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.

 

[epsilon84]

The stock configuration for the 12900K is 125W TDP, Intel says so very clerly in their 12th Gen Datasheet:
View attachment 59543

They're also giving recommendations for Tau timers that limit time spent by the CPU @ PL2=241 in order for the CPU to actually behave like a 125W TDP CPU in the long-run:
View attachment 59544

So if your intent is to advocate that the 12900K be judged at the peak of its performance then by all means, don't let me interrupt. But please stop promoting the idea that stock config for 12900K is 241W TDP, it's even more ludicrous to say that considering the 12900KS is defined by Intel as a 150W TDP SKU.

125W is by definition measured at the base clock (3.9GHz) EDIT: (Actually, base clock is 3.2GHz) If you want to insist that a 12900K be judged on its base clock configuration, then by all means, don't let me interrupt you either.

Intel Core i9-12900K Alder Lake Tested at Power Limits between 50 W and 241 W

We test Intel's Core i9-12900K at various TDP levels all the way down to 50 W to determine how much efficiency is really in the new Alder Lake core, and how these power limits affect performance. Competing with the efficiency of AMD's Zen 3 Ryzen lineup is just two settings changes away.

www.techpowerup.com

Meanwhile, in the real world:

For the i9-12900K, this is 241 W. Unlike past generations of processors that were constrained by the Tau time value to hold maximum power draw, or PL2, Alder Lake processors now run at maximum power draw indefinitely if the load demands it and as long as the processor doesn't hit the thermal limit of 105°C. This is done without inventing a new system; Intel simply tweaked the PL1 and PL2 values and set them both to 241 W, which effectively means the processor can run at 241 W all the time as long as it doesn't overheat. The "125 W" limit now only exists on paper and in marketing documents.

My personal usage also backs this up. Not sure what else there is to say on this. For all intents and purposes, a 12900K has a 241W TDP... or unlimited PL2 of 241W if you want to argue semantics.

 

[Thunder 57]

Again, I re-iterate my point earlier, but on the GPU side: Do AMD fans ask for a 450W 3090 Ti to be power limited to 300W so to make it 'fair' against a 300W 6900XT? See how absurd that argument becomes? Yes, you *can* artificially cap a 3090 Ti to 300W, in which case it may well perform worse than a stock 6900XT. But good luck finding a reviewer to do that...

Of course not, because NVIDIA isn't trying to say it's really only 300W because the 3090Ti has a base clock of x. As you later go on to point out, Intel screws around with it's TDP to make it willfully confusing.

Also, anyone who disagrees with you has to be an "AMD fan"? Nice, I guess no one can try to disagree with you objectively.

 

[JoeRambo]

125W is by definition the base clock (3.9GHz) TDP. If you want to insist that a 12900K be judged on its base clock configuration, then by all means, don't let me interrupt you either.

That is not true and has nothing to do with "clocks" and if You took time to look at your benchmark source, You would see that 125W/125W config is beating 5950x in multiple benchmarks and is very near 241W/241W setup in others. The clocks CPU will run at certain wattage depends on how taxing workload actually is. The actual definition of 'base clock" is very esoteric and is not even worse case, rather than mix of workloads known only to Intel are guaranteed to run no less than that clocks @125W.

12900K is pushed beyond efficient clocks and fed insane volts to win MT benchmarks, but whole lineup should not be judged by stock configuration of SKU designed by marketing morons.

 

[epsilon84]

Of course not, because NVIDIA isn't trying to say it's really only 300W because the 3090Ti has a base clock of x. As you later go on to point out, Intel screws around with it's TDP to make it willfully confusing.

Also, anyone who disagrees with you has to be an "AMD fan"? Nice, I guess no one can try to disagree with you objectively.

It was a general comment, not aimed at any poster in particular.

I guess if you don't understand the difference between base power and turbo power then the 125W/241W difference can be confusing:

Intel® Core™ i9-12900K Processor (30M Cache, up to 5.20 GHz) Product Specifications

Intel® Core™ i9-12900K Processor (30M Cache, up to 5.20 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.

ark.intel.com

• Processor Base Power 125 W
• Maximum Turbo Power 241 W

Again, I'll repeat: For all intents and purposes, a 12900K will consume up to 241W indefinitely when under full load - this is by design.

Asking for it to be capped at 125W, or 142W, or whatever arbitrary figure, means it is then measured in a non stock configuration.

This has already been done and the data is there for anyone interested, which I'll link again:

Intel Core i9-12900K Alder Lake Tested at Power Limits between 50 W and 241 W

We test Intel's Core i9-12900K at various TDP levels all the way down to 50 W to determine how much efficiency is really in the new Alder Lake core, and how these power limits affect performance. Competing with the efficiency of AMD's Zen 3 Ryzen lineup is just two settings changes away.

www.techpowerup.com

 

[epsilon84]

That is not true and has nothing to do with "clocks" and if You took time to look at your benchmark source, You would see that 125W/125W config is beating 5950x in multiple benchmarks and is very near 241W/241W setup in others. The clocks CPU will run at certain wattage depends on how taxing workload actually is. The actual definition of 'base clock" is very esoteric and is not even worse case, rather than mix of workloads known only to Intel are guaranteed to run no less than that clocks @125W.

12900K is pushed beyond efficient clocks and fed insane volts to win MT benchmarks, but whole lineup should not be judged by stock configuration of SKU designed by marketing morons.

Well, that is Intel's definition of 'base power':

Processor Base Power
The time-averaged power dissipation that the processor is validated to not exceed during manufacturing while executing an Intel-specified high complexity workload at Base Frequency and at the junction temperature as specified in the Datasheet for the SKU segment and configuration.

Upon actually looking this up, I realise I got the base frequency wrong in my previous post, for some reason I thought it was 3.9GHz, it is in fact 3.2GHz as stated on Intel's ARK page.

Performance-core Base Frequency 3.20 GHz

I can't confirm or deny these figures as I have never seen my 12900K clock anywhere near 3.2GHz under normal usage. I'm guessing you will have to have power limits set at 125W and perform some form of AVX workload for it to be running in this 'worst case' scenario.

 

[Zucker2k]

The stock configuration for the 12900K is 125W TDP, Intel says so very clerly in their 12th Gen Datasheet:
View attachment 59543

They're also giving recommendations for Tau timers that limit time spent by the CPU @ PL2=241 in order for the CPU to actually behave like a 125W TDP CPU in the long-run:
View attachment 59544

So if your intent is to advocate that the 12900K be judged at the peak of its performance then by all means, don't let me interrupt. But please stop promoting the idea that stock config for 12900K is 241W TDP, it's even more ludicrous to say that considering the 12900KS is defined by Intel as a 150W TDP SKU.

It seems you arrived late at the party, typically. Where have you been all these months?

 

[tamz_msc]

Intel's definition of TDP isn't a problem. They've been more transparent than before by actually specifying PL2 this generation. They're far better than AMD who don't specify the PPT - which is the actual max power use - in their product page.

The problem is that Intel gives too much leeway to motherboard makers by not enforcing those definitions and allowing them to do as they please.

 

[coercitiv]

125W is by definition the base clock (3.9GHz) TDP.

There is no other TDP. Show me the "241W TDP" wording in official Intel spec.

If you want to insist that a 12900K be judged on its base clock configuration, then by all means, don't let me interrupt you either.

Again, Intel defines the stock specs for the 12900K as follows:
PL1 - 125W
PL2 - 241W
Tau - 56s

This means they recommend the CPU be allowed to boost to 241W for a limited period of 56 seconds, after which it should be throttled to 125W for the remaining duration of the workload.

By contrast the 12900 has the following parameters:
PL1 - 65W
PL2 - 202W
Tau - 28s

Notice the considerable cut in Tau timer, they have clearly thought this through in terms of stock specs.

My personal usage also backs this up. Not sure what else there is to say on this. For all intents and purposes, a 12900K has a 241W TDP... or unlimited PL2 of 241W if you want to argue semantics.

It's only semantics until you start accusing other forum members of moving the goalposts based on your personal usage when it should be datasheets. It's bad enough Intel has managed to confuse the heck out of everyone with their hocus-pocus TDP/Base Power/Maximum Turbo Power shenanigans, we don't need personal interpretation of TDP numbers thrown into the mix.

I clearly stated I have no problem with the argument of testing the 12900K at top performance settings (PL1=PL2=241W). I also asked you not to state this as "stock configuration" because it isn't. Let's not help normalize Intel's disingenuous behavior on this matter.

Meanwhile I kindly ask you to consider that 5950X @ stock settings clocks under 4Ghz under MT load while a ~250W configuration can clock at 4.6-4.7Ghz. There's room for higher dynamics on AM4 side as well, only this vendor chose to enforce stock power specs for their product.

 

[coercitiv]

It seems you arrived late at the party, typically. Where have you been all these months?

I've been playing with various ADL-S core & power configurations, overclocking RAM on early Z690 firmware and discussing V/f curves and scheduler behavior with other members. How about you?

 

[epsilon84]

There is no other TDP. Show me the "241W TDP" wording in official Intel spec.

It appears Intel actually stopped using the term 'TDP' on their ark pages, it is now simply classed as power, whether it be @ base frequency or turbo freqeuncy.

• Processor Base Power 125 W
• Maximum Turbo Power 241 W

Which I think is actually a lot less confusing than a generic '125W TDP'.

So your argument is that a 12900K that runs at 241W indefinitely (as most motherboard makers seem to set it) is not 'stock' behaviour?

So when the vast majority of users install a 12900K, and it runs at 241W, are you claiming it is some form of unauthorised / unofficiall 'MCE' (multicore enhancement) unwittingly forced upon them by motherboard makers ignoring official Intel specs?

I'm just trying to clarify your position on this, because I may have to agree to disagree with you about what we consider is 'stock' for a 12900K.

As far I I'm concerned, upon installation and first boot up, without touching any BIOS settings, my 12900K ran at 241W PL2 uncapped, and it seems this is the default setting for the vast majority of Z690 boards.

 

[Zucker2k]

I've been playing with various ADL-S core & power configurations, overclocking RAM on early Z690 firmware and discussing V/f curves and scheduler behavior with other members. How about you?

Countering the FUD being spread in this very thread that your 12700k consumes 300 watts consistently in multithreaded workloads. Seen those posts?





Inflammatory posting is not allowed here but you already knew that.


esquared
Anandtech Forum Director

 

[Zucker2k]

It appears Intel actually stopped using the term 'TDP' on their ark pages, it is now simply classed as power, whether it be @ base frequency or turbo freqeuncy.

• Processor Base Power 125 W
• Maximum Turbo Power 241 W

Which I think is actually a lot less confusing than a generic '125W TDP'.

So your argument is that a 12900K that runs at 241W indefinitely (as most motherboard makers seem to set it) is not 'stock' behaviour?

So when the vast majority of users install a 12900K, and it runs at 241W, are you claiming it is some form of unauthorised / unofficiall 'MCE' (multicore enhancement) unwittingly forced upon them by motherboard makers ignoring official Intel specs?

I'm just trying to clarify your position on this, because I may have to agree to disagree with you about what we consider is 'stock' for a 12900K.

As far I I'm concerned, upon installation and first boot up, without touching any BIOS settings, my 12900K ran at 241W PL2 uncapped, and it seems this is the default setting for the vast majority of Z690 boards.

It's the TAU = Indefinite Time part that is enforced by mobo makers. It's not official Intel spec.

 

[DrMrLordX]

Because the moment you start changing parameters on a CPU and out of its stock configuration, you're just moving the goal posts to suit your agenda, and its not really an apples to apples stock vs stock comparison anymore.

If your agenda is "I don't want to deal with 241W of heat output under Blender" or whatever, how is that invalid? Plenty of people are in that boat. Also see comments from @coercitiv but let's be honest, enough boards allow a limitless tau and/or different PL1 values making the Intel specs wishful thinking.

 

[tamz_msc]

As far I I'm concerned, upon installation and first boot up, without touching any BIOS settings, my 12900K ran at 241W PL2 uncapped, and it seems this is the default setting for the vast majority of Z690 boards.

That's stock behavior, but not the in-spec behavior. The problem is that in-spec behavior is not enforced; out-of-spec behavior is allowed by Intel. In particular, they don't view unlimited tau as overclocking.

MCE is something different altogether, it's a motherboard setting not something that Intel specifies.

 

[epsilon84]

That's stock behavior, but not the in-spec behavior. The problem is that in-spec behavior is not enforced; out-of-spec behavior is allowed by Intel. In particular, they don't view unlimited tau as overclocking.

MCE is something different altogether, it's a motherboard setting not something that Intel specifies.

OK, then I am willing to concede that expected 'stock behaviour' and actual 'stock specs' are 2 different things.

I considered the uncapped 241W PL2 as 'stock' since that is how my 12900K ran out of the box.

Since we are on the topic of power consumption, I'd like to see what other people here think of the claims in this video that a 12900K is more power efficient than a 5950X under a 'mixed use' scenario?

it certainly goes against the grain of what most reviewers and users say WR to power use. Not trying to start a flame war, I'm just geniunely surprised at these results and his conclusion. Does he have merit here?

 

[DrMrLordX]

Since we are on the topic of power consumption, I'd like to see what other people here think of the claims in this video that a 12900K is more power efficient than a 5950X under a 'mixed use' scenario?

Absolutely, a 5950X has high base power draw thanks to I/O requirements. Nothing Intel produces has that problem. You're really not taxing either CPU if both of them are staying below 100W. Also, for workloads in that power envelope, the 12900k will probably win everything. Or nearly everything.

 

[lobz]

125W is by definition measured at the base clock (3.9GHz) EDIT: (Actually, base clock is 3.2GHz) If you want to insist that a 12900K be judged on its base clock configuration, then by all means, don't let me interrupt you either.

Intel Core i9-12900K Alder Lake Tested at Power Limits between 50 W and 241 W

We test Intel's Core i9-12900K at various TDP levels all the way down to 50 W to determine how much efficiency is really in the new Alder Lake core, and how these power limits affect performance. Competing with the efficiency of AMD's Zen 3 Ryzen lineup is just two settings changes away.

www.techpowerup.com

Meanwhile, in the real world:


My personal usage also backs this up. Not sure what else there is to say on this. For all intents and purposes, a 12900K has a 241W TDP... or unlimited PL2 of 241W if you want to argue semantics.

You're dead wrong, but the problem is, you're having a factual debate with intel and not with us AMD fans, yet you're lashing out your personal opinions not at Intel. Strange.

 

[JoeRambo]

Guys, why is the focus on 12900K? Or 12900KS? These are marketing CPUs, period. Meant for your average incompetent reviewer, that runs CB23 and calls it a day, or just reports GB5 ST/MT and that's it. To win the benchmarks at whatever cost.

For example 12900KS is using 49 watts of power in single threaded workload. Pumping insane volts to reach 5.5Ghz. That is FIFTY WATTS guys.

So any results that involve these ridiculous SKUs are irrelevant in any discussion that involves "efficiency", be it single threaded, low threaded or full on load like rendering.
I think it is very impressive, that in some tests like Phoronix, 12900K does incredibly well even in this "marketing moron" edition both in performance and efficiency overall.

 

[epsilon84]

Guys, why is the focus on 12900K? Or 12900KS? These are marketing CPUs, period. Meant for your average incompetent reviewer, that runs CB23 and calls it a day, or just reports GB5 ST/MT and that's it. To win the benchmarks at whatever cost.

For example 12900KS is using 49 watts of power in single threaded workload. Pumping insane volts to reach 5.5Ghz. That is FIFTY WATTS guys.

So any results that involve these ridiculous SKUs are irrelevant in any discussion that involves "efficiency", be it single threaded, low threaded or full on load like rendering.
I think it is very impressive, that in some tests like Phoronix, 12900K does incredibly well even in this "marketing moron" edition both in performance and efficiency overall.

Ouch, does that mean I'm a fool for marketing as a 12900K owner? LOL. I get your point though, it was designed to win benchmarks at whatever cost (to power), and the 12900KS just pushes that to the extreme.

Insane wattage for performance crowns seems to be the trend lately, at least when it comes to Intel and nVidia. AMD is relatively much better behaved in comparison!

What is a more suitable SKU to focus on then? A 12700? 12400? 12100?

 

[JoeRambo]

I am fellow 12900K owner and i've bought it to get unlocked SKU with 30MB of L3 and a good bin of cores and IMC. I ran it stock, laughed when it went ballistic with power usage and thermals and then tweaked it to my liking.

 

[epsilon84]

I am fellow 12900K owner and i've bought it to get unlocked SKU with 30MB of L3 and a good bin of cores and IMC. I ran it stock, laughed when it went ballistic with power usage and thermals and then tweaked it to my liking.

In my defence, I bought it for a killer price of $400 ($550 AUD) off someone who bins their own 12900Ks and mine unfortunatlely (fortunately for me) didn't make the grade with a lowly SP80 rating. How accurate are these ratings anyway? I haven't tried overclocking, seems rather pointless on a CPU thats already pushed to its limits. I do miss the old days of overclocking, not gonna lie..

What kind of tweaks did you make to your 12900K? I'm running it stock but with a 50mV undervolt, will try for a bit more when I have some free time to see how 'efficient' I can make it. That seems more worthwhile than trying to squeeze out another 200MHz for an extra 100W of heat LOL

 

[JoeRambo]

What kind of tweaks did you make to your 12900K? I'm running it stock but with a 50mV undervolt, will try for a bit more when I have some free time to see how 'efficient' I can make it.

Disabled E-Cores, fixed P-Core clock to 5Ghz, set uncore to 4.5Ghz, set "whatever this mode is called in MSI BIOS" => 1.2V voltage and -0.145V offset. Ends up running 1.15-1.2V under full load.
Most important ones are memory tuning, too many to list, but end result is DDR4 3800CL15 with timings like these:

 

[epsilon84]

Disabled E-Cores, fixed P-Core clock to 5Ghz, set uncore to 4.5Ghz, set "whatever this mode is called in MSI BIOS" => 1.2V voltage and -0.145V offset. Ends up running 1.15-1.2V under full load.
Most important ones are memory tuning, too many to list, but end result is DDR4 3800CL15 with timings like these:

View attachment 59546

Thanks, I'm surprised you're stable at 5GHz with such a low voltage! How much power are the P cores pulling?

I have some Trident Z 3600 B Dies so might try for those timings as a baseline. I'm guessing you're also running B Dies to get those kind of timings at DDR4-3800? What kind of voltage are you running for the RAM?

Finallly, how much do those timings improve performance compared to your run stock standard XMP timings?

 

[JoeRambo]

I am using one of the top bins of dual rank B-DIE => 4000CL16 from G.Skill. They don't even work XMP, so no idea of the difference and i would not recommend anyone touching XMP in BIOS for anything above 3200, as motherboards will set nasty SA / IO voltages and use incredibly relaxed secondary/tertiary timings to ruin performance and drive thermals up. Time consuming manual tuning is required. I am at 1.28SA/ 1.25 IO TX / 1.47V VDIMM.
I think for ADL top performance starts ~45ns of memory latency and anything below 50ns is palatable (AIDA testing).

 

[epsilon84]

I am using one of the top bins of dual rank B-DIE => 4000CL16 from G.Skill. They don't even work XMP, so no idea of the difference and i would not recommend anyone touching XMP in BIOS for anything above 3200, as motherboards will set nasty SA / IO voltages and use incredibly relaxed secondary/tertiary timings to ruin performance and drive thermals up. Time consuming manual tuning is required. I am at 1.28SA/ 1.25 IO TX / 1.47V VDIMM.
I think for ADL top performance starts ~45ns of memory latency and anything below 50ns is palatable (AIDA testing).

Fair enough, I've never tweaked the secondary / tertiary timings because it seemed so tedious to test so many different settings for stability etc for minor performance gains.

I'll play around with the sub timings a bit then and use your settings as a rough guide but with slightly more relaxed timings, no idea if my B Dies are near the top bins or not, guess I'll find out.

Cheers