Z690 Chipset Motherboards and Alder Lake CPU

[etorix]

Sorry, where can I find this one, if it's possible?

Oops! Attachment added…

 

[dehjomz]

Interesting.
I'm a Frontend developer, compiling Webpack (Javascript bundler) can consume 100% cpu usage for 1min or more.
Other than that, I don't have any other real 100% cpu demanding tasks
I'm really not sure, I do have a budget for the i9-12900k but I wonder if an 120mm AIO is enough

Some people on youtube only look at the peak values and then claim the 12900K 'gooses voltage and power efficiency to beat the m1 and Ryzen.'

But what they fail to consider is the average power consumption over time. Intel just has a wider dynamic range than the Ryzen processors (lower idle and higher peak). But since most people aren't engaging in an all-core load most of the time, simply focusing on peak power consumption is a very myopic view of the world.

By the way, power companies bill residential units for their kWh (average power) usage not peak kW. Companies add on other charges in the USA like wholesale electric transmisison charges, and if you live in a power market area then average LMP prices from purchased generation, and some other surcharges like fuel costs and renewable energy credits possibly as well.

So even if the Intel system peaks briefly at 212W, if the use case over the course of an hour is such that it is idling (or using a few cores) at a lower wattage than the Ryzen system, then the Ryzen kWh will be higher than the Intel kWh.

So the power consumption conversation should focus on peak yes, but also on average power consumed over the course of time as well for a more accurate depiction of what's going on.

 

[CaseySJ]

Yes, I see what you mean:

• With cores enabled = 8, thread count will be 16, and all 16 threads will be real.
• After a little breakfast, I'll try to patch the kernel to set core_count = 8.
  • Q: Is core_count in cpu_thread.c referring to "cores enabled" or "total physical cores"?

nLThreadsPerCore = thread_count / core_count = 16 / 10 = 1;
nLThreadsPerPackage = nLCoresPerPackage * nLThreadsPerCore = 10 * 1 = 10;
nCPUs = nPackages * nLThreadsPerPackage = 1 * 10 = 10;

If we patch kernel so that core_count = 8:

• nLThreadsPerCore = 16 / 8 = 2
• nLThreadsPerPackage = 8 * 2 = 16 (this assumes nLCoresPerPackage is dependent on core_count as well)
• nCPUs = 1 * 16 = 16

@Elias64Fr

This was tricky to patch, but setting cpu_info.core_count = 8 causes fast and furious boot failure! Back to the drawing board. Probably okay to leave this topic for now...

 

[CaseySJ]

Some people on youtube only look at the peak values and then claim the 12900K 'gooses voltage and power efficiency to beat the m1 and Ryzen.'

But what they fail to consider is the average power consumption over time. Intel just has a wider dynamic range than the Ryzen processors (lower idle and higher peak). But since most people aren't engaging in an all-core load most of the time, simply focusing on peak power consumption is a very myopic view of the world.

By the way, power companies bill residential units for their kWh (average power) usage not peak kW. Companies add on other charges in the USA like wholesale electric transmisison charges, and if you live in a power market area then average LMP prices from purchased generation.

So even if the Intel system peaks briefly at 212W, if the use case over the course of an hour is such that it is idling (or using a few cores) at a lower wattage than the Ryzen system, then the Ryzen kWh will be higher than the Intel kWh.

So the power consumption conversation should focus on peak yes, but also on average power consumed over the course of time as well for a more accurate depiction of what's going on.

Yes, that's the entire point of the YouTube video I referenced!

 

[dehjomz]

Yes, that's the entire point of the YouTube video I referenced!

Yep!

This cpuid.c, et. al puzzle is very interesting. There has to be a way forward.

 

[MatiCohen]

@MatiCohen
I was wrong, it is possible to enable HT for each cores following BIOS Setup (Intel® 600 Series) from Gigabyte :

View attachment 534769

Interesting, what if @CaseySJ tried the following
6P, enable HT only on 2 cores
4E
total 12 threads. lol

Did anyone notice a breakthrough in the Russian forums? I guess vit and others are more active there

 

[dehjomz]

For those interested, here’s a video preview of the Z690 Aero G from Gear Seekers. A beautiful board indeed.

 

[starghost1100]

Interesting, what if @CaseySJ tried the following
6P, enable HT only on 2 cores
4E
total 12 threads. lol

Did anyone notice a breakthrough in the Russian forums? I guess vit and others are more active there

I don't think that we (Russians) have a lot of hackintosh forums )

 

[Elias64Fr]

@Elias64Fr

This was tricky to patch, but setting cpu_info.core_count = 8 causes fast and furious boot failure! Back to the drawing board. Probably okay to leave this topic for now...

View attachment 534775

View attachment 534774

On cpuid.c, you have highlighted an interesting info :
info_p->core_count = bitfield32 ((uint32_t)msr, 31, 16);
info_p->thread_count = bitfield32 ((uint32_t)msr, 15, 0);

On AppleIntelInfo, at MSR_CORE_TREADCOUNT (0x35), I have on my laptop :

Data = 0x6000C
lower Word (15-0) = 0x000C = 12 thread_count
higher Word (31-16)= 0x0006 = 6 core_count

core_count and thread_count are supplied by MSR_35

 

[Middleman]

Want to report another discovery I made earlier today.
Performance on the Asus even on Windows was strangely low.
Have a look, this was using a Balanced profile:

whereas this one was using a High Performance setting, including using a Turbo setting in BIOS.
All had P + E Cores AND Hyperthreading on!!!
In fact the Geekbench scores on macOS were on average higher! (9800-10,000 on Geekbench 5)

This was the Balanced setting scores on Windows 11:

and High Performance setting scores on Windows 11: