Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[igor_kavinski]

Is there any evidence that such a thing exists? It's mostly on the OS anyways.

The OS is taking hints from the Intel Thread Director. What if the ITD is giving the wrong hints and there is nothing that can be done to improve those hints maybe because it's hardwired into the silicon? The evidence is the annoyance of losing performance in background rendering/encoding tasks when they switch to E-cores the moment something simple like foreground browsing is done. The ITD should be smart enough to know that if a process is 100% taxing a P-core, switching it over to E-cores will only make it slower and last longer.

The P-cores and E-cores should ideally work in tandem. If the P-cores are busy, the E-cores handle the task. As soon as one P-core is freed, the task switches over from the E-core to the P-core to finish it faster. What happens in Windows 11 is that CPU intensive tasks get thrown on the E-cores as soon as they lose foreground focus.

 

[Tech Junky]

ITD

I think there's more coming and this switching is just to make use of both core types. ITD is basically just the first release at this point and will evolve in due time. To keep track of the evolvement look to Linux kernel releases for hints of improvements. It's all os based instruction anyway and not hard coded in the hw. That's why the focus is on W11 since MS wants to keep moving forward and hasn't ported it to W10 at this point.

 

[Exist50]

The OS is taking hints from the Intel Thread Director. What if the ITD is giving the wrong hints and there is nothing that can be done to improve those hints maybe because it's hardwired into the silicon? The evidence is the annoyance of losing performance in background rendering/encoding tasks when they switch to E-cores the moment something simple like foreground browsing is done. The ITD should be smart enough to know that if a process is 100% taxing a P-core, switching it over to E-cores will only make it slower and last longer.

The P-cores and E-cores should ideally work in tandem. If the P-cores are busy, the E-cores handle the task. As soon as one P-core is freed, the task switches over from the E-core to the P-core to finish it faster. What happens in Windows 11 is that CPU intensive tasks get thrown on the E-cores as soon as they lose foreground focus.

As far as I'm aware, most of the feedback from the hardware is relative performance/efficiency levels. If that was egregiously off, we'd know by now. Same applies more generally to the fear that Thread Directory is fundamentally broken, so I really don't see why you're pushing this angle.

 

[igor_kavinski]

We'll only know when Raptor Lake comes out and if it gives a better experience. If ITD is broken, how can we know for sure? There isn't anything else out there to compare it with. With Raptor Lake, if Intel improves ITD, it will tout that as a feature to push ADL users to upgrade.

 

[coercitiv]

As far as I'm aware, most of the feedback from the hardware is relative performance/efficiency levels. If that was egregiously off, we'd know by now. Same applies more generally to the fear that Thread Directory is fundamentally broken, so I really don't see why you're pushing this angle.

There's no evidence so far that Thread Dictator is broken. It makes a number of compromises, and we know some of it's "weird" behavior is intentional.

Now usually when I’m dealing with video exports, it’s the video throughput that is my limiting factor. I need the video to complete, regardless of what I’m doing in the interim. By defocusing the video export window, it now moves to the slower E-cores. If I want to keep it on the P-cores in this mode, I have to keep the window in focus and not do anything else. The way that this is described also means that if you use any software that’s fronted by a GUI, but spawns a background process to do the actual work, unless the background process gets focus (which it can never do in normal operation), then it will stay on the E-cores.

In my mind, this is a bad oversight. I was told that this is explicitly Microsoft’s choice on how to do things.

I was however told that if the user changes the Windows Power Plan to high-performance, this behavior stops. In my mind this isn’t a proper fix, but it means that we might see some users/reviews of the hardware with lower performance if the workload doing the work is background, and the reviewer is using the default Balanced Power Plan as installed. If the same policy is going to apply to Laptops, that’s a bigger issue.

All (empirical) evidence I've seen so far is this type of behavior is software controlled, not hardware. IMHO anyone claiming Intel hardware guided scheduling may be broken because we don't have bulletproof evidence to the contrary is just pushing FUD and should be ignored.

 

[ashFTW]

We'll only know when Raptor Lake comes out and if it gives a better experience. If ITD is broken, how can we know for sure? There isn't anything else out there to compare it with. With Raptor Lake, if Intel improves ITD, it will tout that as a feature to push ADL users to upgrade.

The ITD, as is true of all chip components, will likely get more sophisticated in the future heterogeneous generations. It currently may not handle all cases, but that doesn’t mean it’s broken.

Edit: Looks like @coercitiv said something similar just as was typing my reply.

 

[igor_kavinski]

Sign in to your account

techcommunity.microsoft.com

Markfw complained about this mouse lag issue too. Either Windows scheduler has a bug or it is getting wrong feedback from ITD.

 

[igor_kavinski]

The ITD, as is true of all chip components, will likely get more sophisticated in the future heterogeneous generations. It currently may not handle all cases, but that doesn’t mean it’s broken.

Hardware that works mostly but not always is not what I would pay for.

 

[ashFTW]

Hardware that works mostly but not always is not what I would pay for.

There are heuristics baked in the ITD. Heuristics by nature “work mostly but not always”. It’s not like it’s resulting in wrong computational results.

 

[Plasma1996]

Sign in to your account

techcommunity.microsoft.com

Markfw complained about this mouse lag issue too. Either Windows scheduler has a bug or it is getting wrong feedback from ITD.

This sounds more like an OS power-saving feature. But in any event, it's something I've never seen on my i9 12900k no matter how heavy or light the CPU or system load.

 

[Markfw]

This sounds more like an OS power-saving feature. But in any event, it's something I've never seen on my i9 12900k no matter how heavy or light the CPU or system load.

Well, I have on my 12700F. Its not imaginary.

 

[Plasma1996]

Well, I have on my 12700F. Its not imaginary.

What kind of mouse and keyboard?

 

[Markfw]

What kind of mouse and keyboard?

USB, why ??? There is no reason for a USB mouse or keyboard to lockup like this. PS2 is dead.... From what I read, this is exactly the problem with big.little and the new Microsoft thread director.

 

[scannall]

There's no evidence so far that Thread Dictator is broken. It makes a number of compromises, and we know some of it's "weird" behavior is intentional.



All (empirical) evidence I've seen so far is this type of behavior is software controlled, not hardware. IMHO anyone claiming Intel hardware guided scheduling may be broken because we don't have bulletproof evidence to the contrary is just pushing FUD and should be ignored.

It does seem pretty quirky at the moment. But being V1 of something new, that isn't surprising. It's interesting, and likely useful to quite a few people so I am not saying it sucks. But V2 will work better. It's just how these things go. Build something new, see it in use in the wild. Gather your data on what works and what doesn't then go improve it.

It isn't something I am interested in atm, and that's fine too. Having just put together a Ryzen 5800 (Non X) system I'm good for a while.

 

[Markfw]

It does seem pretty quirky at the moment. But being V1 of something new, that isn't surprising. It's interesting, and likely useful to quite a few people so I am not saying it sucks. But V2 will work better. It's just how these things go. Build something new, see it in use in the wild. Gather your data on what works and what doesn't then go improve it.

It isn't something I am interested in atm, and that's fine too. Having just put together a Ryzen 5800 (Non X) system I'm good for a while.

Exactly why I bought this, to see how it works, from a personal perspective. Yes, the P-cores are fast. The e-cores, power usage and software just sucks.

 

[igor_kavinski]

Intel Core i9-12900K (Alder Lake) Review

The Intel Core i9-12900K combines two types of processor cores onto a single die. The 16-core, 24-thread chip can reach up to 5.2GHz.

www.storagereview.com

Love the Be Quiet! rig. Wouldn't mind it at all if it showed up on my doorstep somehow

The review is weird though. Desktop vs. workstation? 24 threads vs. 128? What is the point of this comparison??? Instead of buying a workstation with ECC memory, get an overpriced rig with a gaming mobo to run your professional workloads?

 

[Plasma1996]

The OS is taking hints from the Intel Thread Director. What if the ITD is giving the wrong hints and there is nothing that can be done to improve those hints maybe because it's hardwired into the silicon? The evidence is the annoyance of losing performance in background rendering/encoding tasks when they switch to E-cores the moment something simple like foreground browsing is done. The ITD should be smart enough to know that if a process is 100% taxing a P-core, switching it over to E-cores will only make it slower and last longer.

The P-cores and E-cores should ideally work in tandem. If the P-cores are busy, the E-cores handle the task. As soon as one P-core is freed, the task switches over from the E-core to the P-core to finish it faster. What happens in Windows 11 is that CPU intensive tasks get thrown on the E-cores as soon as they lose foreground focus.

Thread Director isn't a scheduler that is handled by the OS. If the OS needs to move a cpu intensive task to an E-core to maintain system responsiveness it will do it that has nothing to do with any feedback it gets from ITD.

 

[Zucker2k]

USB, why ??? There is no reason for a USB mouse or keyboard to lockup like this. PS2 is dead.... From what I read, this is exactly the problem with big.little and the new Microsoft thread director.

Source?

 

[Markfw]

Source?

Page 99 - Question - Alder Lake - Official Thread

Page 99 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Exist50]

User inputs are almost always highest priority. Would be surprised if it doesn't run on a P-core, much less the choice of cores causing issues.

Gracemont is Skylake-class IPC. There's no way it should cause a mouse to lag.

 

[Markfw]

User inputs are almost always highest priority. Would be surprised if it doesn't run on a P-core, much less the choice of cores causing issues.

Gracemont is Skylake-class IPC. There's no way it should cause a mouse to lag.

I have no idea what is causing it. But on win 11 (that alone could be the problem) on my 12700F, it has happened 2 or 3 times. In a month.

 

[controlflow]

I'm not denying that this alleged mouse issue exists but I can't say I've ever encountered it on my 12900K and MSI Z690 and Win 11 setup.

 

[DrMrLordX]

Gracemont is Skylake-class IPC. There's no way it should cause a mouse to lag.

Could be an interconnect issue. Skylake CPUs didn't have the same interconnect layout, and if user input requests get constantly bounced between core clusters, there could be some lag.

 

[igor_kavinski]

Could be an interconnect issue. Skylake CPUs didn't have the same interconnect layout, and if user input requests get constantly bounced between core clusters, there could be some lag.

Gracemont has Skylake-like performance. Not Skylake-like architecture, does it?

 

[DrMrLordX]

Gracemont has Skylake-like performance. Not Skylake-like architecture, does it?

. . . sort of. I haven't done an exact comparison, but it looks like Gracemont is stronger in integer workloads and weaker in fp. But that is not the point. We've had CPUs more than fast enough to handle user input requests from USB devices for over a decade. The performance of the underlying architecture itself is likely not the source of these problems.

The fact that Alder Lake has an exotic core layout and a thread director scheme that behaves peculiarly might be responsible for the problem.