Linux Fix Pending For Annoying Intel Lunar Lake Laptop Problems

[phoronix]

Phoronix: Linux Fix Pending For Annoying Intel Lunar Lake Laptop Problems

A patch is working its way to the mainline Linux kernel for addressing an annoyance affecting new Intel Core Ultra 200V "Lunar Lake" laptops...

https://www.phoronix.com/news/Intel-Lunar-Lake-Monitor-Bug

[mlau]

How is something like this not detected during CPU bringup?! Esp. since it's a known bug of the atom cores! I mean samples for validation arrive usually at least 1 year before market release, but noone thought about testing ubuntu on this thing? Intel just threw windows at it, it didn't immediately bluescreen and then they called it a success?

[milkylainen]

How is something like this not detected during CPU bringup?

(Not defending Intel even in the slightest here)

Because CPU/SoC development is like any other feature-driven software shit-show.
And buyers buying glitz, you get what you indirectly asked for.

Tbh, I'm actually surprised how well these multibillion transistor feature compexity monsters even work.
Atleast with respect to the pace and cut-throat competition if you miss a product development deadline.
Add nanoscale lithography with a bunch of quantum related problems and yeah. Not trivial any more.

[AdrianBc]

How is something like this not detected during CPU bringup?! Esp. since it's a known bug of the atom cores! I mean samples for validation arrive usually at least 1 year before market release, but noone thought about testing ubuntu on this thing? Intel just threw windows at it, it didn't immediately bluescreen and then they called it a success?

Yes, it is extremely annoying when such important functionality does not work as expected, and I completely agree with you that it is unbelievable when this is not discovered by quality control.

The use of the MONITOR/MWAIT pair for synchronizing threads is one of the extremely few original features of the Intel/AMD ISA.

The vast majority of the improvements brought to the x86/x86-64 ISA had existed for many years or decades in the instruction sets of other computers.

Only for MONITOR/MWAIT, which have been introduced by Intel in February 2004, in Pentium 4 Prescott, I am not aware of a predecessor. Later, a few other ISAs, e.g. IBM POWER, have copied Intel introducing equivalent instructions.

MONITOR/MWAIT remain a significant advantage of the Intel/AMD CPUs versus competitors, which allow a much faster communication between different threads and cores than when using inter-processor interrupts.

Failing to implement correctly and also to test exhaustively one of the most important competitive advantages of the Intel CPUs is hard to excuse.

[cutterjohn]

How is something like this not detected during CPU bringup?! Esp. since it's a known bug of the atom cores! I mean samples for validation arrive usually at least 1 year before market release, but noone thought about testing ubuntu on this thing? Intel just threw windows at it, it didn't immediately bluescreen and then they called it a success?

original ryzen... hangs on heavy loads using linux.... I had original batch and had to send for replacement falling back to my old trusty i7-3930k, which still kicked ass back then...

[EDIT]
...and yes, unlike so many proclaimed, I did NOT get cross-shipped a replacement 1800X, I had to twiddle my thumbs for a freaking month! THANKS AMD! Although I didn't bitch out loud until recently since they were basically broke...
[/EDIT]

[EDIT2]
I also didn't bitch until later about their XMP memory support being borked or nearly so, never got it to work for me, and hand tuning only got it so far, never to the 3200 advertised XMP rating, merely 2933 stable... above that would run for a bit then crash... zen4 OTOH 6k chef's kiss, mostly... and from early reports I understand that ECC if you use two sticks hit 6k(or whatever) then add a second matched pair it should 'mostly' run target 'bandwidth'... (ECC is still too expensive for me given that DDR5 is only now becoming reasonably priced w/o ECC, slowly... hopefully it lasts long enough to REALLY become 'reasonably' priced ECC or not... like DDR3 did...
[/EDIT2]

[AdrianBc]

original ryzen... hangs on heavy loads using linux.... I had original batch and had to send for replacement falling back to my old trusty i7-3930k, which still kicked ass back then...

[EDIT]
...and yes, unlike so many proclaimed, I did NOT get cross-shipped a replacement 1800X, I had to twiddle my thumbs for a freaking month! THANKS AMD! Although I didn't bitch out loud until recently since they were basically broke...
[/EDIT]

[EDIT2]
I also didn't bitch until later about their XMP memory support being borked or nearly so, never got it to work for me, and hand tuning only got it so far, never to the 3200 advertised XMP rating, merely 2933 stable... above that would run for a bit then crash... zen4 OTOH 6k chef's kiss, mostly... and from early reports I understand that ECC if you use two sticks hit 6k(or whatever) then add a second matched pair it should 'mostly' run target 'bandwidth'... (ECC is still too expensive for me given that DDR5 is only now becoming reasonably priced w/o ECC, slowly... hopefully it lasts long enough to REALLY become 'reasonably' priced ECC or not... like DDR3 did...
[/EDIT2]

I have also been one of those who have preordered a Ryzen 1800X. I have also bought an expensive motherboard and 64 GB of ECC memory for it.

Then I have discovered immediately that despite an impressive speed, the first Ryzen CPU was unusable garbage as any multithreaded application that fully used the CPU in Linux would crash very frequently.

I have not attempted to obtain a replacement, because I live in Europe and shipping the bad CPU would have had a non-negligible cost and there was the risk that the replacement CPU would also be bad, because AMD has never explained which was the real cause of the bug.

So I have dumped the expensive Ryzen 7 CPU and motherboard, losing the equivalent in EUR of almost $1000. I have kept only the ECC memory, which I have reused in a motherboard for an Intel Kaby Lake Xeon, which I have bought to substitute the non-functional Ryzen.

Nevertheless, I have not hold a grudge against AMD and I have considered that I have made a charity donation to them, to support competition in the CPU market.

A couple of years later, I have built my main desktop computer around a Ryzen 7 3700X, which has worked flawlessly, also with ECC memory. Later, I have replaced only the CPU in that computer with a Ryzen 9 5900X, which has brought a very useful and large increase in performance, and it has also always worked flawlessly. Now I am looking forward to upgrade the system to a 9950X. I also have an older server with an AMD Epyc CPU and multiple AMD GPUs, which has also always worked flawlessly.


So despite the great money loss caused to me by the first Ryzen, overall I am content with the value provided by AMD. Because there are still a few features of the Intel CPUs that are not implemented yet by AMD and which I like, every time when Intel launches a new CPU I still watch attentively, with the hope that perhaps this time there will be again a reason for me to buy an Intel CPU.

Unfortunately, since the beginning of 2019, not much less than 6 years ago (when I have bought a couple of Intel NUCs), after every Intel product launch I had to conclude that for my needs there are better solutions provided by AMD. Now I still have some hope for the launch of Intel Arrow Lake H at the beginning of 2025, which should be used to update the NUC-sized computers that currently use Meteor Lake, making them much more competitive with AMD Strix Point. An AMD HX 370 is a much better CPU+GPU than any Intel Meteor Lake, but nobody has presented until now any mini PC with HX 370 that can compete in functionality with an ASUS NUC with Meteor Lake. So if I were to buy a mini PC now I would have to choose between either having an obsolete CPU or having inadequate peripherals. So Intel might still have a chance with me if they do not botch the Alder Lake H launch. A bug like that just reported on this site for Lunar Lake, that makes MONITOR/MWAIT unusable, would completely disqualify any CPU for me.

[mlau]

original ryzen... hangs on heavy loads using linux.

This was different: too optimistic binning. The lunar lake thing is a known bug in most previous-gen atom cores. After so much time, I'd have expected intel to have this fixed, since monitor/mwait are important for power savings.

[Anux]

How is something like this not detected during CPU bringup?!

Well it's Intel, but they are somewhat fair to each OS because with the current W11 build you can't even use the iGPU.