Are games ever going to scale with multicore?

If you can shit out a piece of software that is capable of interpreting and properly parallelizing any workload in real-time without any developer intervention, you won't need games anymore with the income you'll be pulling.

>Is multithreading really that hard
when you deal with a lot data that is shared between threads that is also rapidly changing, yes it's really hard and complex. new programming languages are starting to make it easier by implementing safer ways to write concurrency.

>good made by decent developers
they aren't good, nor are their developers decent

their programmers are bottom barrel who don't understand task oriented multithreading, they're stuck in the paradigm of one thread for rendering, one thread for game simulation, and one thread for I/O streaming content from disk into RAM. none of the game development schools go into depth on parallel algorithms or concurrent programming. management and other non-programmers who make decisions do not understand concurrency either, they think investing time into doing it right is a waste.

that's not how it works.

also you can upgrade engines to support multicore so activision/valve/epic really have no excuse.

battlefield3 didn't scale with multicores but bf4 did at launch but it made it crash massively so they took it out then right when BF1 was about to come out they worked it out and put it back in to BF4 I think.

frostbite is a 10year old engine from 2008 and they added multicore scaling to it in like early 2016 to bf4 and bf1 beta so that's like 7years after it was originally made..

so valve could totally add it to source or epic to unreal or blizzard to overwatch .. but imo they just cant be fucked paying the massive salary these specialized guys demand like they literally can earn huge amounts more working in other industries adobe top computer science developers earn literally like 1million a year while at even top paying gaming job would be like 160k I honestly think that's why they shill their studios as inclusive and progressive because they know these keno programers are throwing like 800k away a year working as a game dev so they try to sweeten the deal saying they might meet a cute trap at the office.

makes sense when you think about it EA fired the guy that said if you didn't like the SJW shit in bf5 don't buy it but then had a internal event months later just for the staff bragging about the SJW shit internally like the sales didn't matter. that shit is staff retention imo and all the SJW stuff in games lately is just marketing retards taking the internal politics too seriously and pushing it on fans. obviously they know that stuff wont fly with the masses and most of the hate is based on internal tweets and office tour videos that are for prospective staff not random 'gaymers"

>Is multithreading really that hard
Yes. Parallel/concurrent computations are pretty damn hard. Basically, the only general scalable framework we have is the actor model, but that's because its usecase is very different from any sort of number crunching or real-time constrained reacting.

What do you suggest, some sort of cooperative multitasking?

Do you run on sentences in real life too?

just dedicate a core to each NPC/wall/shrub

John carmak said in a lecture that's on YouTube in 2017 at some Texas uni that mutlicore game programming is impossible to do with out adding considerable latency I wonder how much more bf5 lags on 8core than 2core the 100 more fps you get on 8core would make up for it but if you locked the fps and measured it it would be big dif I think

In johns diagram he implied every next core you add would add like 6ms assuming it was to scale to his other drawings

The added lag is probably why gears 4 and bf5 do it and others don't thous games are already laggy mess so you don't notice

NT's scheduler sucks donkey dick at multicore. Android, Mac, iShart, and GNU/Linux all have much saner internals.

No, games are synchronized programs. An OS doesnt care if your browser loads faster than it renders a frame of video, a game does.
You can use multiple cores but the speedup is smalls since it all has to sync. Best it can do it let you do more things at the same speed, not fewer things faster.

I wish the i3 was called the i2 actually I wish they still made single core ... 8th gen i3 have 4 core now thou its not fair

>didn't mention the best

Attached: download.png (225x225, 1K)

It's a software problem.

As long as games keep being written in C++ we won't see massively parallel engines.
While multi threading is possible and in fact done commonly...on some level; it's just still too difficult to do it right in c++

>will they ever scale...?
Just something to consider: Right now 28% is running a dual-core, 58% is running a quad-core.
store.steampowered.com/hwsurvey/cpus/

It takes years to develop games. I'm guessing people who are looking at these numbers and starting development of some game will want to make sure it can take advantage of four cores. Multi-core will improve but it'll take time.

That's not how multicore programing works it has it or it doesn't you don't make it for specific core counts

>Is multithreading really that hard?
Yes. At least if you have to manually handle multithreading. All global variables are shared between threads.
Think of it with a simple example:
You define X as a global int.
In the first thread you have:
x = 1;
x = x + 2;
print(x);

In the second thread you have:
x = 10;
x = x * 4;
print(x);

So, you would expect to have as outputs 3 and 40 only, but in reality after some iterations of the program, you will have outputs like 12, 4, 6, etc. because of how the program (therefore, the threads) are stopped using the cpu (asuming there is one cpu) and when they are called back, maybe another thread replaces the one which was working, so it continues where it was before, for example, using x = 10 and doing x = x + 2.

Wouldn't a state table fix this?

It may have been touched upon in this thread, but just to lay it out flat, for something to be able to be processed efficiently in parallel, the workload itself has to lend itself well to that type of processing, which games do not (for the most part).
The reason why Ashes has been developed as an RTS as the perfect storm for multithreading, is because it’s largely the only game type that can really leverage more cores. You have hundreds of AI actors that all behave semi-autonomously. Most other games are not proactive in the amount of processing they have to do, that is if the player isn’t applying any input, the engine isn’t really doing anything.
And while there are highly-parallel tasks that are often implemented in games (physics, particle systems, ray casting) many of these are more efficiently done on the GPU anyway.

There's more.
Say, you have two globals a = 0 and b = 0.
Thread 1:
a = 1;
print(b);

Thread 2:
b = 1;
print(a);

Even on x86, an architecture with very strong coherency guarantees, it's legal for result to be
0 0
This isn't a problem at all if you use mutexes and barriers, but it puts things in perspective.

Do you have a opinion on the latency it would add ?

We are literally talking about dice you retard thou AFIK the engine development isn't done by dice any more but a EA studio some place els. Probably the only reason bf1-5 and 4 retroactively got multucore support is some EA uber programmer updated it for FIFA I don't think BF even has the market share to justify the cost of that work and when dice tried to add it them self they fucked it up at BF4 launch and had to remove it in a patch wasn't till like 20months later it was readded

>but what about games that are actuall fun and good made by decent developers like Blizzard Valve Bethesda Epic Bluehole and shit?
Bait.

Blizzard is a joke since their recent releases, Bethesda can't make games (It still amazes me how much manpower and effort you can put in something and get such awful results), Epic is basically irrelevant except for their engine and some Zoomer Garbage, the same Zoomer Garbage Bluehole produces.

ignore the names I just mean normal devs. not ones with EA/Microsoft/sony money.

>I just mean normal devs. not ones with EA/Microsoft/sony money.
So you don't mean Blizzard Bethesda Epic, who all have ALOT of money behind them

>games that are actuall fun and good made by decent developers like Blizzard Valve Bethesda
AHAHAHHHHHHHHHHAAHHAHAHAHAHAHAHAHAHAHAHAHAHHHAHH

read the fucking thread in full and come back.

(((((((((((((((Intel Core i9)))))))))))))))))))

Attached: 1445873750669.jpg (250x250, 15K)

>read the fucking thread in full
No. I was just on the toilet and needed a couple minutes of distraction.

There's not that much that can scale with multicore in videogames.
AI, draw calls, collision detection, physics, networking? How can you possibly need more than 4 cores?
I can't imagine the spaghetti code that actually utilises more than 4 threads. It's not a programming problem, it's a hardware problem. Our processors are too weak to do everything in 4 threads perfectly.

This idea games scale up to 4 and don't need more is wrong there are games that scale all the way to 8 or more or ones that perform the same at 2 there is no inbetween

If you have an 8-core CPU you could run your 4-core game, run the streaming software on top of that and open some Youtube video for background music at the same time. That's one reason to not get a 4-core CPU. Also it won't immediately choke your setup if the game suddenly learns how to use two more cores. I think 8 cores are the sweet spot in current year. With AMD it can even be 8 cores 16 threads with a reasonable amount of money. That's future proof.

that's not how windows multitasking works it doesn't just put dif programs on dif cores it gives them all to the program with the highest priority in task manager you sperg. even if you running a single threaded program background stuff can still fuck it up depending on what its doing. pleb level knowlage

That's what I'm saying, it's some sort of spaghetti code under the hood. Unless you have dedicated threads, you will be making constant synchronisations which adds latency, hence microstuttering.
I'd rather come back in 5 years with a brand new CPU and play without microstuttering.

ohk

>what are game engines

EA added multicore to frostbite 7years after it was released as a engine.. so you have no point. BF3/4 didn't have support for it (thou bf4 got a working version of it patched in when bf1 launched) frostbite was first used in 2008 on bad company 1 and it is the same engine as in bf5

Many heavy tasks, such as rendering, physics and audio are already throughly parallel. However arbitrary game logic is not necessary supposed to gain benefits from parallelization.

Attached: 640px-AmdahlsLaw.svg.png (640x500, 31K)

Just wait until the PS5 with 6 true cores. Suddenly every game will take advantage of at least 5 cores and they won't run on dual cores at all

Idiot play 120Hz plus and watch that cpu go to town. 9800X 8c 16 thread with 1080TIs in sli (16x) lanes per card make it a non micro stutter and smooth play. Try rainbow 6 siege. Black ops 4 or tomb raider

This. A lot of draw calls and frames called to the cpu from a single 1080Ti even is making my 165Hz display show me stutters because my cpu is hitting 100% at 4.5Ghz

> neither bo4 or tomraider scale with more than 2cores
> rainbow 6 siege and other Ubisoft games using same engine AC/etc don't scale past 4core and its been confirmed they are not multithreaded at all and just programmed badly

>still kike-ing this hard

>what is passing the physx workload to the cpu
>what is core unparking, setting decicated cores
>what is tweaking the bios, registry and configs to properly share the workload so the gpu has more resources to deal with the rendering

Go play your gaemes at 5 more fps and 100% cpu usage. Meanwhile, the others will get by just fine with their multicores that don't even currently hit more than 60-80% in demanding gaems.
>b-b-but muh week one benchmarks

lol buthurt 8HARDcore gamer owner.

>but what about games that are actuall fun and good made by decent developers like Blizzard Valve Bethesda Epic Bluehole and shit?

>good decent developers
>Diablo mobile, bfa shitfest, HoTs shutdown, ActivisionBlizzard

>Fallout76, letsreuseoldshitandrewrapitlmao, Bethesda

Both of those are just as bad as EA.

>ubisoft games
>don't scale with more cores
just...
>if they run better on multicores it's just programmed badly. No need to look at what's the limiting factor in this specific case.
kys

>t./v/-tard that copy pasts configs from leddit and (((overclocks))) with msi afterburner; Yet, manages to blow up his 2080ti

Can you at least try to type coherently? Jesus fuck, that was a pain to read.

read it and weep
resetera.com/threads/assassins-creed-odyssey-origins-high-cpu-usage-is-not-drm-related.74402/
confirms its not multithreaded.

you are probably some retard that brought at 16core to game on thou so who cares … we are not debating this shit im asking why more games arnt. I would like them to be but wondering what the limitation factors are.

>decent
>Bethesda
Bethesda is the most retarded and incompetent group of "developers" in the market. Just look at all the embarrassingly pathetic issues with fallout/Skyrim

>but what about games that are actuall fun and good made by decent developers like Blizzard Valve Bethesda Epic Bluehole and shit?
>that are actuall fun and good
>Blizzard Valve Bethesda Epic Bluehole and shit

Thing is, game software can be designed so it elegantly takes advantage of parallelism and multiple threads. Look at Doom and Idtech 6+ or DICE's Frostbite. The problem is most game programmers are stupid.

>decent developers like Blizzard Valve Bethesda Epic Bluehole and shit
shit makes the best games out of all of those studios

>The problem is most programmers are stupid.
ftfy

Do you have any sources to that? Doesn't match my personal experience at all. When moving from 4c to 8c I was able to open anything I wanted at the same time and game FPS was unaffected.

Exactly.

Maybe learn how to write software you dumb fuck. Too many people that either have average IQs in this thread who think they're smart or amateur developers who think their 2 years or python or C++ makes them expert. Come back after 15 years of programming in the trenches.

>blizzard
>valve
>Bethesda
>epic
>blue hole
>making any good games
Having a fucking hearty day after christmas laugh right here

>forgot he also got a faster CPU and formatted

This isn't /v/ cunt we arnt debating gameplay I obviously just mean the common mans game dev

>single task: provide sources
>won't provide sources, makes a sperg blog post
Yeah you don't sound like a logical programmer either. Are you one of those mentioned in your post who think they're smart? That would kinda make sense.

I mean I'm sure theres plenty of skilled programmers in the industry but those fucking deadlines and corporate asses diminish it and they have to cut corners.

You're a fucking idiot that's obviously never written any multi threaded software if you're too daft to have even observed multiple single-threaded processes maxing out a core each simultaneously. Window's scheduler schedules per-thread, and will run threads from different processes simultaneously.

Its not the same person replying too you retard

There are plenty of skilled programmers but they are spread thin within different teams being a rockstar and carrying their team across those deadlines.

>guy opens task manager once thinks he knows everything

No it doesn't

And i'm not any of the previous posters, but both the quoted posters were agreeing that "software doesnt work like that". Too you retard right back at you.

Attached: 129848582235.jpg (500x300, 93K)

Smaller teams using engines like unity won't scale up with more cores as unity only supports a single thread.

can you stop being stupid, thanks

Sounds like anecdotal bullshit, if a high priority process only has 1 thread it can only hog 1 core. If you have some kind of evidence the windows scheduler is doing otherwise then please post it, it would be interesting if it was a bug or actual behavior (a lot of which isnt documented). Sure other problems can come into play with large core counts, such as OS process management overheads and locks, but that is not what you're describing, and doesnt negate the fact that surplus cores means less need for context switching.

all those game devs have proven theyre EA tier garbage you bow legged sheep

Another user here.
That means that on Windows, if I make a pure number crunching program that uses 100% of CPU, then during its quantum ONLY this program can run. Therefore, if I launch, say, 4 of them, one for each core, I would not see a speedup at all, the result won't be different from a serial execution of one program after another.
Yeah, I call bullshit on that one.

Fuckloads of big studio games dont scale too thou the only ones that do are

Ea 50%+ performance per core
Stardoc 20%+ performance per core
Ubisoft 5%+ performance pee core

You can't say only indies don't know how to do it when only 3 engines frostbite anvilnext and some stardoc engine (who are indie) know how to do it.....

The fact you say things like 100% of CPU mean you don't know what you are talking about and a entry level larper

If you can't infer what I meant by that (no waiting, sleeping, or jumping to kernel-space at all, optimal use of ALUs and caches) means that you are the one being a retard, dear user.

Ah yes the age old personality test how do you fair ...

Does high CPU useage mean the program is optimised or unoptimised

Half people say it means its bad the other half say it means its made well.

Little do both parties know that task manager is about as accurate a metric as a loading bar on a program or os install and literally means nothing .

You realise Intel made its CPUs to report CPU use on other cores even when they are not being used to sell tacked on chips and amd followed suit... That info literally means nothing and is wrong I can run a program from 1998 and show 70% load across 16 cores you spergs

Also, to confirm my (and other sane user's) claim.
microsoftpressstore.com/articles/article.aspx?p=2233328&seqNum=7
>Windows implements a priority-driven, preemptive scheduling system—the highest-priority runnable (ready) thread always runs
Key word: thread. The scheduler works on thread granularity.

Lol you realised task manager had threads before dual cores even existed. Its not referring to what you think it is. And a multi threaded program doesn't show up as multiple threads in task manager it shows as a single one.

>Mm bunch of clueless chaps trying to work out complex masters degree level shit with the windows GUI because they nolife and spend lots of time in it. This should end well

Why are you so fixated upon task manager, never once did I mention it. This is about OS scheduling, not... whatever you think it is.
>task manager had threads before dual cores even existed
Yes. You can have multiple threads even with 1 physical processor and core. Please, read on what a thread is, what multitasking is and what types of multitasking exist and what types of, specifically, preemptive multitasking exist. You seem like you have no idea about anything.
But you're just trolling at this point.

You linked a article about windows task manager mate

A big article about Windows scheduling with a small section about affinity and how to set it with task manager. Read, nigga, read.

Windows thread =/= hyper threading

Sell ya 8core before Clarence steals it

Vietnam and China both don't let Intel brand their multi core CPUs as "the best ever" or fastest gaming" ...

Both communist countries really makes you think maybe they are the only places redpilled enough to stand up to their tricks

Its a country based discalmer on all Intel marketing documents really makes ya think

its other way around
>single process
>single thread
>every core is fucked
>imagine compressing files, its single thread but it fucks over windows gui, even if that is running in other thread. its not a bug, its windows feature

ur in luck pentium gold is the new core i3 ur welcome