I am very glad to pay more money to support ED's works.But I need a performance improvement answer.I am ready to buy 40 Series GPU card,I hope I can get 60FPS in game with 2K.

Edited September 19, 2022 by zhouyut001

[dutchili]

8 hours ago, cfrag said:

 

 

Thanks for adding the nuances!

Edited September 19, 2022 by dutchili

[funkster]

At least 2yrs on from the first discussions and approx a year since testing was supposedly taking place. Patience is a virtue but it also runs thin.

Lets hope we get some solid news soon (this year!) Or that someone has enough sense to see the gap in the market and produce an alternative product. 

[Rudel_chw]

1 hour ago, funkster said:

someone has enough sense to see the gap in the market and produce an alternative product. 

What market? People seems to forget just how much of a small niche this market actually is, DCS currently has 1500 players on the last day at Steam, IL-2 only 300 players, while a game like Dota2 has over 700,000 ….
(Data from https://steamdb.info/graph/ )

 

See the difference?  We are so lucky to have ED on this market, that it annoys the sh.. out of me when I read some of the petty complains that show up on this Forum.

Edited September 20, 2022 by Rudel_chw

[Tom Kazansky]

25 minutes ago, Rudel_chw said:

We are so lucky to have ED on this market

This.

It even scares me when I see how much money is spent and earned with LOL or FN skins, thus I appreciate every single day that there are people at ED and partners, dedicated to their dream which is coincidently matching mine.

So I'm eager to get news on multithreading as soon as ED has something to share. (Just to get back to topic.)

Edited September 20, 2022 by Tom Kazansky

[Germane]

vor 5 Stunden schrieb Rudel_chw:

What market? People seems to forget just how much of a small niche this market actually is, DCS currently has 1500 players on the last day at Steam, IL-2 only 300 players, while a game like Dota2 has over 700,000 ….
(Data from https://steamdb.info/graph/ )

 

See the difference?  We are so lucky to have ED on this market, that it annoys the sh.. out of me when I read some of the petty complains that show up on this Forum.

 

Relaxe dude! He only says that he would like to see another simulator. There is nothing wrong with wanting to have an alternative.

By the way. I don't think 1500 players only on Steam is that little.

[HoBGoBLiNzx3]

9 hours ago, Rudel_chw said:

What market? People seems to forget just how much of a small niche this market actually is, DCS currently has 1500 players on the last day at Steam, IL-2 only 300 players, while a game like Dota2 has over 700,000 ….
(Data from https://steamdb.info/graph/ )

 

See the difference?  We are so lucky to have ED on this market, that it annoys the sh.. out of me when I read some of the petty complains that show up on this Forum.

 

You don't think that might have something to do with it's lack of core features that a large number of people have been complaining about for 4 years not being addressed, or at least not in a somewhat reasonable timeline? I think most DCS players, especially VR users, and those with mid tier components, have been pretty patient but I know quite a few that have just given up on the sim, at least until it's figured out. When ED says that multicore support is in the final stages of testing over a year ago, and announced it over 3 years ago, you can't help but feel strung along. If even 3 people are working on this project.. that's how many hours between them since it's been started... or even since it's been advertised at almost complete?

That's over 6000 hours in just ONE year... let that sink in. 

[funkster]

Yeah i agree no market at all in flight sims. MS2020 passed 2 million sales in a few months of release not to mention massive sales in thier internal marketplace addons. 

Oh and nice of you to call other members opinions "petty". I can see the name is very apt. 

That aside. I think hobgoblin sums it up nicely.

 

Edited September 21, 2022 by funkster

[draconus]

16 hours ago, HoBGoBLiNzx3 said:

You don't think that might have something to do with it's lack of core features that a large number of people have been complaining about for 4 years not being addressed...

Yeah, I don't think so. (btw, why only 4 years?) Even in this niche of niche most people fly casually, autostart their aircraft and just want to blow stuff. I doubt more than 5% can do a cold start, AAR and Case I. So no - some performance problems in VR, simple ATC or lack of DC is not the sale breaker - it's just rare for people to really want to become serious military pilots... as IRL.

[sirrah]

Although this is what happens on a forum (and I guess what they are meant for in the first place), this whole discussion is pretty pointless if you ask me..

 

In the end, we're all on the same page and wish for improved performance within DCS. The sooner the better also. ED already gave their answer, so what's the point in trying to fish for more? For obvious reasons, ED will only provide us with news on this huge matter, when they are ready to do so.

 

I don't understand why every week several people seem to think, that it helps when they ask for an update. Clearly the entire community is waiting for a news update, so why not just get in line and wait patiently like the rest of us. I think the main reason why some use strong wording in this, or any of the other threads on this topic, is that it gets a tad annoying. Like the kids in the back asking if we're there yet. Asking ain't gonna speed up the process and yes, ED knows this is what the community screams for. Don't be scared, ED won't forget telling us whenever there's something to be told.

[St4rgun]

2 hours ago, sirrah said:

I don't understand why every week several people seem to think, that it helps when they ask for an update. Clearly the entire community is waiting for a news update, so why not just get in line and wait patiently like the rest of us.

All right, a small supplement for your kind words: https://www.psychologyofgames.com/2019/10/loading-when-were-willing-to-wait/

The essence is this:

Quote

The idea is pretty simple: show people what’s taking so long and they’ll be more satisfied with the experience. Sometimes they may be even more satisfied despite a longer wait.

[linside]

4 hours ago, draconus said:

Yeah, I don't think so. (btw, why only 4 years?) Even in this niche of niche most people fly casually, autostart their aircraft and just want to blow stuff. I doubt more than 5% can do a cold start, AAR and Case I. So no - some performance problems in VR, simple ATC or lack of DC is not the sale breaker - it's just rare for people to really want to become serious military pilots... as IRL.

I think more that 5%.... This game is near one hobby. This (like other sim) has a lot options to upgrade or change the environment. 

[HoBGoBLiNzx3]

16 hours ago, funkster said:

Yeah i agree no market at all in flight sims. MS2020 passed 2 million sales in a few months of release not to mention massive sales in thier internal marketplace addons. 

Oh and nice of you to call other members opinions "petty". I can see the name is very apt. 

That aside. I think hobgoblin sums it up nicely. 
 

 

 

We aren't supposed to nate other "games" but being that they already released what we have been waiting for and then now just posted this.. I think it's fair that it's shared...

 

Edited September 22, 2022 by HoBGoBLiNzx3