THIS NEW Linux Scheduler BOOSTS Performance of the CPU!! (UP TO 200%)



The Nest Scheduler for the Linux Kernel has a Performance Boost of CPUs up to 200%. Let’s explore what Nest Scheduler does and how it works with the Linux Kernel to bring you these gains in hardware. This was presented at the Linux Plumbers Conference! Special Thanks to all the contributors.
______________________________________________________________
Content Creator??
https://editbulk.com
Try Free Today!

Exploring Linux??
Download a Linux Cheat Sheet and 25 Page Checklist here:
📚 https://learn.savvynik.com

Want more info/content?
https://savvynik.com
_______________________________________________________________
Discord Community:
Discord Link: https://discord.gg/zZD5q92

Useful Links:
Slides – https://lpc.events/event/16/contributions/1198/attachments/983/1909/plumbers.pdf

#cpu #linux #performance

36 Comments

  1. Unlikely. This is so simple, and not related to performance but ‘power’ it has not somehow been overlooked by millions of engineers worldwide.

    Reminds me of a child trying to explain Core scheduling without any knowledge.

  2. The only amazing thing about this video is that there are this many viewers who have even the remotest interest in a nonsensical idiotic garbage fan boi video like this.

    Boring videos like this is why youtube sucks.
    High praise in the comments section is why humanity sucks.

  3. does this apply only to relatively new architectures? i've heard this only applies for ryzen cpu's, and if i have sandy bridge xeon, will this scheduler make a difference at all?

  4. Given the appearance of Oracle, I guess this would also be very interesting for Ampere ARM CPU which Oracle invested but also mobile ARM since the big little architecture requires core migration, so taking the core hotness would surely impact the scheduling decisions by avoiding transfer to bigger cores. Thus this new scheduler might only work for SMP processors.

  5. 2x = 100% improvement, not 200%
    If you start with 100% then have a 2x improvement, you have a total of 200% which is 100% more than what you started with.

  6. The title of this video is extremely misleading.
    The performance benefit on consumer grade CPUs (with typically less than 16 cores) is not at all discussed, and it is likely that the performance difference of nest vs CFS is negligible on such CPUs.

  7. Is it 2x or 3x more performant? 100% more performance is 2x faster, so 200% more performance is 3x faster. So which is it?

  8. I'd assume this is great for Ryzen too where we have chiplets, if we are only assigning tasks to cores that have already started, it will more likely be given to a single chiplets that is already active while the other can remain in sleep state until required

  9. wow up to 200%? my pentium have only one core tho so i dont know i dont know bro

  10. Wonder how this will affect performance for some pretty crappy CPUs, such as the AMD FX and descendants.

  11. I feel this report is like saying, when we change the square tires off our car and replace them with round tires it drastically improves the gas mileage.
    How was this not seen in the last 15 years? We would expect the CPU scheduler to be some of the most efficient code on the planet, and now we find out it's something like square Wheels. At this point, let's just get an AI algorithm, to find every performance boost and energy consumption boost that's possible under a variety of tasks and CPUs.

  12. There is one detail not explored in the video and that is: Why the hell took it so long to implement this obvious idea? It's well known fact, that keeping thread on the same core improves performance because local cache contains data, which are going to be needed for future computations. Software developers had to use thread pinning to work around this deficiency of kernel scheduler.

  13. Basically linux kernel developers learned about LIFO queue and decided to practice using this data structure in OS Scheduler. That's kind of bold move, if you don't mind me saying.

  14. I'm sorry but did you dumb it down or just presented it at your level of understanding. The original presentation was already simplified as possible but this video is just useless. You actually explained nothing about how it achieves better performance and power consumption. What's the point of making such videos? The whole concept is about using the boost frequency as constantly as possible not letting resources be wasted and the energy savings on their machines was due to the fact they had high power consumption unrelated with the core usage (she mentioned high uncore frequency). So if the task is finished quicker the less energy is wasted. But only per this task. In the same period of time the CPU won't save energy due to the Nest.

Leave a Reply

© 2023 53GB