Alder Lake CPUs common discussion

[PolRoger]

looks almost like alder lake "K" CPUs get treated like non K SKUs.

View attachment 224647

What else would you expect? Infinite CPU clock?

that i get an unlocked CPU when i buy an unlocked CPU?

It is unlocked, but it will only run X MHz by default

I'm wondering if this is now a BIOS related issue for Gigabyte motherboards? Or is this some kind of Intel ADL "limit" that applies to all of the other motherboard makers? If ASUS, MSI, ASRock can design a BIOS to circumvent a type of ADL default rules like what we are seeing here then GIgabyte should also be able to come up with a BIOS that does the same?

EDIT:
Maybe this is just Gigabyte BIOS issue and is related to AVX??...

AIDA stress utilizes AVX and so Does Prime95 and they both will drop the clocks on various cores while stressing with manual all core OC. Disable AVX instructions on Prime95 and I'm not seeing the drop in core speeds. WCG which I believe doesn't use AVX will also run without lowering clocks on all core with E-cores disabled (which is due to WCG W10 scheduling bug).

 

[GerKNG]

I'm wondering if this is now a BIOS related issue for Gigabyte motherboards? Or is this some kind of Intel ADL "limit" that applies to all of the other motherboard makers? If ASUS, MSI, ASRock can design a BIOS to circumvent a type of ADL default rules like what we are seeing here then GIgabyte should also be able to come up with a BIOS that does the same?

EDIT:
Maybe this is just Gigabyte BIOS issue and is related to AVX??...

AIDA stress utilizes AVX and so Does Prime95 and they both will drop the clocks on various cores while stressing with manual all core OC. Disable AVX instructions on Prime95 and I'm not seeing the drop in core speeds. WCG which I believe doesn't use AVX will also run without lowering clocks on all core with E-cores disabled (which is due to WCG W10 scheduling bug).

optimum techs 12600k review has the same problem.
it's on a z690 Unify.

the throttling happens completely random.
sometimes at idle on the desktop, ALL the time in cod vanguard for some reason.
NEVER in Warzone.
every couple minutes in WoW.

i have every single option in the bios set to "unlock" every single thing that might throttle the CPU.
above 4.8 Ghz = down to 4.8 Ghz.
at 4.8 Ghz = stays for ever at 4.8 Ghz.

looks more like a bug to me tbh.

 

[phanbuey]

optimum techs 12600k review has the same problem.
it's on a z690 Unify.

the throttling happens completely random.
sometimes at idle on the desktop, ALL the time in cod vanguard for some reason.
NEVER in Warzone.
every couple minutes in WoW.

i have every single option in the bios set to "unlock" every single thing that might throttle the CPU.
above 4.8 Ghz = down to 4.8 Ghz.
at 4.8 Ghz = stays for ever at 4.8 Ghz.

looks more like a bug to me tbh.

Really does sound like a bug... I am going to try COD vanguard to see if this is happening for me.

 

[unclewebb]

@GerKNG - The IA: Max Turbo Limit throttling reason can get triggered when a CPU is idle. This one is usually not that important or meaningful.

Run a full load stress test. Take a screenshot of the limit reasons while the CPU is fully loaded and while the CPU is throttling.

@PolRoger - Your screenshot indicates that it might be related to IccMax or PL4. Have a look in the BIOS to see what IccMax is set to. Setting IccMax to the maximum, 255.75, used to work to eliminate this type of throttling for 10th and 11th Gen CPUs. I am not sure if 12th Gen use the same maximum. Both the CPU core and cache need to be set to the same value. If there is a bug in the BIOS, it might only be setting the core properly and not the cache / ring.

It might not be possible to adjust PL4 in the BIOS but have a look for that setting.

 

[GerKNG]

@GerKNG - The IA: Max Turbo Limit throttling reason can get triggered when a CPU is idle. This one is usually not that important or meaningful.

Run a full load stress test. Take a screenshot of the limit reasons while the CPU is fully loaded and while the CPU is throttling.

View attachment 224664

@PolRoger - Your screenshot indicates that it might be related to IccMax or PL4. Have a look in the BIOS to see what IccMax is set to. Setting IccMax to the maximum, 255.75, used to work to eliminate this type of throttling for 10th and 11th Gen CPUs. I am not sure if 12th Gen use the same maximum. Both the CPU core and cache need to be set to the same value. If there is a bug in the BIOS, it might only be setting the core properly and not the cache / ring.

It might not be possible to adjust PL4 in the BIOS but have a look for that setting.

Your PL4 limit is for the uncore/Ring
This is constantly a thing because it is dynamic to 4.5GHz but only runs at 3.6 under load.

My throttle reason is in the screenshot i posted. Max turbo frequency ( just like when it would be a non K chip)

 

[phanbuey]

I dont actually know what's going on anymore....

I just keep increasing multiplier and the numbers keep going up. Scores starting to get to 'joke' territory.

 

[RandallFlagg]

View attachment 224683


I dont actually know what's going on anymore....

I just keep increasing multiplier and the numbers keep going up. Scores starting to get to 'joke' territory.

You are getting close to +50% over my 10850K in single thread CPU-Z.

Can you tell a difference in the UI response with that massive single thread performance, or was it already in the realm of not noticeable to normal humans?

 

[phanbuey]

You are getting close to +50% over my 10850K in single thread CPU-Z.

Can you tell a difference in the UI response with that massive single thread performance, or was it already in the realm of not noticeable to normal humans?

I think it's mostly placebo but it for sure feels faster .

5.4 failed unfortunately - died after some time with the LinPack... back down to 5.3 now and going to run it for a few days before moving on.

Havent touched the E cores yet but in 3900x cinebench mt territory.

 

[RandallFlagg]

> 2000 single core in Cinebench R23 with a 12600K, good grief.

I guess the days where the XX600K were just defective XX900K chips and generally couldn't clock as high are gone. That's a different die on that chip from the 12700K/12900K.

 

[tabascosauz]

> 2000 single core in Cinebench R23 with a 12600K, good grief.

I guess the days where the XX600K were just defective XX900K chips and generally couldn't clock as high are gone. That's a different die on that chip from the 12700K/12900K.

Isn't the other type of ADL die only 6+0, so 12600K is still just the same die as the other two? Nevertheless, very impressive.

I'm always jealous how comprehensive the performance data and reporting (IA limit reasons etc) are on Intel since even Haswell.

 

[RandallFlagg]

Isn't the other type of ADL die only 6+0, so 12600K is still just the same die as the other two? Nevertheless, very impressive.

I'm always jealous how comprehensive the performance data and reporting (IA limit reasons etc) are on Intel since even Haswell.

I think the 12600K is the top tier of the 6+4 die. The yet to be released lower SKUs like 12400 6+0 is supposedly a defective 12600K.

From what I understand, Intel has two dies one is 8+8 (12900K) and its lower rated skus and defects, the other 6+4 (12600K) and its lower rated SKUs defects.

I put it rather harshly, but this is ofc the way all the vendors work including AMD and NVidia.

In short, the old way is why top SKUs usually had the highest OC potential in single thread, despite having more points of failure (more cores). This is true of Zen also. In this new model, the 12600K may wind up hitting the highest Ghz / OC, since it has fewer things to fail as it is a 6+4 instead of 8+8 die to start with.

 

[tabascosauz]

From what I understand, Intel has two dies one is 8+8 (12900K) and its lower rated skus and defects, the other 6+4 (12600K) and its lower rated SKUs defects.

Do you have a source for this? All I can find are various outlets all unanimously reporting on the same MSI ADL overview, which showed one 8+8 and one 6+0. Made a big hubbub about it because the dies are located a bit differently so the hotspot may be different on upcoming 6+0:

 

[RandallFlagg]

Do you have a source for this? All I can find are various outlets all unanimously reporting on the same MSI ADL overview, which showed one 8+8 and one 6+0. Made a big hubbub about it because the dies are located a bit differently so the hotspot may be different on upcoming 6+0:

Yeah I may be mistaken, I thought I had seen a die shot of the 6+4 that showed 6 golden coves and a 4 core gracemont cluster. Looking again all I find is the same as your post :

 

[phanbuey]

Good to know either way.

 

[PolRoger]

I think it's mostly placebo but it for sure feels faster .

5.4 failed unfortunately - died after some time with the LinPack... back down to 5.3 now and going to run it for a few days before moving on.


View attachment 224689

Havent touched the E cores yet but in 3900x cinebench mt territory.

I don't think my sample is as strong as yours... Mine needs more voltage for ~100MHz less at ~52x... (52P/38E)

Also... If people think there're going to be running these new ADL cpu at 50x to 53x (all core) with long term heavy loads on average air coolers and thin 240 AIO's... I think they will be disappointed.

CPUZ bench (52P/38E):

Stock defaults with XMP enabled Gear 1:

 

[unclewebb]

Mine needs more voltage

This is not a fair comparison. CPU-Z screenshot that @phanbuey posted is showing Core VID voltage.

Your screenshot shows actual Core Voltage. That is the important number.

 

[RandallFlagg]

I don't think my sample is as strong as yours... Mine needs more voltage for ~100MHz less at ~52x... (52P/38E)

Also... If people think there're going to be running these new ADL cpu at 50x to 53x (all core) with long term heavy loads on average air coolers and thin 240 AIO's... I think they will be disappointed.

CPUZ bench (52P/38E):

View attachment 224705


Stock defaults with XMP enabled Gear 1:

View attachment 224706

That's actually a hefty boost from default, looks like +10% single and Multi on ZPU-Z.

Can you guys show the peak power / temp while running that?

This is my 10850K- note I've got more running than this as it's a workday so it's not a clean bench but is ballpark. It basically runs 210W during the MT test :

 

[phanbuey]

My friend just picked up 12700K well and it's overheating like crazy on his AIO... here is his 12700K at 240W-280W at 4:9 all core, cannot get 5.0 stable at the moment, here is his cinebench @ 4.9 95C temps:

Seems the 12700Ks are much harder to push? W1zz's was 5.0 at 1.4V ~ seeing 3 other reviews of 12700K deeming it 'not worth' to overclock. Could just be the lottery/bioses? Seems weird for 2 more cores to have such a difference (altho TPU 12600K also didn't really OC).

@PolRoger -- what motherboard are you running?

 

[RandallFlagg]

My friend just picked up 12700K well and it's overheating like crazy on his AIO... here is his 12700K at 240W-280W at 4:9 all core, cannot get 5.0 stable at the moment, here is his cinebench @ 4.9 95C temps:


View attachment 224710

Seems the 12700Ks are much harder to push? W1zz's was 5.0 at 1.4V ~ seeing 3 other reviews of 12700K deeming it 'not worth' to overclock. Could just be the lottery/bioses? Seems weird for 2 more cores to have such a difference (altho TPU 12600K also didn't really OC).

@PolRoger -- what motherboard are you running?

240W is insane for 4.9. It sounds like the BIOS isn't tuned well to me. He is just running 4.9 on the P-cores right? Just sayin...

PolRoger is using an Aorus Ultra ITX :

 

[Devon68]

reviews deeming it 'not worth' to overclock.

It might get even better after the Beta testing is done , new motherboards might come out as well as better coolers .

 

[Outback Bronze]

Interesting read here for Alder Lake CPU's:

Maximus Z690 and Alder Lake: Modern CPU's require Modern Overclocking Solutions. - CPUs, Motherboards, and Memory - Linus Tech Tips

 

[JustBenching]

Fair enough. Sometimes I can't read. The 5600X did spank the 10900K in all but a few games though.

Nah, it didn't. I mean, in stock maybe, yeah. The 10900k has an insane overclocking headroom, be it cache at 5.0ghz, ram at 4400c16-4600c16, freequency at 5.1-5.3 ghz. The 5600x doesn't hold a candle. And yes, I assume people that spend ~500 for a 10900k, will oc the crap of it. I have some numbers from my 10900k in ingame benchmarks that I can share with you, it regularly beats / trades blows with a highly tuned 5950x.

 

[mstenholm]

5600X and 10900K. Headline is Adler Lake. Take your disagreements elsewhere please.

 

[TheoneandonlyMrK]

5600X and 10900K. Headline is Adler Lake. Take your disagreements elsewhere please.

Probably Should have made it a club then, no offense intended.

 

[phanbuey]

@RandallFlagg -- Everything identical except motherboard + Cpu and bios settings.

12600K @ 5.3 ghz 40 ring 3600mhz Gear 1 Cl 16 not yet tuned: EDIT: i set it to lowest not sure why it said custom... will rerun soon but i didn't touch any settings between the two.

10850K @ 5.1Ghz - 4000mhz ram CL17 4x8GB tuned timings, 44 ring:

The 12600K is super fast and powerful... but man those old Skylakes are still fast AF. Enjoy your 10850

12600K

10850 was around 7.4 seconds.