Damn, This is why you're the man. Hilarious that you've figured it out before intel and asus. Maybe they'll ship you a red phone, just a direct line to Gelsinger. Your first call too pat could be that XMP isn't "overclocking" if they use it in marketing. You should sell them your crash-cart™ design.
Just an FYI, the Clover boot loader will allow you to change/replace the ACPI tables as returned by the BIOS before booting the OS so if you need to "fix" them you can. I have used this method to fix bugs I found in an HP laptop's ACPI implementation.
Thank you so much 🥰 for doing all this excellent hard work , so Linux keeps growing. It's a shame that the vendors and consortium are not doing their chores. Without people like you Linux would have never took over most of the world's computers. Thanks again and all the best to you 🐧
Well, it's not really him, so much as it is all of us - up to us, that is. Community run softwares, OSs. I think we expect too much and do too little. It's interesting learning about this stuff though! I would love to contribute if I could
Oh my god. Thanks for the heads up. I think my Linux system will remain AMD for awhile. They've been killing it in the GPU driver space too. When I plan my upgrades I think my Windows gaming PC will be an i5 Alder Lake, but only after the cheaper motherboards come out.
Checked my stuff on my MSI pro z690-a ddr4, it seems to be working fine for me, my hyper threads are set to like 38 in the acpi_cppc/highest_perf files, my effencicy cores are set to 63, and my performance cores are set to 64. I do have xmp enabled in the bios so it seems to be fine on this MSI board anyways. Here's an ugly bash one liner to get this info on your boards if any who read this are curious for cpu in $(ls /sys/devices/system/cpu | grep "cpu" | grep -v freq | grep -v idle); {echo $cpu; cat /sys/devices/system/cpu/$cpu/acpi_cppc/highest_perf}
@@Flamewires the reason those cores are higher than the others is because those are the favored cores for even more boosting. An extra multiplier bump on those specific cores.
Might wanna take a few (dozen) more steps back. Bits of skin, bone, hardware and keyboard switches tend to fly far and wide when new core hardware's released... especially when the manufacturers half-arse their Linux support (as usual). :-)
This is good to know. My 12900K/Asus Z690 Prime-P (both "D5" and D4 variants) should be coming in in about a month from now so I'll be able to play around with it then (and looking forward to it). I think for me, compounding the issue is going to be that I am stuck/relegated to running CentOS 7.7.1908 because my HPC apps doesn't support anything newer at the moment. So, we'll have to see how that goes.
On my MSI PRO Z690-A DDR4 running a 12600k i get a 1 for the cat /proc/sys/kernel/sched_itmt_enabled - so it seems enabled. For the acpi_cppc stuff the PCores (core0-11) I get 63, so also for the hyperthreads and for the ECores 36 - I think I justed patched the bios when I got the board. Running Linux Kernel version 5.14.15 And pretty happy with the experience so far powerconsumption in idle (10 docker containers and a vm running) is lower for me while doubling the ram sticks than on my previous ryzen 3600 with an x570 mobo.
Thanks for this content :) I am interested in getting new hardware and I will be using linux for gaming for the first time when I do so this information is very useful for me.
You're pretty much describing the Xeon Phi 7290F. It was a 72-core 1.5GHz chip with built-in 100Gb networking. It was a pretty obscure part for high-end compute, and was only supported by a couple LGA3647 motherboards, but the concept could totally be revived on the new architecture.
@@asm_nop Expect that the micro-architecture of knights landing sucked balls unless all you were doing were AVX512 operations ac-cross a lot of threads. Gracemont is a lot better situated for good server performance and I expect it will beat A72 and trade blows with the A76.
Imagine a world where all ACPI tables were correct, and every BIOS queried the OS with ACPI_OSI on an interface by interface level... Yeah, I like to dream.
Depends who you turn to - vendor. My best guess is this 'new' Intel way of designing is backwards and forwards, it serves to "microfunctionality", background services, updates, task schedulers (e cores)... This is useful in a sense but sorta breaks these "conventions" we keep almost arriving at. This core scheduler of there's, it's a very different abstraction, say, from A new, very different type of CPU. One that uses a vertical stacking cache layer, as opposed to a scheduling layer. Supposedly the performance to energy gains are drastic, so, could imagine thermals as well... Closer to the new Ryzen architectures, which are more traditional but refined, with performance gains, all the while for fair wattage/consumption. The stats are quite interesting, it seems the alder lake is set to want-to use more power... //Alluded to previously. This is kind of backed up by marketing hype around,specifically iro laptops, where the TDPs for the Ryzen laptops (5900H I think) were consistently lower, meaning greater GPU throughput. Though, the metrics for the 12th i9, in-game, were better when compared with the Ryzen. That said, as per one source, there was only one game that the 12th gen couldn't run (red dead 2?), but it did outperform most of the comparisons with Ryzen, while Ryzen failed at running some games if memory serves. That said, there is the new Ryzen architecture set to launch this year, though, many laptops on the market, least, where I live, don't support the new Ryzen CPU hardware just yet.
i'm still waiting for intel to release quad socket server cpus with p and e cores. i'll just sit back and watch the OS developers cry and tear out their hair trying to figure out that scheduling nightmare. they're probably not making those server cpus because consistency across cores but hey i can dream
Not working as it "should". Primarily just assigning threads to the wrong cores, whether wholly or just occasionally. From what I've read/seen to date.
Imho the user interactiveness is janky. You can for sure feel when e.g. the web browser goes from a p core to an e core while lots of JavaScript is running. Its a sort of subtle but noticible jank.
@@wendelltron that's fair, on the MSI board it's working great though, my performance and snappiness is incredible, even without the more advanced patches intel's working on. So idk about other boards but this specific msi board is working great! Maybe one to try?
See, this is why Windows is great and Linux sucks, the distro people don't care about just making things work out the box but talk about mass adoption, Sorry, much pent up frustration with those kinds of comments everywhere lately, needed to let some out.
Manufacturers need to support Linux better. There will be people that say NO, Windows is dominant, why should they support Linux. I'm happy I ditched Windows. Forgetting the Windows nonsense I learned. Re-educating myself in Linux. And don't say "How is Grandma going to cope with Linux" The same way she learned Windows.
Honestly I don't see why linux isn't popular among people who don't know computers at all. My experience with basic stuff like web browsing on linux is you install and go and it just works. I would recommend linux to anyone who just wants to watch youtube and check email.
I keep trying to use Linux, I try every year or so, use it for a week or so, then i break it trying to figure what version of Java need to run,to run this program or that program, and I kill it and/or the programs run wonky.....so it'll get there, eventually! I hate windows soo much, but I have to use it....for now
Java killed it last time, I had run 3 different versions. I cant remember exactly but I tried using tutorials online, and its always janky. Command line copy and paste,and bam its dead.....I ran pop os,Mint and fedora. Those are what I tried from recommendations online, I always break it, then go back to windows.....and I hate windows,but I use it anyways. I like overclocking too,but I can't really do that either on linux, especially GPUs. Just cpu in the bios of course
@@brianmccullough4578 Run a VM of the OS you're running on bare metal. Tinker and faff about in the VM, then transfer that knowledge to the bare metal. Also, these days with Btrfs Filesystem snapshots (done by either timeshift or snapper) will often save the day, just boot from a recent working backup, make that the working copy and carry on. But if you try to force Linux to be like WIndows, it'll always kick you in the teeth. like trying to drive an open-wheel track car on a rally stage, like a track car. It'll chew you up and spit you out. Gotta use the right tools, in the right way for the environment. Just gotta bear in mind. Windows is like a modern road car. Very few "user serviceable parts". Linux is more like a kit-car, it's almost all user-serviceable and you can easily get covered in grease if you so choose, but at the very least gonna get ya fingers dirty. Learn to roll around, then crawl, then stumble, then walk, then run, then double-twisting-backward-somersaults with swan-diving finish. Linux VM on Windows, then Dual Booting Windows - Linux (separate drives) with Linux VM for checking stuff, then ... whatever works for you.
@Peanut - INTP good ideas! I will try that. Right now,I have an old laptop that I play on, just distro hopping, if I screw up,I just start over! Is there a distro that has more features built in, or a more full featured store? Like is the pop shop good? I don't really care about the OS being streamlined, I kind of want it to just work. I'll get there guys,no worries!
You're affected by this bug then. The lulz part is I don't think they realized the i5 is technically not a turbo boost max 3 part. It needs the tbmt machinery to do thread director though
Life becomes much easier when you reject this bleeding edge, poorly tested, bug ridden hardware and live on currentGen -1; Since vendors refuse to release a quality product, I refuse to pay a premium for the latest gear.
This man is doing the lords work. I sure wish vendors would do this level of testing before shipping hardware.
Psshhh. Are you READY for early access motherboards!?!!! Oh wait they're just called motherboards, nvm.
Damn, This is why you're the man. Hilarious that you've figured it out before intel and asus. Maybe they'll ship you a red phone, just a direct line to Gelsinger. Your first call too pat could be that XMP isn't "overclocking" if they use it in marketing. You should sell them your crash-cart™ design.
I think that's an ASUS thing, not an Intel thing
Just an FYI, the Clover boot loader will allow you to change/replace the ACPI tables as returned by the BIOS before booting the OS so if you need to "fix" them you can. I have used this method to fix bugs I found in an HP laptop's ACPI implementation.
Cool. Wasn't clover a bootloader designed to boot OS X on x86? I can imagine that being useful for a hackintosh.
I love how enthusiastic you are about such in depth content! Thank you, you're awesome!!!
Thank you so much 🥰 for doing all this excellent hard work , so Linux keeps growing. It's a shame that the vendors and consortium are not doing their chores. Without people like you Linux would have never took over most of the world's computers. Thanks again and all the best to you 🐧
Well, it's not really him, so much as it is all of us - up to us, that is. Community run softwares, OSs. I think we expect too much and do too little. It's interesting learning about this stuff though! I would love to contribute if I could
Thank you Wendell, started to think something was wrong.
Appreciate the work you do
Thanks for the hard work Wendel ^^
Thank you for doing so much work, for just a few of us
Oh my god. Thanks for the heads up. I think my Linux system will remain AMD for awhile. They've been killing it in the GPU driver space too. When I plan my upgrades I think my Windows gaming PC will be an i5 Alder Lake, but only after the cheaper motherboards come out.
Checked my stuff on my MSI pro z690-a ddr4, it seems to be working fine for me, my hyper threads are set to like 38 in the acpi_cppc/highest_perf files, my effencicy cores are set to 63, and my performance cores are set to 64. I do have xmp enabled in the bios so it seems to be fine on this MSI board anyways.
Here's an ugly bash one liner to get this info on your boards if any who read this are curious
for cpu in $(ls /sys/devices/system/cpu | grep "cpu" | grep -v freq | grep -v idle); {echo $cpu; cat /sys/devices/system/cpu/$cpu/acpi_cppc/highest_perf}
@@Flamewires the reason those cores are higher than the others is because those are the favored cores for even more boosting. An extra multiplier bump on those specific cores.
Do you have a video talking about the issues you had with DisplayPort cables?
Never used linux, just like watching from afar.
Might wanna take a few (dozen) more steps back. Bits of skin, bone, hardware and keyboard switches tend to fly far and wide when new core hardware's released... especially when the manufacturers half-arse their Linux support (as usual). :-)
Zen 3 works very well.
Wow, all I can say it thanks, your ace.
This is good to know.
My 12900K/Asus Z690 Prime-P (both "D5" and D4 variants) should be coming in in about a month from now so I'll be able to play around with it then (and looking forward to it).
I think for me, compounding the issue is going to be that I am stuck/relegated to running CentOS 7.7.1908 because my HPC apps doesn't support anything newer at the moment.
So, we'll have to see how that goes.
Oh, this sounds like it came out hot.
I wish I had more time to mess around with Linux and bios' at this level
_level1_ that is
_engagement_
On my MSI PRO Z690-A DDR4 running a 12600k i get a 1 for the cat /proc/sys/kernel/sched_itmt_enabled - so it seems enabled.
For the acpi_cppc stuff the PCores (core0-11) I get 63, so also for the hyperthreads and for the ECores 36 - I think I justed patched the bios when I got the board.
Running Linux Kernel version 5.14.15
And pretty happy with the experience so far powerconsumption in idle (10 docker containers and a vm running) is lower for me while doubling the ram sticks than on my previous ryzen 3600 with an x570 mobo.
Excellent thank you so much. Is this with xmp on?
@@wendelltron Yes with xmp enabled otherwise almost untouched form the optimized defaults so no manual overclocking or something else
Thanks for this content :) I am interested in getting new hardware and I will be using linux for gaming for the first time when I do so this information is very useful for me.
heyy that cable tester is saving lives
So when is intel going to release a 50+ All E-core SKU to compete with Graviton?
You're pretty much describing the Xeon Phi 7290F. It was a 72-core 1.5GHz chip with built-in 100Gb networking. It was a pretty obscure part for high-end compute, and was only supported by a couple LGA3647 motherboards, but the concept could totally be revived on the new architecture.
@@asm_nop Expect that the micro-architecture of knights landing sucked balls unless all you were doing were AVX512 operations ac-cross a lot of threads.
Gracemont is a lot better situated for good server performance and I expect it will beat A72 and trade blows with the A76.
Imagine a world where all ACPI tables were correct, and every BIOS queried the OS with ACPI_OSI on an interface by interface level... Yeah, I like to dream.
I don't think I've ever seen Wendell so agitated!
Feeding the algorithm with like/sub/bell and comment.
PSA: make a hotkey to help👌
Do you think these issues will be less of a problem for future hybrid architectures?
Depends who you turn to - vendor. My best guess is this 'new' Intel way of designing is backwards and forwards, it serves to "microfunctionality", background services, updates, task schedulers (e cores)... This is useful in a sense but sorta breaks these "conventions" we keep almost arriving at. This core scheduler of there's, it's a very different abstraction, say, from
A new, very different type of CPU. One that uses a vertical stacking cache layer, as opposed to a scheduling layer. Supposedly the performance to energy gains are drastic, so, could imagine thermals as well... Closer to the new Ryzen architectures, which are more traditional but refined, with performance gains, all the while for fair wattage/consumption.
The stats are quite interesting, it seems the alder lake is set to want-to use more power... //Alluded to previously. This is kind of backed up by marketing hype around,specifically iro laptops, where the TDPs for the Ryzen laptops (5900H I think) were consistently lower, meaning greater GPU throughput. Though, the metrics for the 12th i9, in-game, were better when compared with the Ryzen. That said, as per one source, there was only one game that the 12th gen couldn't run (red dead 2?), but it did outperform most of the comparisons with Ryzen, while Ryzen failed at running some games if memory serves. That said, there is the new Ryzen architecture set to launch this year, though, many laptops on the market, least, where I live, don't support the new Ryzen CPU hardware just yet.
i'm still waiting for intel to release quad socket server cpus with p and e cores. i'll just sit back and watch the OS developers cry and tear out their hair trying to figure out that scheduling nightmare. they're probably not making those server cpus because consistency across cores but hey i can dream
I have the feeling the LTT staff use their cable tester on all the cables that Linus touches.
Please, Satan - you owe me - this man is The Dude. Help him, help us all..! I jest but this is really important work, ty.
UPDATE! WE NEED UPDATE!
You should prioritize processes to cores manually also in Windows anyway.
and affinity
I've been stung way too many times by being a "new hardware pioneer" on Linux. Gonna sit back and wait until the dust settles this time...
I'm waiting for System76 selling a 12th gen laptop.
I guess wsl2 is running it in a VM of some sort?
Yes, yes it is.
Intel, AMD, ASUS, MSI, Linus Torvalds, Red Hat, Microsoft (etc) all need Level1Linux on speed dial.
Broken as in unusable, or broken as in not working as it should, but still usable?
Not working as it "should". Primarily just assigning threads to the wrong cores, whether wholly or just occasionally. From what I've read/seen to date.
Usable, but not ideal, so sometimes high resource usage apps are put on effencicy cores, which will result in lower performance, but it'll still work
Imho the user interactiveness is janky. You can for sure feel when e.g. the web browser goes from a p core to an e core while lots of JavaScript is running. Its a sort of subtle but noticible jank.
@@wendelltron that's fair, on the MSI board it's working great though, my performance and snappiness is incredible, even without the more advanced patches intel's working on. So idk about other boards but this specific msi board is working great! Maybe one to try?
@@kevinpyro3008 yeah that's what I mentioned at the start. My msi unity z690 also works superbly
See, this is why Windows is great and Linux sucks, the distro people don't care about just making things work out the box but talk about mass adoption,
Sorry, much pent up frustration with those kinds of comments everywhere lately, needed to let some out.
Manufacturers need to support Linux better.
There will be people that say NO, Windows is dominant, why should they support Linux.
I'm happy I ditched Windows. Forgetting the Windows nonsense I learned. Re-educating myself in Linux.
And don't say "How is Grandma going to cope with Linux" The same way she learned Windows.
Honestly I don't see why linux isn't popular among people who don't know computers at all. My experience with basic stuff like web browsing on linux is you install and go and it just works. I would recommend linux to anyone who just wants to watch youtube and check email.
I keep trying to use Linux, I try every year or so, use it for a week or so, then i break it trying to figure what version of Java need to run,to run this program or that program, and I kill it and/or the programs run wonky.....so it'll get there, eventually! I hate windows soo much, but I have to use it....for now
What? How?
Java killed it last time, I had run 3 different versions. I cant remember exactly but I tried using tutorials online, and its always janky. Command line copy and paste,and bam its dead.....I ran pop os,Mint and fedora. Those are what I tried from recommendations online, I always break it, then go back to windows.....and I hate windows,but I use it anyways. I like overclocking too,but I can't really do that either on linux, especially GPUs. Just cpu in the bios of course
@@brianmccullough4578 Run a VM of the OS you're running on bare metal. Tinker and faff about in the VM, then transfer that knowledge to the bare metal.
Also, these days with Btrfs Filesystem snapshots (done by either timeshift or snapper) will often save the day, just boot from a recent working backup, make that the working copy and carry on.
But if you try to force Linux to be like WIndows, it'll always kick you in the teeth. like trying to drive an open-wheel track car on a rally stage, like a track car. It'll chew you up and spit you out. Gotta use the right tools, in the right way for the environment.
Just gotta bear in mind. Windows is like a modern road car. Very few "user serviceable parts". Linux is more like a kit-car, it's almost all user-serviceable and you can easily get covered in grease if you so choose, but at the very least gonna get ya fingers dirty.
Learn to roll around, then crawl, then stumble, then walk, then run, then double-twisting-backward-somersaults with swan-diving finish.
Linux VM on Windows, then Dual Booting Windows - Linux (separate drives) with Linux VM for checking stuff, then ... whatever works for you.
@Peanut - INTP good ideas! I will try that. Right now,I have an old laptop that I play on, just distro hopping, if I screw up,I just start over! Is there a distro that has more features built in, or a more full featured store? Like is the pop shop good? I don't really care about the OS being streamlined, I kind of want it to just work. I'll get there guys,no worries!
@@brianmccullough4578 Overclocking AMD GPUS can be done with Corectrl.
You only need openjdk. I can't imagine that breaking your system...
These are not the performance cores you're looking for.
These are not the performance cores we're looking for. Move along!
😍😍😍
Just checked my cores on ubuntu and an asus motherboard and all cores highest_perf value is 255 on 12600k
You're affected by this bug then. The lulz part is I don't think they realized the i5 is technically not a turbo boost max 3 part. It needs the tbmt machinery to do thread director though
Soon™
Oof...
Life becomes much easier when you reject this bleeding edge, poorly tested, bug ridden hardware and live on currentGen -1; Since vendors refuse to release a quality product, I refuse to pay a premium for the latest gear.
1000%, but even then we can't expect them to do all the work.
Wendell is an absolute saint for his contributions to computer janitors across the world
@@popcorny007 yes we can. they want people to buy their product, so they can do all the testing to make sure it works.
@@jmwintenn Sure, but not before the product goes to market :P
i learned this the hard way just now