Specs of Top Intel 12th Gen Core "Alder Lake-S" Processors Surface

[btarunr]

Intel will debut its 12th Gen Core "Alder Lake-S" desktop processors either toward the end of 2021, or early 2022, introducing the LGA1700 socket, 600-series chipset, and more importantly, hybrid CPU core architecture to the desktop space. The 10 nm "Alder Lake-S" silicon features up to eight "Golden Cove" performance cores (P-cores), and up to eight "Gracemont" efficiency cores (E-cores), in a heterogenous CPU core setup rivaling Arm big.LITTLE. Specifications of the top Core i9, fairly-top Core i7, and mid-tier Core i5 parts were leaked to the web on Chinese social media.

The 12th Gen Core lineup will be led, predictably, by the Core i9-12900K, which succeeds the i9-11900K with a maxed out 8+8 (P+E) configuration, unlocked multipliers, the most cache, and the highest clock speeds. The P-cores ("Golden Cove" cores) are clocked up to 5.30 GHz (1-2 cores boost), and up to 5.00 GHz all-core / 8 cores; while the E-cores ("Gracemont" cores), are clocked up to 3.90 GHz (1-4 cores boost), with 3.70 GHz all-core / 8 cores boost. The total L3 cache on the silicon is 30 MB. The i9-12900K has a TDP of 125 W (PL1), with 228 W PL2. Intel will introduce several new overclocking features, including multiple memory gear ratios.



The 12th Gen Core i7 processors will be 8+4 core (P+E), have slightly lower clock speeds than the Core i9 parts, and possibly miss certain boost features. The Core i7-12700K will be the top part in the Core i7 extension. The P-cores feature maximum boost frequency of 5.00 GHz, with 4.70 GHz all-core boost; while the E-cores 3.80 GHz max boost, with 3.60 GHz all-core. The chip has 25 MB of L3 cache, and identical PL1 and PL2 values to the i9-12900K.

The Core i5 series will be made up to 6+4 core (P+E) parts. The unlocked Core i5-12600K is runs the P cores at up to 4.90 GHz, with up to 4.50 GHz all-core; while the E-cores are run at up to 3.60 GHz, with up to 3.40 GHz all-core. 20 MB of L3 cache, and the same 125 W PL1 and 228 W PL2 values as the other two unlocked parts.

There could be variants of the three that lack iGPU, but have identical CPU specs—i9-12900KF, i7-12700KF, and i5-12600KF.

View at TechPowerUp Main Site

 

[Crackong]

Hey there 228W PL2

 

[Prima.Vera]

cannot wait to see how an old 10900K CPU will trash this i9 garbage of 8 Cores + 8 Useless Cores... Specially in games or render apps.

 

[Space Lynx]

cannot wait to see how an old 10900K CPU will trash this i9 garbage of 8 Cores + 8 Useless Cores... Specially in games or render apps.

or better yet, use the weak cores by accident in old old games, limiting max frame rate possibilities for 240hz monitors.

 

[watzupken]

Hey there 228W PL2

I wonder if this is the upper limit, or just some average number than Intel pulled out. To be honest, it is very hard to take their word for it nowadays. We've seen too many numbers from Intel of late, that don't really reconcile with the actual product. The TDP number is by far the most misleading number because the actual product requires between 2 to 3.5x that amount of power to reach that kind of performance that Intel is tooting.

cannot wait to see how an old 10900K CPU will trash this i9 garbage of 8 Cores + 8 Useless Cores... Specially in games or render apps.

I doubt that will happen. Even with the current Rocket Lake, the IPC gain is noticeable over Comet Lake. And if you take Tiger Lake H vs the older Comet Lake H, the gains is even more pronounced in games and in rendering software. While the Alder Lake is capped at 8 performance Golden Cove cores, the efficient cores from what I heard is almost as fast as Skylake (not sure if this is true until we hear more about it closer to launch). So if you have 8 efficient cores helping to make up for the 2 missing high performance cores in the 10900K, I think it is highly possible that Alder Lake will be quite a bit faster. The only hurdle for Intel is AMD's Ryzen 9 series.

 

[dj-electric]

E core and P core are official terms. This seems legit.

 

[nguyen]

So with windows 11 we can choose which application require high performance cores, similar to iGPU vs dGPU?
The only way 8+8cores config like this would sell is that Intel charge similarly to previous 8C CPU and not make a new pricing bracket (i11 and i13 SKUs )

 

[Uroshi]

I wonder if this is the upper limit, or just some average number than Intel pulled out. To be honest, it is very hard to take their word for it nowadays. We've seen too many numbers from Intel of late, that don't really reconcile with the actual product. The TDP number is by far the most misleading number because the actual product requires between 2 to 3.5x that amount of power to reach that kind of performance that Intel is tooting.


I doubt that will happen. Even with the current Rocket Lake, the IPC gain is noticeable over Comet Lake. And if you take Tiger Lake H vs the older Comet Lake H, the gains is even more pronounced in games and in rendering software. While the Alder Lake is capped at 8 performance Golden Cove cores, the efficient cores from what I heard is almost as fast as Skylake (not sure if this is true until we hear more about it closer to launch). So if you have 8 efficient cores helping to make up for the 2 missing high performance cores in the 10900K, I think it is highly possible that Alder Lake will be quite a bit faster. The only hurdle for Intel is AMD's Ryzen 9 series.

When they said the Gracemont cores are 90% of Skylake I wonder if they meant 90% of the 6700K cores (4.2GHz single core turbo, 4GHz all core). That wouldn't be so impressive to be honest.

 

[Selaya]

[ ... ]
I doubt that will happen. Even with the current Rocket Lake, the IPC gain is noticeable over Comet Lake. And if you take Tiger Lake H vs the older Comet Lake H, the gains is even more pronounced in games and in rendering software. While the Alder Lake is capped at 8 performance Golden Cove cores, the efficient cores from what I heard is almost as fast as Skylake (not sure if this is true until we hear more about it closer to launch). So if you have 8 efficient cores helping to make up for the 2 missing high performance cores in the 10900K, I think it is highly possible that Alder Lake will be quite a bit faster. The only hurdle for Intel is AMD's Ryzen 9 series.

No.

IPC is fairly useless a metric for gaming performance. IPC isn't the bottleneck for game fps, not by a long shot. Memory (latency, primarily) is. Which is also the strongest selling point of Skylake, it's got the best IMC since forever basically. That's also why Skylake was holding its ground and at the beginning even eclipsing RKL in game performance - it's got a much better IMC.

 

[Crackong]

I wonder if this is the upper limit, or just some average number than Intel pulled out. To be honest, it is very hard to take their word for it nowadays. We've seen too many numbers from Intel of late, that don't really reconcile with the actual product. The TDP number is by far the most misleading number because the actual product requires between 2 to 3.5x that amount of power to reach that kind of performance that Intel is tooting.

Intel's TDP numbers are far beyond fixing so everybody looks for their "TrUe" power limit numbers now (PL1 and PL2)
PL1 and PL2 number were so far OKAY in varies reviews and testing, generally their CPUs never exceed the PL2 limit without overclocking.

 

[bug]

I'm curious to see how this heterogeneous architecture works. It seems 10nm will kill it regardless, but I'm still curious from a technical point of view. It could be a glimpse of things to come.

 

[AusWolf]

I wonder if this is the upper limit, or just some average number than Intel pulled out. To be honest, it is very hard to take their word for it nowadays. We've seen too many numbers from Intel of late, that don't really reconcile with the actual product. The TDP number is by far the most misleading number because the actual product requires between 2 to 3.5x that amount of power to reach that kind of performance that Intel is tooting.

I guess I'm still in the minority that thinks Intel's TDP is pretty straightforward - 125 W PL1, that means a CPU limited to 125 W power consumption except for short bursts of Tau. The turbo values are "up to", which means you can achieve them within PL2, or with unlocked power limits. Nobody ever said your CPU will run at 5.3 GHz all the time. The only reason people are crying is because 4-core chips (up to 7th gen) could maintain max turbo speeds within PL1, while modern 6 and 8-core chips can't - but officially, turbo has always been an "up to" value with no guarantee.

AMD's TDP formula on the other hand...

 

[The Quim Reaper]

cannot wait to see how an old 10900K CPU will trash this i9 garbage of 8 Cores + 8 Useless Cores... Specially in games or render apps.

The 8 big cores have Hyper Threading...so there will be 16 high performance threads plus an additional 8 low power threads, giving 24 threads in total.

A 10900K will not be faster.

 

[TheLostSwede]

E core and P core are official terms. This seems legit.

E for Economy? P for Premium Economy?

The 8 big cores have Hyper Threading...so there will be 16 high performance threads plus an additional 8 low power threads, giving 24 threads in total.

A 10900K will not be faster.

Wouldn't that depend on the application? Not all things will be able to take advantage of the Atom cores.

 

[Prima.Vera]

I wonder if this is the upper limit, or just some average number than Intel pulled out. To be honest, it is very hard to take their word for it nowadays. We've seen too many numbers from Intel of late, that don't really reconcile with the actual product. The TDP number is by far the most misleading number because the actual product requires between 2 to 3.5x that amount of power to reach that kind of performance that Intel is tooting.


I doubt that will happen. Even with the current Rocket Lake, the IPC gain is noticeable over Comet Lake. And if you take Tiger Lake H vs the older Comet Lake H, the gains is even more pronounced in games and in rendering software. While the Alder Lake is capped at 8 performance Golden Cove cores, the efficient cores from what I heard is almost as fast as Skylake (not sure if this is true until we hear more about it closer to launch). So if you have 8 efficient cores helping to make up for the 2 missing high performance cores in the 10900K, I think it is highly possible that Alder Lake will be quite a bit faster. The only hurdle for Intel is AMD's Ryzen 9 series.

Actually the "old" i9 10900K is faster in games than the i9 11900K due to extra 2 Cores. Reviews are all over the internets...

 

[AusWolf]

E for Economy? P for Premium Economy?

E for Ehh... and P for Pff...

 

[persondb]

I think this has a way bigger potential in the market below the i9 and Ryzen 9, since well, if you compare an i5/7 to an Ryzen 5/7 then you would get the same number of big cores plus some additional little cores.

Depending on how Intel does their SKU for i3s, it could be really good as say 4+4 i3. It would probably be way better for marketing for office PCs too as the comparison would be lower core count AMD equivalents. For the higher end though, odds are that AMD will continue to be better.

That is, assuming that Intel doesn't fuck with perfomance and doesn't do terrible pricing.

 

[bug]

I guess I'm still in the minority that thinks Intel's TDP is pretty straightforward - 125 W PL1, that means a CPU limited to 125 W power consumption except for short bursts of Tau. The turbo values are "up to", which means you can achieve them within PL2, or with unlocked power limits. Nobody ever said your CPU will run at 5.3 GHz all the time. The only reason people are crying is because 4-core chips (up to 7th gen) could maintain max turbo speeds within PL1, while modern 6 and 8-core chips can't - but officially, turbo has always been an "up to" value with no guarantee.

AMD's TDP formula on the other hand...

The "problem", iirc, is that tau is not set in stone and some mobo makers just set it to forever, which makes the TDP limited only by the cooling solution (since PL2 is not sustainable under normal circumstances).
I don't hold this against intel, but who am I to stand in the way of a dose of good ol' hatin'?

 

[AusWolf]

The "problem", iirc, is that tau is not set in stone and some mobo makers just set it to forever, which makes the TDP limited only by the cooling solution (since PL2 is not sustainable under normal circumstances).
I don't hold this against intel, but who am I to stand in the way of a dose of good ol' hatin'?

To be honest, I'm not quite sure about that. Hardware Unboxed did a pretty extensive review of a whole bunch of B560 and Z590 motherboards. In it, they listed my Asus TUF B560M-Plus Wifi in the "comes with unlocked Tau out of the box" category, which is only true if you enable the Asus Optimiser in BIOS, which you have to do manually. If you don't touch the BIOS, you'll have Intel default PL and Tau values, so HU's categorisation is pretty much flawed.

In my opinion, the problem is reviewers 1. not being specific enough about how they test CPUs, or what settings they change to achieve the given numbers, and 2. testing every CPU with a 360 mm AIO or a custom loop, which most people don't have. They focus too much on peak performance, and forget about the fact that every use case is different. TPU is pretty much the only somewhat reliable source in this regard.

 

[The Quim Reaper]

Wouldn't that depend on the application? Not all things will be able to take advantage of the Atom cores.

I've No idea, until we see them in action.

It could be a similar scenario to when Hyper Threading was first introduced, the software had to catch up with the hardware.

It looks like AMD are going to do something similar as well in the future, so 'big' and 'small' cores seems to be the future.

 

[BorisDG]

It looks like AMD are going to do something similar as well in the future, so 'big' and 'small' cores seems to be the future.

Future for portable/laptop devices, but I don't see why such thing can be "future" for desktop based systems. And they are doing it, because Intel is doing it. I mean it's cool way to "test the waters" for mobile market, but still.

 

[bug]

To be honest, I'm not quite sure about that. Hardware Unboxed did a pretty extensive review of a whole bunch of B560 and Z590 motherboards. In it, they listed my Asus TUF B560M-Plus Wifi in the "comes with unlocked Tau out of the box" category, which is only true if you enable the Asus Optimiser in BIOS, which you have to do manually. If you don't touch the BIOS, you'll have Intel default PL and Tau values, so HU's categorisation is pretty much flawed.

In my opinion, the problem is reviewers 1. not being specific enough about how they test CPUs, or what settings they change to achieve the given numbers, and 2. testing every CPU with a 360 mm AIO or a custom loop, which most people don't have. They focus too much on peak performance, and forget about the fact that every use case is different. TPU is pretty much the only somewhat reliable source in this regard.

Like I said, hate brings in page clicks. Who am I to stand in the way of that?

 

[AusWolf]

Like I said, hate brings in page clicks. Who am I to stand in the way of that?

And the reason behind hate is misinformation 99% of the time.

 

[The Quim Reaper]

but I don't see why such thing can be "future" for desktop based systems. And they are doing it, because Intel is doing it. I mean it's cool way to "test the waters" for mobile market, but still.

Even desktops have their limits as to how big a silicon package you want to put inside a PC.

Silicion Die shrinkage is going at a slower pace than the ability/need to put more cores on them, so new ways have to be found to keep increasing processing power whilst keeping thermal & power requirements at a sane level.

..if 8 stripped down 'small cores' only use, as an example, 20-30% of the power of 8 'big' cores but have the equivalent processing power of 3 or 4 'big cores' then its the only way to do it.

 

[TheLostSwede]

I've No idea, until we see them in action.

It could be a similar scenario to when Hyper Threading was first introduced, the software had to catch up with the hardware.

It looks like AMD are going to do something similar as well in the future, so 'big' and 'small' cores seems to be the future.

I thought it had already been confirmed that unless you're on Windows 11, the Atom cores brings zero performance to the system, as Windows 10 doesn't seem to use them for anything.
Obviously other operating systems aren't supported as yet so...

HyperThreading was such a bust at launch. I remember when we tested that at PCW and it offered zero performance improvement. Intel tried so hard to find something where it would show an improvement, but failed.

Yes, seen those rumours too. It makes more sense in mobile devices imho, as I'm not really sure what the benefit would be in a desktop system except some power saving, which isn't that relevant.