Intel Thread Director: Lunar Lake Optimizations Explained | Talking Tech | Intel Technology
Вставка
- Опубліковано 3 лип 2024
- Intel Thread Director technology works with the OS in real-time to recommend the most performant cores for a given workload. In this episode of Talking Tech, we sit down with Intel Fellow Rajshree Chabukswar to talk about optimizations in upcoming Intel Core Ultra processors that make that data more granular, and those recommendations more consistent with platform performance and power expectations. The inverview also covers how workload classification has evolved to take full advantage of Lunar Lake’s highly capable low power island.
ITT Live Session - The Magic of Intel Thread Director Technology: • Architecture All Acces...
Learn more about Lunar Lake in our Talking Tech overview: • Lunar Lake Overview: I...
Chapters:
00:00 - Intro
01:12 - What is Thread Director?
03:47 - What's new for Lunar Lake?
06:29 - Scheduling evolution
09:09 - Thermal and power management
10:19 - Classifying AI workloads
12:01 - Containment Zones
15:25 - Consuming platform intent
18:20 - Balancing performance and power
20:37 - The future of Thread Director
21:48 - What is the low power island?
23:31 - Conclusion
About Intel Technology: Intel has always been at the forefront of developing exciting new technology for business and consumers including emerging technologies, data center servers, business transformation, memory and storage, security, and graphics. The Intel Technology UA-cam channel is a place to learn tips and tricks, get the latest news, and watch product demos from both Intel and our many partners across multiple fields.
Connect with Intel Technology:
Visit Intel Technologies WEBSITE: intel.ly/IntelTechnologies
Follow Intel Technology on TWITTER: intel.ly/3MmhqET - Наука та технологія
Congratulations Rajshree and team!
I love the effort Microsoft and Intel have put together in partnership to make Thread Director work with OS scheduling. Having hybrid hardware/software scheduler is smart engineering. I find this more impressive from an engineering perspective than the partnership of Qualcomm and Microsoft in PRISM on Windows on ARM.
I love the direction we're seeing with Microsoft pushing Windows OS priority on efficiency with SOC design. Starting with efficiency then scaling up for performance is a better strategy.
Lunar Lake is impressive and it has been a long time since I could say that about Intel
Amazing progress.
Thread director is gonna lift the already powerful and efficient Lunar lake to new heights.
maybe you try in practice before speak
@@petrm2234 Exactly. There is some useful information in here for sure. It's not just pure marketing which is good. But the only thing that matters is how it works in practice.
@@petrm2234we haven’t tried any of the upcoming chips, does that mean we are just supposed to not speak about them at all
does the thread director also do the core parking?
no he employs a valet to park his cars, theres no point being a director if you have to park your own car
Intel will optimize service, hardware, software... to bring the best experience to customers, sounds good.
so unlock bclk and igpu overclocking again like it was before, more customizing
User-mode scheduling (UMS) and apps which schedule their own threads/fibers or use Isolated User Mode / Thread local storage (TLS) (maybe thread pool API), is sounding as if this is about to require some level of recoding and optimization to support handling in process. Recompiling a few gazillion apps? I'm not so sure I like the sound of the implications!
i would add that intel and microsoft having between them an outsized role in code optimization at a fine-grained level...... well lets just say i really hope my software doesnt end up less optimized than my competitors, or less accurately profiled wrt to strengths and weaknesses of the various core types. thats the sort of thing im watching for in this rollout
why is this not out by july
This is a mood lifter!
Intel should do a monolithic power management and thread director software package that handles EVERYTHING. Leaving too much to OEMs makes no sense with algorithms getting more advanced.
Instead the whole software should be open source, so the industry, even AMD can standardize and iterate around these ideas, so that windows actually works.
Doing a fragmented approach won't cut it against apple, as the api boundaries will become a limiting factor as chip architechtures and paradigms change faster and faster.
C-States and the Thread Director, are built into the processor as "Microcode". These are not something which can be put on Github and share as "Open Source". The reason this is being brought forward, is major changes in Lunar Lake to the threading model, handled by physical cores, opposed to SMT as it has been done in mostly all previous generations.
in a sense i would that they do through their control over microcode + the base bios/eocode + the base bios/uefi image + the silicon itself
the oem or the bios vendor has the ability primarily to choose between the options that intel presents to them
however, the solution as a whole presented by x86 is incredibly complex, i think its hard to argue that one more complication in intel puts them in a different place than AMD .... but when you compare x86 to apple's ecosystem and try and project the cost of qualification and verification - not to mention potential impacts on release schedules - this is where the risk resides for this type of approach. that said there are also significant advantages here and i think its fair to say that no one outside of intel is in a good position to weigh the two
It just doesn't work
“Maybe you try in practice before speak” -your words