in development Vulkan API Discussion

1 hour ago, Buzzer1977 said:

Modern APIs like Vulkan enable developers to make use of multiple cores ...

At least that's what NVidia says ...

Nvidia

 

If the problem is the computing power for IA behavior, the triggers, the trajectory of bullets, the damage model, the système, the flight model...Do you really think using Vulkan (dedicated to rendering) will help a lot ?

Vulkan = improvement for scenario for which I already have good fps. What I need is imrovement for scenario for which I have wrong fps and you know what ? They are CPU bounded.

And last thing : there is a room for improvement without Vulkan, as I improved fps for some cases which are more bugs than a real issue

So having devs spending their time on Vulkan instead of working on real improvement would be a waste of time and resources...

1 hour ago, lefuneste01 said:

If the problem is the computing power for IA behavior, the triggers, the trajectory of bullets, the damage model, the système, the flight model...Do you really think using Vulkan (dedicated to rendering) will help a lot ?

Vulkan = improvement for scenario for which I already have good fps. What I need is imrovement for scenario for which I have wrong fps and you know what ? They are CPU bounded.

Why just limit parallelization to some AI stuff and create another bottleneck, by touching all that data again in another thread only to pump it to the GPU, if you can handle it all in parallel and pump the data to the GPU directly from each thread?

On 3/28/2022 at 4:05 AM, SkateZilla said:

Or they could use the Vulkan Hardware Assisted or Ray Tracing SSAA and only have to program a function once and work on all GPUs from 2016+

There's a big fat radio silence about the new engine / Vulkan implementation. I really hope that they not reached to some uresolvable issue since the last news, or just stopped the progress to redesign previously done parts...

I know that it's not much what BIGNEWY wrote here sometimes that "they are making good progress", but I'm missing it. More than five months passed since the last "Development report".

16 hours ago, 3WA said:

Well, there goes my GPU.

Mine's from 2010

vor 13 Stunden schrieb St4rgun:

There's a big fat radio silence about the new engine / Vulkan implementation. I really hope that they not reached to some uresolvable issue since the last news, or just stopped the progress to redesign previously done parts...

I now that it's not much what BIGNEWY wrote here sometimes that "they are making good progress", but I'm missing it. More than five months passed since the last "Development report".

Yeah,

an update would be really nice.

OK, let's ask them!

Dear @BIGNEWY, can we have any info on the engine rewrite progress? It would be particularly interesting, because:

• right now the GPU prices are still hugely inflated, but there's a hope that they will be lowered soon (https://videocardz.com/newz/asus-announces-price-cuts-of-up-to-25-in-april-for-geforce-rtx-30-gpus)
• the new generation GPUs will come to the market this fall, so if someone calculates on still buying from the existing generation nVidia GPU they have to do it in the right moment / price
• it would be really a must to know WHICH GPU is optimal for DCS with the new graphic engine especially in VR with high resolution headsets.

It's easy to advice to buy the strongest GPU (3090), case solved. But it's far from optimal solution price/performance wise. I'd like to make a decision to only buy the neccessary GPU, because right now DCS in VR is the most resource hungry application for me. For every other task my existing hw is more than enough.

Do we have the possibility to have ANY performance numbers connected to the new engine BEFORE fall this year? I don't ask for any deployment date for Vulkan / Multicore / VR optimization, just preliminary ballpark numbers to help us to make decisions in these hard times.

Thanks for any info!

31 minutes ago, BIGNEWY said:

Hi, 
sorry I can not give any details yet or performance test results, we will have to continue to be patient there. Work is progressing, and I think we will see multicore before vulkan, but I can not give a timeline. If you are running DCS on your current hardware and it is smooth I think you can look forward to an improvement, I dont want to say its going to be a magic bullet but I am hopeful. Regarding hardware, the fastest GPU and the card with the most VRAM will always be the best choice. 

Thank you very much for the quick response. The performance test results are not needed right now, I asked if it's feasible to get ANY performance data before fall.

3 hours ago, St4rgun said:

Thank you very much for the quick response. The performance test results are not needed right now, I asked if it's feasible to get ANY performance data before fall.

Massive changes in the core take massive efforts and massive time, but these changes don't pay the bills like new modules do. 
I think they know they have to fix the core with multithreading and Vulkan and i also bet they will do their best to get there.

On 3/29/2022 at 4:28 PM, BIGNEWY said:

Hi, 
sorry I can not give any details yet or performance test results, we will have to continue to be patient there. Work is progressing, and I think we will see multicore before vulkan, but I can not give a timeline. If you are running DCS on your current hardware and it is smooth I think you can look forward to an improvement, I dont want to say its going to be a magic bullet but I am hopeful. Regarding hardware, the fastest GPU and the card with the most VRAM will always be the best choice. 

Distributing the thread's workload across multiple cores increases CPU performance. Vulkan can make use of all the available cores by redistributing the workload across a number of threads and scale the work without just generating more work. In order to be truly scalable, the Vulkan API spreads its workload across multiple threads with as close to zero overhead as possible.

Vulkan does not scale multiple cores, because this is an explicit API after all, but Vulkan provides the tools and mechanisms needed in order to allow applications to scale as they want. Some of this can be achieved by simple design choices and yet more by providing objects that explicitly cater to multi-threading.

I sincerely wonder why MultiCore and Vulkan shall be introduced in discrete way. Although two different APIs, as far as I understand MultiCore and Vulkan are mutually dependent to yield the best performance and efficiency result. I am eager to hear some more detailed explanation from ED - thank you.  

46 minutes ago, Fastbreak said:

Distributing the thread's workload across multiple cores increases CPU performance. Vulkan can make use of all the available cores by redistributing the workload across a number of threads and scale the work without just generating more work. In order to be truly scalable, the Vulkan API spreads its workload across multiple threads with as close to zero overhead as possible.

Vulkan does not scale multiple cores, because this is an explicit API after all, but Vulkan provides the tools and mechanisms needed in order to allow applications to scale as they want. Some of this can be achieved by simple design choices and yet more by providing objects that explicitly cater to multi-threading.

I sincerely wonder why MultiCore and Vulkan shall be introduced in discrete way. Although two different APIs, as far as I understand MultiCore and Vulkan are mutually dependent to yield the best performance and efficiency result. I am eager to hear some more detailed explanation from ED - thank you.  

Vulkan is Multi-Threaded, therefore Multi-Core if Multiple Cores are available.

ie Vulkan is Multi-Threaded Code, But the Scheduler (Whether it's Windows or the Chipset) controls what thread's go where.
Whether it scales multiple cores or is confined to a single core is up to the Scheduler and how much load each core can handle,

A Single core can handle more than 1 thread, but it will not process all threads simultaneously, the Scheduler will cycle thru them to allow for the tasks to process, For Example, the average Windows 10 Computer with just common startup programs runs about 3,000 Threads, and the Scheduler controls which thread gets Cycle time etc.

Typically Windows Scheduler will not put multiple Vulkan (or DX12) threads on the same core if other cores are available, it will put the threads across 2 or 3 cores easily.

As long as the code is written to be parallel, and multi-threaded, the Scheduler in Windows/Chipset will take care of the rest.



Re-coding the DCS Core to be parallel and multi-threaded will take everything that DCS Does and allow the Scheduler to spread it out, so DCS will not force 1 or 2 cores to process EVERYTHING, including Draw Calls:
-Object Physics: for Aircraft, Vehicles, Weapons, Weather,
-AI: Pathing, Decisions, interactions
-Emitter Physics: Radar, FLIR
-Events/triggers:

All of these take significant CPU Cycles, and process continuously, 
Now if everything relies on these items to be processed individually, then one thread will get clogged up waiting for everything to complete, causing GPU Commands to get bogged down and FPS to drop, or lag/desync delay in commands being registered.

Allowing the Core to expand to multiple threads vs 1 or 2 will allow everything to process more fluidly.
then you add Vulkan, which takes the GPU Commands that were DX11 and saturated the CPU with draw calls, and remove the CPU Overhead, and the single thread saturation and allow the CPU to process commands across multiple threads, and send GPU Processes directly to the GPU freeing up the CPU to do other work.

I've played ED's flight simulators since the Flanker days.  I think DCS is absolutely amazing and I've invested a lot of money and time in the game.  With that said I am however growing impatient of the lack of news of the implementation of Vulkan. It's been talked about for several years now and nothing's changed. It's a bitter pill to swallow having to keep waiting and waiting when people like myself have literally spent thousands on their gaming rigs just so they can get what can only be described as mediocre framerate performance when in VR.

Reading the latest newsletter about polished features and bug fixes with no mention of Vulkan is incredibly frustrating.  As much as I love DCS if Vulkan doesn't land sooner than later I'll moth balling DCS and migrating to MSFS2020 because I just cannot justify spending so much money on hardware when DCS delivers such poor performance with its current engine. ED really do need to prioritise Vulkan and keep people regularly up to date with the development, otherwise people will stop investing in the game and eventually they'll drop it completely.

Careless about Vulcan, multi threading is what’s needed and long overdue.

15 hours ago, Infiltrator said:

I've played ED's flight simulators since the Flanker days.  I think DCS is absolutely amazing and I've invested a lot of money and time in the game.  With that said I am however growing impatient of the lack of news of the implementation of Vulkan. It's been talked about for several years now and nothing's changed. It's a bitter pill to swallow having to keep waiting and waiting when people like myself have literally spent thousands on their gaming rigs just so they can get what can only be described as mediocre framerate performance when in VR.

Reading the latest newsletter about polished features and bug fixes with no mention of Vulkan is incredibly frustrating.  As much as I love DCS if Vulkan doesn't land sooner than later I'll moth balling DCS and migrating to MSFS2020 because I just cannot justify spending so much money on hardware when DCS delivers such poor performance with its current engine. ED really do need to prioritise Vulkan and keep people regularly up to date with the development, otherwise people will stop investing in the game and eventually they'll drop it completely.

I agree that DCS does need engine improvements to keep going but I have very different experience to yours. I find (VR) DCS runs pretty well and better than MSFS. However DCS does really struggle and drop significantly when there are lots of units placed down, especially lots of moving or shooting units, many radars, lots of weapons being deployed. This drains the CPU quickly and results in FPS tanking. However considering than none of these things are even possible in MSFS I cant see how changing sims would make it different.

Multi-core processors have been out for 17+ years and today's modern games can utilise 6+ cores. ED can produce as much content as they like & release all of those stunning trailers. But if they don't implement Vulkan I'm gone. Some people have spent £2,500 on a GPU, and for what? A game that needs a lot of fine tuning in order to run 45 frames in VR? 

I was once into Arma 3. That game was release almost a decade ago & still Bohemian Interactive's flogging that dead horse with their single core engine by continuing to release sticking plaster content. I fear ED will follow their lead in continuing to polish the game & add more content on a outdated engine in order to try to keep their audience & to "keep the lights on" as so many like to mention.

I've literally just spent £550 on an i9 12900K CPU. Why would anyone settle for anything less than multi-core gaming in this day & age when investing so much money in components?

If you're reading this Wags tell the Devs at ED to pull their finger out & to get that Vulkan implementation prioritised!  People don't buy supercars to drive them on single-track bumpy lanes! 

In Norway we do!

Sent from my MAR-LX1A using Tapatalk

1 hour ago, Infiltrator said:

Multi-core processors have been out for 17+ years and today's modern games can utilise 6+ cores. ED can produce as much content as they like & release all of those stunning trailers. But if they don't implement Vulkan I'm gone. Some people have spent £2,500 on a GPU, and for what? A game that needs a lot of fine tuning in order to run 45 frames in VR? 

I was once into Arma 3. That game was release almost a decade ago & still Bohemian Interactive's flogging that dead horse with their single core engine by continuing to release sticking plaster content. I fear ED will follow their lead in continuing to polish the game & add more content on a outdated engine in order to try to keep their audience & to "keep the lights on" as so many like to mention.

I've literally just spent £550 on an i9 12900K CPU. Why would anyone settle for anything less than multi-core gaming in this day & age when investing so much money in components?

If you're reading this Wags tell the Devs at ED to pull their finger out & to get that Vulkan implementation prioritised!  People don't buy supercars to drive them on single-track bumpy lanes! 

 

Vulkan and Multicore have their own teams.

Rewriting a Graphics Engine from scratch takes time, as well as rewriting a core engine from a single serialized thread to parallel threads.

DCS is a Diverse Simulation, it's not call of duty, everyone's system needs tweaking to find the balance of performance and visual definition that they are comfortable with.

The main draw of 6, 8, 12, 24 core CPUs wasn't to force everyone to create parallel coded multi-threaded simulations ASAP, it was to allow a User to use multiple high demand instances of Software at the same time.

By all means if we go back to 2002 with a Single Core Athlon 64 and try to run a game, record and compress the video footage, while live streaming in 1080p to Twitch and YT, the system would have a stroke.

At the risk of sounding like a bad guy, gamer's today are spoiled compared to 90's gamers, I literally had to create a boot disk for nearly every high end sim just to free up enough memory to run the sims, but each sim had different drivers and memory requirements. even after games moved from DOS to Win95 it was the same routine, had to create a custom script to kill extra processes in windows.

8 minutes ago, SkateZilla said:

If you're reading this Wags tell the Devs at ED to pull their finger out & to get that Vulkan implementation prioritised!  People don't buy supercars to drive them on single-track bumpy lanes! 

I love the Supercar reference, so so true

16 hours ago, rcjonessnp175 said:

Careless about Vulcan, multi threading is what’s needed and long overdue.

when you realize how much of the CPU is being used to process DirectX draw calls you'll change your mind.

1 hour ago, Infiltrator said:

… I fear ED will follow their lead in continuing to polish the game & add more content on a outdated engine…

There, exactly and precisely there, you lost it . Do you realize this is a thread talking about a new (another one, I can't remember how many I've known in DCS) game engine and multithreading the game, right? It's happening, they are working on it since several years, hence they are well aware of it (since several years ago, if not more) mate. Fear not .

7 minutes ago, SkateZilla said:

when you realize how much of the CPU is being used to process DirectX draw calls you'll change your mind.

Sorry I don't, as its been shown with other "games"  and "Sims"  I know its not the messiah, Vulcan that is.  I just want and look forward to our "Combat" sim to be multithreaded so we can finally have Strategic/Theatre  wide Ai that Fills in the "World"  of this Digital Combat Sim...  I'm not a API fan boy,  but what ever the actual on the ground answer is, aka Multi-Threading which will get us to that Dynamic Campaign world we all been dreaming about in Blackshark and now DCS.  Hope it Finally comes this year.  In my Opinion the graphics are as good as they need to be, we need the AI behind the curtain to be able to run at a full theatre wide scenario and not tank.  Some Brilliant Bought and paid for campaign's not flyable in VR, let alone in 2d in some cases will be much enjoyed once this happens.  Multi-Core is being thrown around allot lately on Discord chats from ED folks, so hopefully its a sign its near release.

 Look forward to Multi-threading

literally had to create a boot disk for nearly every high end sim just to free up enough memory to run the sims, but each sim had different drivers and memory requirements.

On 4/26/2022 at 4:54 PM, SkateZilla said:

Vulkan and Multicore have their own teams.

Rewriting a Graphics Engine from scratch takes time, as well as rewriting a core engine from a single serialized thread to parallel threads.

DCS is a Diverse Simulation, it's not call of duty, everyone's system needs tweaking to find the balance of performance and visual definition that they are comfortable with.

The main draw of 6, 8, 12, 24 core CPUs wasn't to force everyone to create parallel coded multi-threaded simulations ASAP, it was to allow a User to use multiple high demand instances of Software at the same time.

By all means if we go back to 2002 with a Single Core Athlon 64 and try to run a game, record and compress the video footage, while live streaming in 1080p to Twitch and YT, the system would have a stroke.

At the risk of sounding like a bad guy, gamer's today are spoiled compared to 90's gamers, I literally had to create a boot disk for nearly every high end sim just to free up enough memory to run the sims, but each sim had different drivers and memory requirements. even after games moved from DOS to Win95 it was the same routine, had to create a custom script to kill extra processes in windows.

 

This is what I played my first PC game on.

On 4/26/2022 at 3:43 PM, Infiltrator said:

Some people have spent £2,500 on a GPU, and for what?

Dunno, the game is CPU limited. Some of you people have money burning a hole in your pockets.

On 4/27/2022 at 10:21 PM, 98abaile said:

Dunno, the game is CPU limited. Some of you people have money burning a hole in your pockets.

We're just PC enthusiasts who want the most immersive flight simulation experience possible.  I personally got my 3090 for MSRP. I wouldn't have paid retail for it. I managed to beat the scalper's bots. If you're a VR user like myself, getting a 3090 is worth it for the eye candy. The 24GB of RAM also eliminates VRAM bottlenecking.

Efficient & minimal mission creation seems to be what is handling the CPU limitation ATM. My only concern is ED will keep delaying the implementation of Vulkan & instead keep creating more content & features as a means of diverting people's attention from the elephant in the room. 

Finance is obviously an issue as ED have to pay their bills & wages & Vulkan isn't paid DLC which in my mind keeps it at the bottom of their priority list. ED's priority is to create content & modules to keep the money rolling in, which is understandable as they have to 'keep the lights on'.

I'm just concerned Vulkan will keep being pushed back. I know little to nothing about coding. But from what I gather going from one API to another is an enormous undertaking which requires many thousands of manhours of work.

Vulkan is now 6 years old & we've been discussing Vulkan's implementation into DCS for over 4 years so it's understandable that some people like myself who have spend a lot of money on gaming rigs are now beginning to become impatient. 

With Sims like Star Citizen Chris Roberts knows he has to release an abundance of news with Star Citizen's development because it keeps its player base happy with at least knowing what's going on behind the scenes.  Sure, things are pushed back, but at least people are kept informed.  With ED all we get is a newsletter & a newsletter that hasn't even mentioned Vulkan for a while now.

An ETA with some development news would be nice & reassuring instead of a blanket of what seems to be secrecy with a: "don't worry it's coming" to boot. I love DCS in all its splendor, but there's only so much one can take when shelling out money on components in anticipation for what seems to be an imminent implementation of the new API that will unlock people's hardware's potential.

I now find myself digging deeper into keeping myself entertained when playing DCS. Meaning I'm constantly flying my favourite module & taking on harder missions to keep myself distracted from the frustration of not knowing when.