This game is very single thread locked in regards to performance. There is 1 master thread that controls everything in the game, and that IS the bottleneck and why DX12 will not do anything for the game in regards to performance.
This is not a graphical API issue, its not a worker thread issue, its the main control thread. And this was explained on Reddit some months back by an Anet Game Engine Dev.
yes single thread is dx 9.
NO, the engine uses a single thread to prepare information to be passed on to the GPU by dx9, where it is calcualted by the GPU’s threads
dx 12 is multi core
NO dx12 is a programextension that can help implement I/O and basic OS capabilities for easier integration of 3rd party progrmas on an OS, as such it handles the ways information can be send to CPU core’s, GPU, audio, HID and more.
It works as a library so the game writers can write a progrma which can be played on any PC with sufficient capability, ( when the hardware compatibility is ok).
All core can talk to each other and send the job more fast to the gpu.
- Yes this is called hypertransport, and is an exchange mechanism so 1 processor can use data from another core or CPU. it’s completely dx12 independant and part of the CPU infrastucture(as it is made by the CPU manufacturer and needs to be implemented on motherborads for CPU to CPU communication).
- The speed to GPU will not increase, speed is governed by bandwidth whcih is governed by card bitwidth, bus link speeds and bus width. The datatransfer to the GPU stays the same. as DX 12 doesn’t change the maximum transferrates over the bus. It could however improve multithreaded tasks, and relieve the CPU bottleneck. But only if the game engine allows for this. Which in the case of GW2 isn’t so.
When using dx12 also the dev can have a more and deeper control.
The dev makes a program and is already in full control over the program, He’s allowed more resources; Yes. But this is not really relevant on an older software platform or game engine.
On the part that was control by the api.
Dx 12 will make more fast the game that are single threaded have lots of detail lots of complexity and that are cpu bound or cpu heavy.
NO, -single- threaded engine have no benefit whatsoever from DX12, these engines must be completely rewritten to be able to use multicore for the graphis rendering thread
but the important factor with dx 12 is that it all depends on the gaming devs job the implementation is in their hand to code the game and send the job through all the core to the gpu. handle the memory etc…
Yes and if a program’s engine cannot accept multithreading you can implement dx12 without rebuilding an engine from scratch, but it would not improve the performance. The engine is the communication between the program and the directx components whcih are handling the acces to the hardware responsible for building the frames you see on your PC. If it’s single threaded, it is single threaded, even though dx12 gives it acces to as many threads as the PC has available.
to help you understand if you did not read all the article that talk about cpu heavy game the problem is that there is to much computing on the cpu using 1 thread of a core with over head. (= game engine limit, and not easily changed)
the cpu fill up to quick and gets to 100 percent and the gpu is waiting at about 30 or 40 percent to receive the job. *(Correct, this is the symptom)*
difference with dx 12 use all core all thread on the system.
Problem is: Even if directx 12 could be accessed the game engine is only capable of filling 1 rendering thread. This would make the program run on directx 12 but still only acces the 1 rendering thread, as the programs limitation being the game(-engine)) cannot distribute data to more cores….
over head is remove the api control is given to the game dev for them to implement those interaction the same as when a game dev code for console gaming.
Which is very useful when you are writing a new game engine. but with a 15 years old engine this causes some problems.
right now pc have high end graphic card and multicore and thread cpu a lot more powerful then console but because of the coding it is equivalent with console because console coding is done better they know for what hardware it is use and do not have to let the api run that part of the software for them like in dx 9 dx 10 dx 11. when dev do a good job of coding in dx 12 you will get all those gain.
Again the engine, so the original program isn’t capable of distributing the data to more cores.
even if you come back with the main single thread control over and over it is normal the game is single thread dx 9 api. it is normal in those day it was the cpu ghz that gave the speed to the system and the new graphic card api is the bridge or the cop that handle the trafficic between cpu and gpu and memory allocation. now ghz as reach the limit and pc use multiple core and thread to gain speed it is like splitting a job so that it gets done more fast between the core and thread and they can all talk together and the dev decide where they send the job with coding.
Yes this is the problem now… but 15 years ago when this engine was coded for dx8 originally this wasn’t a problem.
hope this help you. because I have seen that line about the main control thread is master thread is single thread pass by often enuff. it is normal in those time there was only 1 thread. so you would all code the software to go by one main thread. dx 11 support multiple thread but still use only 1 thread. dx 12 use all thread they can talk to each other to do the job and it is all in the hands of the game dev.
You still expect people to remake the whole engine. writing an average engine takes 2-5 years, for a dedicated company
so that single thread can be split by the dev to use different thread. and dx 12 can solve the cpu bound heavy that the game suffer from. that is performance that it bring to the game if the game dev do a good job.
No it cannot, without rewriting the whole game engine
hope it answer that part. that many do not seams to get.
It seems there are many seams, and I hope I have answered the questions which arise
- You have the program, in this case a game Game (gw2.exe) This controls acces to the game engine. security, account acces, acces to the BLTP and so on, and allows for acces to all image, sound and other datalibraries used ingame
- You have the Game engine (Unreal modified engine, with Havok Physics engine) this controls the build up of the world, the placement of objects and the management of objects in the world, this also places the objects and tells the GPU skins and options and how the game world should be build up, You are interacting ingame with the engine. This also places the information in a stream to the GPU, directX 9 does this in order, DX12 does this parallel, while retaining order. problem is if the engine is written for dx9 it cannot send it to different threads. The different processes can be different threads though, (read the program, the engine and I/O’s). The engine is started by the program but is a program in it’s own right. As are the networking services and other I/O’s (including the database(s) with sprites, textures and skins)
- You have the API layer (directX,Vulkan, OpenGL) this allows the input and output to all hardware without so implementation of the game can be simplified. This will take over some hardware control from the OS for direct acces. This transfers data to the hardware and allows communications with the engine. so you type and direct X sends this info to the engine directx allows the use of any keyboard or mouse, or GPU, or CPU audio, I/O and so on.
- You have the OS which handles I/O and runs the software. it is the platform the program runs on which handles the engine and the platform whcih allows for the API to run so this works together with API. This is also why directX versions are often coupled with OS-es nowadays.
- You have the hardware which is the physical limits to the games capabilities. (RAM, GPU type, CPU’s , audio, LAN, HID’s) which are unaccessible without an OS
https://en.wikipedia.org/wiki/Game_engine
The speed to GPU will not increase, speed is governed by bandwidth whcih is governed by card bitwidth, bus link speeds and bus width. The datatransfer to the GPU stays the same. as DX 12 doesn’t change the maximum transferrates over the bus. It could however improve multithreaded tasks, and relieve the CPU bottleneck. But only if the game engine allows for this. Which in the case of GW2 isn’t so.*_
it gain speed because of multithreading and the over head being remove also since the devs will make better code to replace the election of the api that need to validate and control the load balancing will also save cpu power and process. since the code will do that job. in dx 9 dx 10 dx 11 devs have no access to that. dx 12 bring direct control of that level.(The primary feature highlight for the new release of DirectX was the introduction of advanced low-level programming APIs for Direct3D 12 which can reduce driver overhead. Developers are now able to implement their own command lists and buffers to the GPU, allowing for more efficient resource utilisation through parallel computation. Lead developer Max McMullen, stated that the main goal of Direct3D 12 is to achieve “console-level efficiency on phone, tablet and PC”)(and what tell you that gw2 does not? I have not read any where that devs have try to optimize gw 2 to dx 12 or try to do something similar. any source for this?
Even if directx 12 could be accessed the game engine is only capable of filling 1 rendering thread. This would make the program run on directx 12 but still only acces the 1 rendering thread, as the programs limitation being the game(-engine)) cannot distribute data to more cores….*_
what is needed for the engine to be able to use more then one thread or core? is it a coding language limitation or game engine structure limitation(like it as only one x to use one thread and you would need to add more X to the engine to use more thread? game engine is a program a program you can add stuff to it I guess since you said that it is the same then gw 1 but modified if they modified it once to add what they needed they could probably do it again.
- You have the API layer (directX,Vulkan, OpenGL) this allows the input and output to all hardware without so implementation of the game can be simplified. This will take over some hardware control from the OS for direct acces. This transfers data to the hardware and allows communications with the engine. so you type and direct X sends this info to the engine directx allows the use of any keyboard or mouse, or GPU, or CPU audio, I/O and so on.
yes you are right it is simplified so that the dev do not see what is happening behind the api door this is changing in dx 12. dev have more control. yes often devs do not want to play with hardware. dx handle the hardware part. depending on what program language they use sometime devs will also have abstraction layer keeping them in the dark on the process happening under the layer. also some coding language have some limitation.
but the fact was that yes it is normal that the thread is single since it use dx 9 and it was in use when cpu where single core.(at that time you could not tell the data to go to another core there was only one.) and yes dx 12 is multicore(meaning you can send data using coding to multiple core to do more job at once meaning a speed gain.) you can use more then 1 core and it can solve the bottle neck of cpu heavy software. saying no and that the game engine is not able to use more thread in its current state (using only one thread)with dx 12 does not means that dx 12 does not support it and that it does not bring this improvement. it just means that the game engine would need a upgrade to be able to use multiple core.
we all ready know that we are talking about single core game that need a upgrade and that can benefit of dx 12 improvement. Dx 12 will make more fast the game that are single threaded have lots of detail lots of complexity and that are cpu bound or cpu heavy.
_*NO, -single- threaded engine have no benefit whatsoever from DX12, these engines must be completely rewritten to be able to use multicore for the graphis rendering thread
so we agree that with single core game that have no optimization will not gain the improvement but if they have the optimization it will solve the bottle neck issue.
we are saying the same thing it is just that we understand it differently. most of the no becomes yes if the engine is optimized. we could spend day adding stuff that say the same thing and the issue would always be that people do not understand what is said. and some complain that we talk to much about it. yet many repeat the same thing and often do not understand what they are saying or talking about or they would not copy paste things written like this (that IS the bottleneck and why DX12 will not do anything for the game in regards to performance.) with out more explanation about it. since read like that this is absolutely false.
The key takeaway from all of this is section 2, “Parallel Execution State”. Designing systems for functional decomposition, coupled with data decomposition will deliver a good amount of parallelization and will also ensure scalability with future processors with an even larger amount of cores. Remember to use the state manager along with the messaging mechanism to keep all data in sync with only minimal synchronization overhead.
The observer design pattern is a function of the messaging mechanism and some time should be spent learning it so that the most efficient design possible can be implemented to address the needs of your engine. After all, it is the mechanism of communication between the different systems to synchronize all shared data.
Tasking plays an important role in proper load balancing. Following the tips in Appendix D will help you create an efficient task manager for your engine.
As you can see, designing a highly parallel engine is manageable by using clearly defined messaging and structure. Properly building parallelism into your game engine will give it significant performance gains on modern and all future processors.
https://software.intel.com/en-us/articles/designing-the-framework-of-a-parallel-game-engine/
also example: with havoc engine.
