Windows 10 & Intel Core i9-12900K Alder Lake Performance

[Wirko]

Well, that wouldn't be an AL specific problem (were we talking AL? I lost track).
Fwiw, I've always been wondering whether batching threads that do light work (as opposed to spreading them across as many execution units the CPU presents) is a pro or a con. I still haven't been able to figure that out.

Not AL specific. I wanted to point out that the nasty scenario you mentioned is very much possible when all cores are equal, if you have HT.

I'd like to see more testing with, for example, 2-, 4- and 8-thread compiling on P cores with HT vs. E cores. Results for ST indicate that E performance is about 2/3 of P performance. It would be very interesting to see how this scales. Two E cores about the same as one P core?

You're probably not hammering MySQL all day long a desktop, so an E core is the perfect place for it to loiter. Of course, we'd want that load to move to a P-thread as soon as it starts to, you know, work. A damn good example of a workload that's not that easy to figure out, thank you very much.

Not all day long but maybe for minutes on end, if you have a development machine and full software stack on it.

 

[bobmllr]

That usually depends on the compiler flags and the project itself. But if you have work for 24 threads, they would end up on all cores, whether it's Win10 or Win11.

The real nasty scenario is when you don't have work for all the cores. Win10 will randomly select a core and Win11 will relegate your compile work to E-cores as soon as you click away from the compiler window.

Thanks for that info. I wonder how each spawned process when compiling/linking is handled... non-UI processes?

 

[bug]

Thanks for that info. I wonder how each spawned process when compiling/linking is handled... non-UI processes?

View attachment 235970

If you don't specify a priority when you create it, the process inherits the priority of the parent.

 

[ADVenturePO]

Hi Techpowerup,
Intel #Alder #Lake is not working properly in some very demanding apps on #Windows #21H2. Period.
y-cruncher. Corona Render - most recent known issues.

You've tested on 21H1. Try 21H2. You'll see.

The funniest thing. 21H2 is totally optional after 2-3 days of using Windows 10.
I, by my own, recommended my client to update. Well, all performance went to sh*t.
Sorry but that's just what I feel.

Funnier. On newest Windows 11 version 12700k and 12900k is demanding more Vcore and thus running hotter.

I do not recommend installing W10 21H2 with Alder Lake. This, like many other updates, is really broken.

Cheers

 

[CyberCT]

What's the current scenario here?

I'm itching to build a new PC for the living room and go with a 12600K .... I'm used to Windows 10 and not sold on upgrading to Windows 11 yet ... but for the new build I might use Windows 11 from the start if it's really that much better with these new CPU architectures.

I have retro games from the 90s and early 2000s and hope this architecture doesn't break them (like Rise of the Trial 1996, Duke Nukem 3D, etc).

When spending hours researching yesterday ... I found one study from November 2021 where disabling E cores caused much higher power draw ... which surprised me. Is this still the case?

I'm going to be using this PC to game and web browse.

 

[bug]

What's the current scenario here?

I'm itching to build a new PC for the living room and go with a 12600K .... I'm used to Windows 10 and not sold on upgrading to Windows 11 yet ... but for the new build I might use Windows 11 from the start if it's really that much better with these new CPU architectures.

I have retro games from the 90s and early 2000s and hope this architecture doesn't break them (like Rise of the Trial 1996, Duke Nukem 3D, etc).

When spending hours researching yesterday ... I found one study from November 2021 where disabling E cores caused much higher power draw ... which surprised me. Is this still the case?

I'm going to be using this PC to game and web browse.

I've just disabled E-cores on mine, until they start being used properly. Now all cores are the same, so not a hint of a problem in Win10.

 

[Why_Me]

What's the current scenario here?

I'm itching to build a new PC for the living room and go with a 12600K .... I'm used to Windows 10 and not sold on upgrading to Windows 11 yet ... but for the new build I might use Windows 11 from the start if it's really that much better with these new CPU architectures.

I have retro games from the 90s and early 2000s and hope this architecture doesn't break them (like Rise of the Trial 1996, Duke Nukem 3D, etc).

When spending hours researching yesterday ... I found one study from November 2021 where disabling E cores caused much higher power draw ... which surprised me. Is this still the case?

I'm going to be using this PC to game and web browse.

i7 12700 / 12700F + B660 board + Windows 11

 

[CyberCT]

Do any of you play 90s games or found any games that are still broken having both cores active as "2 separate processors" and what are thoughts on Windows 11 these days? I heard there were still issues with it. Is it 100% backwards compatible with all Windows 10 programs?

Also ... do any of you have a KillaWatt meter and is power consumption way worse with e cores disabled in games and web browsing / general Windows use?

 

[arni-gx]

i7 12700 / 12700F + B660 board + Windows 11

OMG..... my cpu is too old for PC gaming today.......

 

[npelov]

running it on P-cores would be faster, and probably use less energy, completing sooner.

Running on P-cores will most likely NOT use less energy. I've done lots of energy benchmarks - performance per watt. Running at boost always takes more Wh for specific task. There is a sweet spot for every cpu when it comes to efficiency. That's why some of the server CPUs run at lower clocks and rely on more cores. My guess is (I haven't tried), unless you use every single feature of the power cores, efficient cores will be more energy efficient. I think (again guessing) that power cores have more extended hardware instructions which are emulated in the P-cores (using microcode). You can check if a task uses the extended hardware of the P-cores by fixing both E- and P-cores clcok speed to the same value and use affinity to run the same number of threads first on P- then on E-cores using affinity (which is not always an easy task). In linux you can set affinity prior executing process using taskset command. The child processes will inherit the affinity, so for software compiling you can specify cores even though lots of short lived processes are spawned. So if for the same frequency P-cores are faster then the process uses more of their features. Which is more efficient - you can only tell by benchmarking while measuring power at the wall. In most cases E-cores will be more efficient even though they will be slower.