Game Dev Explained: Why Games don't use all your CPU cores.
HTML-код
- Опубликовано: 20 авг 2020
- In this new video series I try to answer questions about game development for non game developers. It's by no means a deep dive, rather then a high level explanation of why things are like they are in game making.
Say hi on my twitch: www.twitch.tv/brainshack
It's all about the game engines, usually they are not ready and even have limitations preventing such optimizations from working. Devs don't go on low-level programming, that's why the engines exists. We need to wait to the engines to automatically makes more use of the threads, as standard is NOT up for the game devs to do that, making game engines are completely different than using them to create games. Of course you can modify the engines or use some function to do that, but any messing around with threads are always a headache (trust me), it easily leads to errors, that's why no game dev will mess with that if they don't need to, so in the end it's up to the engines devs to optimize the engines.
There is another issue too, nobody wants to take that job to optimize since the majority of gamers doesn't have more than 4 cores (look at Steam Survey), a very few percentage have 6 or more cores, so there is no point on optimizing it if in the end it must play perfectly with 4 cores. I believe that's is the major point, they will only optimize it once the majority of players has more cores on their game machine.
But that's no reason to be concerned, as the entry-level CPUs begins to have more threads, these engines will be updated, up to recently entry-level CPUs has 4 threads, it was just NOW that 10th gen intel and Ryzen 3000 began shipping 4 cores / 8 threads CPUs on their i3/r3. Which means now cheap CPUs has up to 8 threads available, so we will start to see better scaling up to 8 threads (today usually is 4 or 6)
I have 64 cores and a 128 threads so basically I might aswell take a pair of tweezers to it and leave 1 core in and increase the ghz to 20,000
Excellent thanks you
a PERFECT example of optimization is teardown btw, that game fucking runs smoothly on my shitass 4 core 4ghz, 1050 with 4gb shared mem and 8gb of ram laptop 😎altho i gotta set the res to 75% and low settings, but still looks great and runs most importantly smooth as hell as i said 😎rly impressive ngl. and i think the reason for that is bcuz the engine and the game was made by 2 ppl, so they had lots of freedom on tweaking and creating the engine to their liking obv
No devs are just to lazy to do it instead they rather just give you a cash grab game knowing the sheeple will buy it up.
i am a dev (not a game dev though), that makes no sense, is just that such optimization requires a lot of work, and actually if you are not into an big company it just wouldn't worth doing it, the effort needed may be greater than just developing the game itself... it doesn't make sense to do that. Also devs usually don't go on low-leveling programming, that's the engine to do automatically, i assure you no dev will take that time unless the engines themselves help with it.
@@rangelebert3049 I have to disagree on this. First of all, there are a significant bunch of cashgrabbers nowadays especially since Early Access become the norm. Second of all, "The effort needed may be greater than just developing the game itself." The effort may be huge however when the engines have better support for multicores you can create way more sofisticated games, alot better than what we have today. More cores is the future, we just need a backbone in the games that support it.
For example simulators like racing sims. Their biggest holdback is physics calculations that more or less has to be done on a single core. If the engines would make multicore better supported simulators would benefit greatly from this.
Imagine how easy it would be for AMD/Intel to come up with a new lineup of cpus. All they would have to do is add more cores and boohm, performance goes through the roof.
@@alicedad6171 More than that, the used market would instantly explode. There are tons of old xeons with great core counts floating around out there from outdated industrial servers. And whole slews of multi cpu motherboards for them. If the software standard was scaleable to your system's core count, by having support baked into engines rather than added by individual programs, then you could put together a top quality gaming system for almost nothing, using chips that haven't seen relevant use for a decade. E waste cut, massive spike of technology-enabled working class around the world, and, most critically, less profits for silicon. No more market bifurcation along server and consumer lines, no more product hyperspecification, no more half-market artifical scarcity, no more hostage consumer base to planned obsolescence, no more artificial cap on consumer performance. That last one's really important, cause of the value, and capitalist secrets, there are in big data crunching. Imagine if everyone on earth could monitor a massive chunk of the web on their home computer, just like corporations do? a small group of 100 people or less, using bargain bin pcs could learn and reveal *everything* that companies are doing with our data.
so it's not gonna go that way any time soon. artificial processing cap will stay on, unless you're rich (capitalist) enough to afford something like threadripper.
btw, the same is true for why no-one makes multi gpu work right.
Yep,
They are just lazy and fat,
That's is why they are unsuccessful, not only at their job, but also their personal lives.
Battlebit is on unity and does multi threading well,
The developer is not fat and lazy...(he is fit and has learned martial arts growing up, which echoes through his life as discipline)
Coincidence? I think not
Oh that explains why RaceRoom is terrible on my 8 core 16 thread CPU , Plays much better on my 3770 ,, iracing plays better on the new PC 😊