I'm surprised with some of these comments. I found this to be a great and informative chat. How do you follow a tech channel and find this boring? Also, english is my second language, and I understand everything she said. It's funny how english natives show themselves to be so ignorant of the world the live in.
yes being one of those english natives, I sadly agree. the intolerance for these nations is sometimes sad. Indian people are some of the educated, intelligent on the planet - specifically for coding, mathematics and other very complex areas of data manipulations.
Lots of people who follow these channels are gamers, not tech people. And most western audiences have never been exposed to different accents, or rarely, where as you, a foreign speaker, have more time hearing foreign sounds. Westerners don't do well abroad on average, as they just are insulated from the world and rarely look outside themselves.
Leo, ignore all the rude comments. This was an excellent video, and thanks to Intel and Rajshree for providing us with this information. I thought it was a great video. She is very smart and very insightful, she has a commanding understanding of what OSes need to do to schedule threads on the various cores and how the OS can benefit from hardware guided information on the state of the CPU etc. I suppose a follow up to this interview could be with Microsoft and/or Linux and/or even some software vendors (e.g., game engines) to see how they use Thread Director properly. Because even though Alder Lake has been around a while, we still hear from time to time, that certain game engines function better when E-cores are disabled and/or hyperthreading is disabled. Why is that?
Very informative. Didn't know how much logic was built in for thread management. Will be fun to see how these new ecores help power consumption and performance.
Great discussion of Thread Director and Meteor Lake. I have a 13700k and have been thrilled with the E-cores multi-thread performance and have had ZERO issues with scheduling, Intel did a great job with Thread Director.
Poor Leo. Hope you got better. I really enjoyed the interview. Was difficult making her out but she clearly knows what she is talking about. They need all the help they can get right now.
Very cool interview! intel had its own internal interview video with lots of engineers/folks throughout the company, including Rajshree, and I thought they gave great answers in that. This was neat to see going a little more off-script and a little more consumer-oriented questions rather than strictly business-side questions/answers or internal navel-gazing such as in the internal interviews (though the technical answers were in-depth and accessible enough there, too.) Seems to me intel feels proud of this direction with Meteor lake, and they have put together a product where there is a lot to say about it at once. I think it shows a lot of wisdom that, with such technical changes, they are in a large part letting the engineers speak for themselves rather than trying to flip over backwards in marketing department to spin it in outlandish ways. If the product is good and relevant to a lot of real-world needs, IMO letting the engineers speak directly was a great move to make here. (I would think that, As more of an engineer type than a marketing type myself.) In this video, I think the interviewer interrupts a bit much in the first two questions, but interviewer and interviewee quickly found a rhythm to be on the same page and get the information out that consumers/reviewers would want to know and discuss the product and questions effectively/accurately. I still have my doubts about Windows 11, personally! Maybe that's where the OS can take better advantage of thread director. The thing is, Microsoft isn't a perfect partner to showcase intel's new hardware, because Windows 11 coincidentally has more bloatware and has upped the ante for background resource usage, so regardless of CPU quality it will be a less than ideal environment to get best benchmark numbers, I think! Not intel's fault, though. I also see some edge cases where games don't know how to deal with heterogeneous core designs in CPUs, but we have this on AMD's side with the 3D-VCache as well, so again, it's not just an intel thing. We will see how game devs respond. Again, thanks for the interview!
Great stuff! Thanks for this content. I've always resented E-Cores as a cheat, but now I can see how these are actually good for performance all around. Nicely done.
ECores (even on my 12900H) are equal in performance to a true core from a 6700 ; if i'm thermally limited i'd much prefer having 2 fully performant (2x4 e-cores) 6700's than 1 thermally throttled set of 4 P cores ...
For a while I really wanted to see some fresh, not coming from Alder Lake times, informative interview about Intel's hybrid architecture, so great idea on time Kit Guru! People finding it boring, go watch Linus Tech Tips doing some "crazy stuff" ;)
Really usefull material, thanks a lot ! Something I wonder, is why there is not an internal scheduler to classify work on differents core as there in traditional CPU operations. It seems from this speach that it is a question of versatility and power management, but still OS is also running its own QOS for traditional CPU so I'm not convinced by this answer. I would like to see more about this, like deep dive into IPC of each, consumption, type of operation they can handle and comportment inside an OS.
This was a great chat. Thank you so much for sharing this. Best one I've heard yet and I think I finally understand how the Thread Director communicates w/ the OS, at least on a higher level.
I am totally confused hwo the TD take care of core group to cache locality. Is that seen from compiled association code? For compiler coding it seams to me a nightmare. As the Zen core surpass on efficiency and IPC application to architecture favor get a headache.
Very interesting glimpse into the future from 2 very informed people, The question I have is the overhead, as this prioritisation mechanism obviously consumes CPU cycles and time so what is the impact of it compared to a system running without E cores in real terms? I'm still waiting for the day where we will see the OS running on dedicated cores and apps on cores that other processes do not tread....
Would Thread Director be present in Linux, or would Linux use another CPU governor considering ARM chips and even Apple M series are part of the kernel now.
x86 can last forever, Arm has very minimal advantages. The reason Arm powered devices like Apples chips might seem like they have an advantage is because they do but not because of Arm, Apple buys the newest and best nodes. Also Intel has historically focused solely on performance (hence why they easily outperform the M2 ultra) and not tried to balance efficiency until Meteor Lake.
Not really , of a nation with 1.43 billion people there are bound to be clever amongst the populus , compare that with uk 65 million means there's a 22 times greater chance of finding the right brains for the job
@@KitGuruTech Thank you, I have another question. MTL's P core supports hyperthreading, but the E core does not. What was the reason for Intel's decision to disable hyperthreading on the E core?
@@williamstsai1 they simply do not have the hardware needed to support SMT. You require additional transistors to support it, which makes the cores bigger.
I would have thought nanoseconds or even picoseconds - milliseconds is so slow... Though if it is milliseconds, it definitely would highlight why one wouldn't want the thread to change cores very often.
a thread switch is expensive for a cpu - you have to unload and reload the state of various registers (the next instruction on the thread after resuming on any core better work), you've got various pointer stacks, memory vtables, you don't want to discard any speculative execution knowledge, etc .... there is so much thread specific info inside a modern cpu ....
Why Intel did not applying all P-Core for consumer product? It always easier than design " Thread Director " or other folks like these idea. Products with all P-Core way better.
Some tiny low power cores in the SOC for background tasks makes some sense, now try making a good architecture for the rest of the chip like AMD does instead of the current hot garbage.
you sound like you swallowed a frog Leo. I really enjoyed the interview, I am one of the few people who finds thread director very interesting however. Yes thats me.
Sounds to me it would be benefitial to have a completely different OS for AMD or intel based x86 systems. With designs so different you don't want things in your OS your CPU does not know what to do with. It only bloats the OS and opens up reasons for performance issues.
AMD needs more help from OS since they don’t have thread director to help the OS. They simply rely on OS to make judgement calls about when to use the extra tread vs increasing freq of a core. Thread director significantly reduces the OS overhead
i went into this not knowing what to expect, but considering the amount of pathetic tech content on 'channels' today, this really was very very educational. I guess if they dressed her in a low cut top and covered her in make up and put it on the thumbnail it would get more eyes. sad.
Seems a bit overkill the way they are designing these chips with different cores just for different power consumption but guess thats just the way things are now.
except that in thermally throttled workloads there has truely always been types of workload - single thread (1 core go brr) and multi-thread( as many cores as possible - efficiency is king, until you run into single thread limits)
i don't think he has to get used to anything, if she's so smart then she should learn the language she is trying to speak as well as she can learn to do her job, but she never will so you have to resort to stupid comments like that @@j340_official
Seems a bit overkill the way they are designing these chips with different cores just for different power consumption but guess thats just the way things are now.
It is all a race to longer battery life on systems that are basically idling. Only way you can compete vs ARM on x86 cores is to keep coming up with new party tricks how to shut off most of the chip most of the time.
@@Jarnis-v1cThere same party tricks (as well as using the most advanced process nodes) are what makes arm efficient. With tiered core system, the thread director, and advanced process node like 18A will make Intel x86 cores more efficient than contemporary ARM CPUs including the ones from Apple when 18A debuts. Mark my words.
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
I'm surprised with some of these comments. I found this to be a great and informative chat. How do you follow a tech channel and find this boring? Also, english is my second language, and I understand everything she said. It's funny how english natives show themselves to be so ignorant of the world the live in.
I thought it was great. learned a lot. Very educational. Great channel
yes being one of those english natives, I sadly agree. the intolerance for these nations is sometimes sad. Indian people are some of the educated, intelligent on the planet - specifically for coding, mathematics and other very complex areas of data manipulations.
Agree. I’m stunned how many people don’t want to learn about this stuff ! 😮
Lots of people who follow these channels are gamers, not tech people. And most western audiences have never been exposed to different accents, or rarely, where as you, a foreign speaker, have more time hearing foreign sounds. Westerners don't do well abroad on average, as they just are insulated from the world and rarely look outside themselves.
It's literally just one comment, everyone else is interested
Leo, ignore all the rude comments. This was an excellent video, and thanks to Intel and Rajshree for providing us with this information. I thought it was a great video. She is very smart and very insightful, she has a commanding understanding of what OSes need to do to schedule threads on the various cores and how the OS can benefit from hardware guided information on the state of the CPU etc. I suppose a follow up to this interview could be with Microsoft and/or Linux and/or even some software vendors (e.g., game engines) to see how they use Thread Director properly. Because even though Alder Lake has been around a while, we still hear from time to time, that certain game engines function better when E-cores are disabled and/or hyperthreading is disabled. Why is that?
Very informative. Didn't know how much logic was built in for thread management. Will be fun to see how these new ecores help power consumption and performance.
More of this please - with so many reviews on youtube are about extra frames on this and that games. thanks for this.
The complexity in these things is fantastic.
Informative and actually rather interesting. Glad we have Kitguru as many tech channels focus now on purely ‘entertainment’.
Great discussion of Thread Director and Meteor Lake. I have a 13700k and have been thrilled with the E-cores multi-thread performance and have had ZERO issues with scheduling, Intel did a great job with Thread Director.
U suck Rajshree Chabukswar
A very useful video. Thanks to both of you.
Whats going on inside these is very impressive in that can power manage different areas of the CPU by self power managing.
Poor Leo. Hope you got better. I really enjoyed the interview. Was difficult making her out but she clearly knows what she is talking about. They need all the help they can get right now.
Very cool interview!
intel had its own internal interview video with lots of engineers/folks throughout the company, including Rajshree, and I thought they gave great answers in that. This was neat to see going a little more off-script and a little more consumer-oriented questions rather than strictly business-side questions/answers or internal navel-gazing such as in the internal interviews (though the technical answers were in-depth and accessible enough there, too.)
Seems to me intel feels proud of this direction with Meteor lake, and they have put together a product where there is a lot to say about it at once. I think it shows a lot of wisdom that, with such technical changes, they are in a large part letting the engineers speak for themselves rather than trying to flip over backwards in marketing department to spin it in outlandish ways.
If the product is good and relevant to a lot of real-world needs, IMO letting the engineers speak directly was a great move to make here. (I would think that, As more of an engineer type than a marketing type myself.)
In this video, I think the interviewer interrupts a bit much in the first two questions, but interviewer and interviewee quickly found a rhythm to be on the same page and get the information out that consumers/reviewers would want to know and discuss the product and questions effectively/accurately.
I still have my doubts about Windows 11, personally! Maybe that's where the OS can take better advantage of thread director. The thing is, Microsoft isn't a perfect partner to showcase intel's new hardware, because Windows 11 coincidentally has more bloatware and has upped the ante for background resource usage, so regardless of CPU quality it will be a less than ideal environment to get best benchmark numbers, I think! Not intel's fault, though.
I also see some edge cases where games don't know how to deal with heterogeneous core designs in CPUs, but we have this on AMD's side with the 3D-VCache as well, so again, it's not just an intel thing. We will see how game devs respond.
Again, thanks for the interview!
The way these chips are being designed is almost adding an additional abstraction in that they are working as CPUs that decide best how to be a CPU.
Very cool interview getting information from someone who knows what they are talking about.
Great stuff! Thanks for this content. I've always resented E-Cores as a cheat, but now I can see how these are actually good for performance all around. Nicely done.
Yes. I thought the same. Changed my way of thinking.
ECores (even on my 12900H) are equal in performance to a true core from a 6700 ; if i'm thermally limited i'd much prefer having 2 fully performant (2x4 e-cores) 6700's than 1 thermally throttled set of 4 P cores ...
For a while I really wanted to see some fresh, not coming from Alder Lake times, informative interview about Intel's hybrid architecture, so great idea on time Kit Guru! People finding it boring, go watch Linus Tech Tips doing some "crazy stuff" ;)
This was a great deep dive on what's to come Leo. I really appreciate your efforts here.
its good to hear that thread management is getting smarter .........
Really usefull material, thanks a lot !
Something I wonder, is why there is not an internal scheduler to classify work on differents core as there in traditional CPU operations.
It seems from this speach that it is a question of versatility and power management, but still OS is also running its own QOS for traditional CPU so I'm not convinced by this answer.
I would like to see more about this, like deep dive into IPC of each, consumption, type of operation they can handle and comportment inside an OS.
This was a great chat. Thank you so much for sharing this. Best one I've heard yet and I think I finally understand how the Thread Director communicates w/ the OS, at least on a higher level.
Great interview/talk very informative, it's great to see such knowledge people explaining about these topics
Great interview!
Really interesting have to admit tuned out a little bit because the detail is above my understanding.
Are you able to create some content around testing thread director Leo ? I’d find it worthwhile
Same! I want to see standard scheduling vs heterogenous scheduling benchmarks, especially for thin and light laptops.
Excellent information. Now I know how the new generation Intel CPUs work & how Windows 11 makes use of it.
I am totally confused hwo the TD take care of core group to cache locality. Is that seen from compiled association code? For compiler coding it seams to me a nightmare. As the Zen core surpass on efficiency and IPC application to architecture favor get a headache.
This was such a great talk, I had some wrong assumptions. Thanks to the great Rajshree and host Leo!
Very interesting glimpse into the future from 2 very informed people, The question I have is the overhead, as this prioritisation mechanism obviously consumes CPU cycles and time so what is the impact of it compared to a system running without E cores in real terms?
I'm still waiting for the day where we will see the OS running on dedicated cores and apps on cores that other processes do not tread....
Great video and very informative! Thank you!
Glad you enjoyed it!
Wonder how much adaption the kernel of windows and linux will need to take advantage of all this goodness.
Would Thread Director be present in Linux, or would Linux use another CPU governor considering ARM chips and even Apple M series are part of the kernel now.
Quad thread cores and 2 variant efficiency cores , hybrid 3D placement of structure instead of stacked plane space
That thread director sounds very clever.
But it seems like a tool Intel as well as the Windows OS needed long before now...
thread director really is critical to the operation of modern CPU's I know the 2H windows update improved it dramatically
Hats off to Intel for keeping x86 competitive over the years, but how long can this go on?
x86 can last forever, Arm has very minimal advantages. The reason Arm powered devices like Apples chips might seem like they have an advantage is because they do but not because of Arm, Apple buys the newest and best nodes. Also Intel has historically focused solely on performance (hence why they easily outperform the M2 ultra) and not tried to balance efficiency until Meteor Lake.
Super interesting chat!
Fascinating stuff. Thank you for sharing.
The CPU market is getting very interesting indeed. I am excited to see how it goes in 2026 and beyond.
So many Indian people In coding jobs globally. It’s fascinating how they seem born with this ability !
Haha true, smart folks.
Not really , of a nation with 1.43 billion people there are bound to be clever amongst the populus , compare that with uk 65 million means there's a 22 times greater chance of finding the right brains for the job
They’re not born with it, they work hard the same as anyone else. There are a lot of people in India so the talent pool is large.
Compared to multi-threading, is the performance of the thread-director better?
They are two different technologies that work hand-in-hand. Ultimately the Operating System makes the decisions about core and thread allocation.
Leo
@@KitGuruTech Thank you, I have another question. MTL's P core supports hyperthreading, but the E core does not. What was the reason for Intel's decision to disable hyperthreading on the E core?
@@williamstsai1 they simply do not have the hardware needed to support SMT. You require additional transistors to support it, which makes the cores bigger.
I would have thought nanoseconds or even picoseconds - milliseconds is so slow... Though if it is milliseconds, it definitely would highlight why one wouldn't want the thread to change cores very often.
No. The timeslice for each thread to be run on the CPU is in the milliseconds range.
a thread switch is expensive for a cpu - you have to unload and reload the state of various registers (the next instruction on the thread after resuming on any core better work), you've got various pointer stacks, memory vtables, you don't want to discard any speculative execution knowledge, etc .... there is so much thread specific info inside a modern cpu ....
@@andytroo exactly
Interesting approach they are taking to compete with the native efficiency of the arm architecture.
All very cool but need to see these things in the flesh.
I love my arc a 770 btw we did good
Has Leo been shouting at everyone behind the scenes again?
Why Intel did not applying all P-Core for consumer product? It always easier than design " Thread Director " or other folks like these idea. Products with all P-Core way better.
Some tiny low power cores in the SOC for background tasks makes some sense, now try making a good architecture for the rest of the chip like AMD does instead of the current hot garbage.
Big.Little core idea sucks. Intel should focus its study on how to shut down some cores based on needs. GPU should be used as many Little cores.
you sound like you swallowed a frog Leo. I really enjoyed the interview, I am one of the few people who finds thread director very interesting however. Yes thats me.
And yet it still manages to be compatible with x86.
Will it run quake?
Sounds to me it would be benefitial to have a completely different OS for AMD or intel based x86 systems. With designs so different you don't want things in your OS your CPU does not know what to do with. It only bloats the OS and opens up reasons for performance issues.
I think that's what standard scheduling is, though, which runs on Intel, too. Just throw everything to the P-cores, done.
AMD needs more help from OS since they don’t have thread director to help the OS. They simply rely on OS to make judgement calls about when to use the extra tread vs increasing freq of a core. Thread director significantly reduces the OS overhead
i went into this not knowing what to expect, but considering the amount of pathetic tech content on 'channels' today, this really was very very educational. I guess if they dressed her in a low cut top and covered her in make up and put it on the thumbnail it would get more eyes. sad.
So calling it 3D is a bit of a misnomer isnt it because with 3D you immediately think of graphics not architecture in silicon.
AMD have 3D v cache which refers to the silicon architecture
1st
She didn't mention Linux... I wonder why.
Linux natively hands heterogeneous cores, its Windows that needs Thread Director to handle them. Also 95% of consumers use Windows.
Seems a bit overkill the way they are designing these chips with different cores just for different power consumption but guess thats just the way things are now.
except that in thermally throttled workloads there has truely always been types of workload - single thread (1 core go brr) and multi-thread( as many cores as possible - efficiency is king, until you run into single thread limits)
Literally the entire industry is going this route. Intel, Arm, Qualcomm, Apple, and eventually AMD.
quite a lot to unpack there, difficult to make out what she is saying sometimes - but my translation skills for that accent have never been great.
Get used to Diversity where smart people from different cultures contribute to excellent products.
i don't think he has to get used to anything, if she's so smart then she should learn the language she is trying to speak as well as she can learn to do her job, but she never will so you have to resort to stupid comments like that @@j340_official
she is certainly knowledgeable but i have to admit I zoned out after 5 minutes. sorry.
That is very boring.
its just very technical, not very easy for most people to sit and listen to I guess.
If the way tech actually works is boring then why are you perusing a Tech channel?
I need a fkn download link..
Very cool interview getting information from someone who knows what they are talking about.
Will it run quake?
Has Leo been shouting at everyone behind the scenes again?
Seems a bit overkill the way they are designing these chips with different cores just for different power consumption but guess thats just the way things are now.
It is all a race to longer battery life on systems that are basically idling. Only way you can compete vs ARM on x86 cores is to keep coming up with new party tricks how to shut off most of the chip most of the time.
@@Jarnis-v1cThere same party tricks (as well as using the most advanced process nodes) are what makes arm efficient. With tiered core system, the thread director, and advanced process node like 18A will make Intel x86 cores more efficient than contemporary ARM CPUs including the ones from Apple when 18A debuts. Mark my words.