Why are game developers so shit at multithreading?

>>56270190
They just don't care enough to make it work. Plus they know most people are still gaming on dual core shit. I was almost impressed that shovel knight was ~150mb until i remembered the snes games they are trying to emulate are ~4mb.

>>56270258
Probably the higher quality sounds and bigger sprites for higher red displays.

>>56270190
That would embarrass Intel because in mutithreading FX-8350 shits on much more expensive i5's.

>>56270479

it sure does friend :^>

>>56270190
VIdeogames is were all the failures of other industries end up.

Writers, journalists, artists, coders. All people who couldnt make it in proper industries.

>>56270190
Lazy coding, most don't see the use of it.

How do I into thread pool with game development?

>>56270190
Because most of the frame time is spent on rendering and opengl/directx are garbage and can't be multithreaded.

You will see more improvements with vulkan

>>56270190
because it's hard to make a multithreaded game engine work without making it bloated as fuck
it's literally easier to just hardcode 2 threads

>>56270795
Idiot

>>56270745
does that mean that my 24 threads will be useful for gayming some day?

>>56271052
It means don't buy an i3/i5 over an 8350. The 8350 will last you a lot longer.

Things are reaching the peak of duo, so the'll start writing for quad-core soon.

>>56271081
>J-just wait another 5 years anon.

Holy fuck I just can't. I thought AMD waitfags are just a meme.

>>56271094
?? 5 years. All I'm saying is you'll have a great fall back with the extra cores. Sure if you can piss monie up the wall every year get an i5-7. But for those on a lower budget will get a longer life from an 8350, and with what they save they can buy a pre made water loop and give it the clock giving it comparable performance.

We could spout buy the known best all day, but what's more interesting to talk about is getting bang for buck, and longevity.

Sure AMD is behind the ball with nVidia and intell. but that doesn't mean you should dis-consider it as a viable option especially when most cant afford a 1080 and an i7.

At the end of the day spamming the known best products as advice is edging on being parody of payed off tech advisory.

There are already some games designed to use multiple threads. (Witcher 3 for example)
Eventually the major engines will all support it.

>>56270489
i5's are monster multithreaders in your opinion?

>>56270190
It doesn't support DX12 yet.

>>56271191
watercooling is for retards, you dumb piece of shit

File: 1471656806298.gif (2MB, 198x195px) Image search: [Google]

2MB, 198x195px

>>56271191
>Sure if you can piss monie up the wall every year get an i5-7.

I see lots of people on /g/ boasting about still using 2500k's though

>>56270190
>Why are game developers so shit at multithreading
you said it, they're game devs not game programmers. They probably used unity or gamemaker to make the last deus ex.

>>56271252
this. i jumped from a core 2 xeon quad to a 6700k, but i was even considering buying a used 2600k. i just wanted the sweet nvme stuff, otherwise it's pretty much the same performance

>>56270190
Because games are shit at being paralleled.

This is an OpenGL/D3D problem you retards.

>>56270190
The GPU utilization is very high, the game is running at 75fps, and the CPU is 50% or above on all cores. What's the issue?

>>56270190
since they are mostly intel yews.
and since we know that intel is unable to make a CPU with more than 4 cores, we wouldn't be seeing a game that actually requires more than 2 cores for the next 5+ years

1. Most game engines are built in C++, which is notoriously diffuclt to program multi threaded processes effectively
2. Most game engines offload computation to the GPU

The only games you won't bottle neck the GPU first are those shitty console ports.

File: BSM.jpg (47KB, 610x449px) Image search: [Google]

47KB, 610x449px

>>56271249
What are cars?

>>56271252
I'm on a i7 3820 still.
Don't feel the need to upgrade even the professional applications I use won't get a significant boost.
Was overclocked to 4.5 when I got it too. Went away with the overclock as well, there's no point.

GPGPU killed the CPU star.

Path of Exile recently fixed their multithreading, google it

>>56270521
isn't it much harder to become a (good) game dev?
For enterprise projects, you can just say "yeah you need one of our 80000$ systems to run it" or your work at least doesn't need to run at even 30fps. I've worked at a few enterprise companies and <1% of workers there I would trust to make even a simple game, that runs remotely smoothly

>>56270190
Because they get paid by nvidia to boost their new gpu sales

>>56272235
>GPGPU killed the CPU star
For certain workloads anyway.

I upgraded from an AMD A10-7850k to an i7-5820k

So many cores.

>>56271616
>intel is unable to make a CPU with more than 4 cores
>what are those i7s 3930K, 3960X, i7-3970X, 4930K, 4960X, 5820K, 5930K, 5960X, 6800K, 6850K, 6900K, 6950X
>being a poofag

>>56273006
even the i7-970 6 core 12 threads.

We haven't had ONLY quad cores since the Core2Quad era.

>>56270190
Not everything can be multithreaded.

>>56270255
What's the text above the thing he's kicking away saying?

>>56270258
>>56270684

What would you put onto the thread? Everything in a non-procedural game is done frame-by-frame.

>>56270190
Picture not related?

>>56273285
>retard

Because parallelization of CPU workload in a video game is not that simple, especially not in C++. Also you can't multithread everything.

The picture in OP shows excellent scaling on all cores. On my computer the game utilizes all eight threads with HT enabled and very evenly distributed.

>>56272924
This one is part of AMD's Gaming """""evolved""""" program, anon.

>>56270258
Most of that 150mb will be bloat from using some framework or engine.
If they made a bespoke engine for that game then there's something badly wrong.

>>56270190
You think multi-threading is easy?

You think that more than 5% of modern game devs are good programmers?

Ha!

>>56270479
This fucking meme, a i3 is better for gaming.

Depends on the game, games like gta5 or bf4 use most of my 12 threads on my 5820k. Generally 30-40% usage. Keeps the 1080 pegged.

I would imagine bf1 will be even more multithreaded, especially if vulcan is involved.

>>56270489
Indeed, and numerous benchmarks demonstrate that.

>>56273471
CPU benchmark and other benchmarks would like to have a talk with you.

>>56273243
>11 core CPUs

File: Beautiful multithreaded glory.png (126KB, 528x852px) Image search: [Google]

126KB, 528x852px

>>56271262

Just sayin' might wanna reconsider that thought.

DX11 Is terrible at multi threaded workloads but it is worlds better than earlier versions.

>>56273544
idiot

File: zozzle.png (28KB, 608x338px) Image search: [Google]

28KB, 608x338px

>>56270479
just look at those i5s getting shat on
>kek

File: 1471023413611.jpg (112KB, 576x1024px) Image search: [Google]

112KB, 576x1024px

>>56273691
>CineBench

>>56273544
I literally monitor all 12 threads usage in game

>>56273704
>you're not using my cherry picked benchmarks and I have no valid argument, so I'm just going to fall back on the good old fedoras

>>56273749
Proprietary benchmarks are useless as the compilation process and source cannot be easily verified.
Try PostgreSQL, Linux kernel compilation or Blender

>>56273769
are you fucking listening to yourself?

>>56273782
Pick any other libre benchmark if you wish.

>>56273666
>>56273720
Im not going to spoon feed you, if you're too stupid to even use Google, you deserve to pay more for less.

>>56273805
My fault, seen "you" and thought you quoted a different comment.

Even still, amd is for poorfags.

>>56273550
AMD 12 cores with insufficient yields

>>56273789
http://www.anandtech.com/bench/product/697?vs=1198
A few open source ones here. AMD spanked in pretty much every one by a CPU that costs the same.

>>56270258
The majority of the size there is the sound, think how big an album is, even an mp3 album. The soundtrack of the game is bigger than an album alone, now add in sound effects and you're already at over 100mb with no game

>>56273390
>The picture in OP shows excellent scaling on all cores. On my computer the game utilizes all eight threads with HT enabled and very evenly distributed.
What game is that? The stealth mechanics remind me of Splinter Cell but that doesn't fit the camera perspective

I would love to put the game's multicore scaling to the test

>>56273914
Deus Ex Mankind Divided

>>56273691
Isn't Cinebench a load of bullshit anyway? (intel-favored)

>>56273936
Ah, it's not out for Linux yet. I'll wait then

Nobody seems to test whether or not it scales to 32 cores

>>56270521
bait

>>56273967
Doubt it's ever going to see port to Linux. The game is running on DX11 right now and there's a DX12 patch coming up 5th next month. I really don't see them porting it any more than this.

The DX12 patch could reduce the CPU load, depending on what causes the considerably high CPU usage in it right now. The game has at least visually a tremendously high polycount which could relate to a large amount of draw calls, an area where DX12 would help a lot. But if it's something related to AI or other systems it's not going to help one bit. I'm running i7 4970K @ 4,9 GHz and the game uses all eight threads well over 50% in some cases.

>>56270190
What the fuck happened to your colors?

>>56273067
Most games after 2008 would benefit from multithreading

>>56274243
>Doubt it's ever going to see port to Linux.
https://www.gamingonlinux.com/articles/deus-ex-mankind-divided-looks-like-it-may-be-getting-a-linux-steamos-release.7906
https://www.phoronix.com/scan.php?page=news_item&px=Deus-Ex-MD-Linux-Possible
http://boilingsteam.com/deus-ex-mankind-divided-potentially-for-linux/

>>56274506
Alright, cool.

>>56270479
Intel pays every game developer? Good meme my friend.

Because it is a hard task. Especially when a lot of the components rely on each other

>>56277179
Every big one and engine devs, yes

>>56277239
yes my frend we got them now

They are not shit.
Pretty much everything that can be split in to separate concurrent threads is split.
Hell even most physics engines to do without the developer doing anything.

>>56270190
Look up any game in process explorer, you will see way more threads than your CPU has cores. User input is pretty much required to be in another thread. It's just many tasks are inherently non-parallel.

File: 1471459611214.png (582KB, 939x1163px) Image search: [Google]

582KB, 939x1163px

>>56270190
...cuz it was difficult until C++11 simplified it and it is still easy to fuckety-fuck it up.

Plus most americans can't read so are shit coders and the indians are too busy dreaming of raping someone on the bus home to keep their head in the game... thus...

>>56278027
I'd even say majority of tasks. Everything that follows that formula:
1. Do something
2. Do something with the results of the previous step

>>56273769
>Blender
Cycles is a pretty good example of how GPU programming is apparently more convenient for performance than multithreading

>>56270190

I don't know how multithreading works it scares me

>>56278027
Input polling is done in another thread because windows blocks your thread any time your window is being dragged.

>>56274471
Sure, but still not everything can be split up.

>>56278162
Its just fucking multitasking you scared jew

>>56278835

All my life computers start at point A and end at B

>>56273408
Falcon pls I like you but you're Linus tier and this subject is out of scope for you

>>56278058
Pretty much. If you leave a game alone it never does anything on its own, it's all in reaction to the player's input, and often even more specifically in reaction to triggers within the game.
Besides the core game logic and draw calls, the only other thing that takes major CPU time is AI, but many games don't have complex enough or a large quantity of AIs to ever push the limits. Some RTS games come close though, because you're simulating pathing for hundreds or even thousands of units.
Stuff that that people used to think multi-core processing would work great for, like physics, ray-casting and other micro-simulations are much better handled by GPU compute, so our CPUs are once again left to work on a fairly limited data set that doesn't thread well. Even if DX12/Vulkan multi-thread draw calls, it's still only part of the equation.