Over the years Intel has prided itself on its ability to provide processors that fit into the ultraportable, professional market. We’re talking thin and light designs with obscene levels of performance and battery life for the form factor. It’s so important to Intel, that over the years they’ve produced several design and validation standards relating to how the best ultraportables should be developed, such as low power displays, the best connectivity standards, and approaching all-day battery life. It surprised me somewhat that Intel didn’t really discuss its next generation of processors for these devices at CES at the beginning of the year, focusing their keynote almost entirely on the 45 W prosumer and workhorse designs instead. To find out about the more mainstream and ultraportable silicon, we had to dig into the back end of our press deck to get details.

Core Counts Get Crazy

Intel typically addresses the ultraportable market with its U-series processors, which all end in a U in the product name. Depending on the exact configuration, these processors started with dual-core parts at 9 W, scaling to quad-core at 28 W. This time around, Intel is changing it up, partly due to the new Hybrid core design.

Starting with the low power processor at 9 W, Intel classifies this as a U-series processor featuring two performance cores, eight efficiency cores, and 96 execution units of graphics in the silicon. As with the ultra low power Tiger Lake processors, this uses an Intel design known as ‘BGA Type4’, or ‘UP4’, that combines both the CPU and the chipset on the same package. and is the smallest such processor package that they offer.

At 15 W, we’re in the more traditional U-series that we are familiar with. The package size increases to what Intel calls BGA Type3, or UP3, but you’ll notice that the CPU layout looks identical to the 9 W version. That’s because it is the same – up to two performance cores and eight efficiency cores, then 96 execution units on graphics. This is going to be important: for a market readily served today by the Core i9-1195G7 – an 11th Gen 15 W processor with four large cores – Intel is pivoting to fewer performance cores and more efficiency cores for their 12th Gen chips. It’s also worth noting that the chipset is also on die, but offers more functionality than the 9 W chips.

Then as we move to the higher powered parts, Intel is introducing a new series called P, designed to run at 28 W. Intel has technically used P before, but on the desktop, and then replaced it with F. In this case however, the P series stays in that UP3 form factor but gets the larger, more powerful silicon, with up to six performance cores, eight efficiency cores, and the same 96 EU graphics. The idea here is that a vendor can make a UP3 notebook and either equip it with a higher performance P-series chip or a lower power U-series at 15 W.

Perhaps this is why Intel kept this side of the Core 12th Gen away from the main presentation – we’re seeing a fundamental realignment of how the company is classifying its traditional 9/15/28 W segments. What used to be 4 core is now 2+8 core, and that expands into the processor lists as well, where we start with 5-core parts, and end at 14-core processors.

It’s also worth noting that the naming of these processors has changed from the previous generation – they are no longer getting G7, G4, or G1 suffixes to signify graphics configurations. Now Intel is only signifying power, using either a P or a U to respectively.

Intel 12th Gen Core-P at 28 W (UP3)
AnandTech Core
P+E
Base
E
Turbo
E
Base
P
Turbo
P
GFX
EUs
GPU
MHz
L3
MB
i7-1280P 6+8 1300 3600 1800 4800 96 1450 24
i7-1270P 4+8 1600 3500 2200 4800 96 1400 18
i7-1260P 4+8 1500 3400 2100 4700 96 1400 18
i5-1250P 4+8 1200 3300 1700 4400 80 1400 12
i5-1240P 4+8 1200 3300 1700 4400 80 1300 12
i3-1220P 2+8 1100 3300 1500 4400 64 1100 12

At the top of the P-series is the Core i7-1280P, and it’s the only P-series processor with everything enabled. It has six performance cores, eight efficiency cores, and offers a peak turbo frequency of 4.8 GHz. It has all the graphics cores running at a 1.45 GHz peak, and a max turbo power consumption of 64 W.

The other Core i7 parts, the i7-1270P and 1260P, lose two of the performance cores, and in exchange get extra base frequency on both performance and efficiency cores. The loss of two cores also reduces the L3 cache size from 24 MB to 18 MB, which is going to have an effect on some performance metrics, especially if these processors are paired with discrete graphics.  Looking at this table, Intel should have made that i7-1280P at the top a Core i9 to differentiate it.

In the mid-range, the Core i5 processors have the same 4+8 core configurations as most of the i7 parts, but are reduced slightly on peak turbo frequency and graphics, coming down from 96 execution units to 80 execution units. This is a 16% drop, but the frequencies are still equivalent. Intel still lists these processors as 28 W Base, 64 W turbo.

Bringing up the rear is the Core i3-1220P, which offers only two efficiency cores and eight performance cores. Thie integrated graphics are also further cut back, with another drop to 64 execution units, but still the same 28-64 W power window. This Core i3-P looks very much like what the top Core i7-U processors will be in that configuration, but the Core i7-U are much more efficient.

Intel 12th Gen Core-U at 15 W (UP3)
AnandTech Core
P+E
Base
E
Turbo
E
Base
P
Turbo
P
GFX
EUs
GPU
MHz
L3
MB
i7-1265U 2+8 1300 3600 1800 4800 96 1250 12
i7-1255U 2+8 1200 3500 1700 4700 96 1250 12
i5-1245U 2+8 1200 3300 1600 4400 80 1200 12
i5-1235U 2+8 900 3300 1300 4400 80 1200 12
i3-1215U 2+4 900 3300 1200 4400 64 1100 10
Pentium 8505 1+4 900 3300 1200 4400 48 1100 8
Celeron 7305 1+4 900 N/A 1100 N/A 48 1100 8

If the transition from P to U wasn’t a giveaway that we’re coming down the stack, Intel also makes this distinction by putting a 5 as the last digit in every processor. That’s because the 0 parts are the even lower 9 W hardware we’ll cover a bit further down.

The top line Core i7-U at 15 W comprises two processors, each with two performance cores and eight efficiency cores, with the Core i7-1265U having a turbo up to 4.8 GHz. Both parts have 96 execution units for graphics running at up to 1.25 GHz, which is slightly slower than the P series. But the power range is listed as 15 W on base and 55 W on turbo. Ultimately this isn’t a massive difference from the P-series in all honesty, as at the top end the U series is only 9 W less for the loss of four performance cores.

Mid-range Core i5-U at 15 W has the same i7-to-i5 reduction we saw in the P-series. We still get two performance cores and eight efficiency cores, but the integrated graphics come down from 96 to 80 execution units. The power range of 15-55W is still the same.

Now at the low end is where we get interesting. There is a single Core i3, running in a 2+4 configuration with 64 execution units, but it’s the Pentium and Celeron that are worth looking at. Both the Pentium 8505 and Celeron 7305 (notice there’s no U) run with only a single performance core, four efficiency cores, and 48 execution units on graphics. In our lists, Intel has the Pentium with turbo frequencies up to 4.4 GHz, while the Celeron continues to not get any kind of turbo boost ability. What we end up with is a 5 core configuration.

Intel 12th Gen Core-U at 9 W (UP4)
AnandTech Core
P+E
Base
E
Turbo
E
Base
P
Turbo
P
GFX
EUs
GPU
MHz
L3
MB
i7-1260U 2+8 800 3500 1100 4700 96 950 12
i7-1250U 2+8 800 3500 1100 4700 96 950 12
i5-1240U 2+8 800 3300 1100 4400 80 900 12
i5-1230U 2+8 700 3300 1000 4400 80 850 12
i3-1210U 2+4 700 3300 1000 4400 64 850 10
Pentium 8500 1+4 700 3300 1000 4400 48 800 8
Celeron 7300 1+4 700 N/A 1000 N/A 48 800 8

Funnily enough, when we move to the 9 W family of hardware, not a lot changes, each configuration still has the same core count as the 15 W version, although in each case the base frequencies are lower. These 9 W processors also only have a maximum turbo power consumption of 29 W. It is worth pointing out that even though the two i7s look identical, the top i7 supports vPro Enterprise, while the other only supports vPro Essentials. The same delineation exists for the Core i5 parts as well.

Between the three families, there are some other differences.

The 9 W processors support both LPDDR4-4267 and LPDDR5-5200, whereas the 15W and 28 W processors support those as well as DDR4-3200 and DDR5-4800. Usually support for DDR means double the peak memory capacity available, leading to 128 GB on DDR4/DDR5 and only 64 GB on LPDDR4/LPDDR5.

The 28 W processors also have Turbo Boost Max 3.0, which is Intel’s marketing name for favored core. It means that one of the cores on the P-series Core i7 is designated the ‘best’ core, and can turbo up higher than the others, typically +100 MHz or +200 MHz, depending on the configuration. This would push the Core i7-1280P from 4.8 GHz to 5.0 GHz, and I’d check on Intel’s official ark.intel.com database for the details, but despite the link being in the slide deck, the public database has not yet been updated.

Other details are TBD, given that Intel has not provided full breakdowns of the chipset and PCIe support for anything other than the H-series processors at this time. This is despite Intel’s own press release for CES states that U-Series and P-Series design wins will available in the first quarter, Evo-enabled designs will come in the first half of the year, ‘including foldables, 2-in-1s, detachables, and others’.

With AMD pushing its 15 W parts more into the 28 W space stating that it has made a breakthrough in its power gating design, it’s interesting to see that instead Intel has extended its 45 W processor down to 28 W. It’s going to be very confusing for customers to understand what’s in their product for sure, but not only that, it’s going to be interesting to see how OEMs deal with a CPU market that is becoming more fractured when it comes to performance, functionality, power, and design. Meanwhile, there are some shortages going on relating to display controllers and other similar things, causing schedules to slip or portfolios to adjust accordingly.

Comments Locked

91 Comments

View All Comments

  • ChrisGX - Thursday, January 20, 2022 - link

    Yes, and the parts you mention -- M1 Pro and M1 Max -- push well into i9 territory. The higher end versions of the M1 (or, no doubt, M2) aren't targeted at thin and light laptops. Those chips are to be compared to high end Intel parts not the chips mentioned in this article. The M1 and/or M2 represent the right point of comparison for the ADL chips covered in this article.
  • web2dot0 - Thursday, February 10, 2022 - link

    Totally disagree. Intel's CPU is targeting for thin and light because they don't expect those CPUs to be paired with dGPU.

    On the other hand, Apple's M1Pro/M1Max are paired with 16c/24c/32c GPU which are MUUUUCH faster than Intel's iGPU ... Apple take the GPU in the account when they configure their laptops.

    To compare Apples to Apples, M1Pro/M1Max CPU has comparable power draw (albeit even lower power consumption since they don't have "peak power draw" figures. Their CPU simply max out at 30W (40W CPU SoC total draw including RAM,MB,etc..).

    This is simply on par with Intel's U/P series "thin and light" CPU's power numbers to be honest.
  • kwohlt - Thursday, January 20, 2022 - link

    On ARM, Apple has 2 laptops, 1 of which comes in 3 sizes. And they have one mini computer that uses laptop parts. Of course they're going to have significantly less CPU SKUs than Intel, who makes a wide range of chips for everything thing from embedded devices, to thin and lights, to standard laptops, desktop replacements, desktops, workstations, servers, etc.

    Of course they're going to have significantly more SKUs not only for different types of devices, but for different price points within these types.

    If it made sense for Intel to offer only 2 chips to meet all client computing needs and budgets, they would. But it doesn't at all.
  • ChrisGX - Thursday, January 20, 2022 - link

    Well, the M1 is for thin and lights and no chip mentioned in this article is for embedded devices, desktop replacements, desktops, workstations or servers. Your comments on those points are irrelevant. The ADL chips covered in this article are for thin and light laptops. And, that huge profusion of chips is there for the sake of Intel's margins, nothing more. There is no benefit for users in any of it.

    How exactly to users lose out by choosing the M1 over all of these ADL parts? (Please note, I have conceded that in terms of genuine consumer needs a more energy efficient and lower power option would be useful. Perhaps, I'm wrong, maybe even two non-premium options might be needed.)

    Personally, I wouldn't equate Intel's profit margins with user needs. If that was valid Intel oughtn't to be experiencing any problems today but it is one of the few major semiconductor companies that isn't growing.
  • Calin - Friday, January 21, 2022 - link

    "workstations, servers"
    The servers and workstations have their own processors, not included there. Usually they have a bit more memory bandwidth for the High Performance Computer (one more fast memory channel) or several more channels of slower memory for servers.
  • Spunjji - Friday, January 21, 2022 - link

    Yup. Two chips and perhaps three bins of each chip for good yields and some price distinctions - fastest, not-as-fast, fast-enough.
  • Sahrin - Thursday, January 20, 2022 - link

    …AMD is able to run a Zen 3 core at 3.5GHz at less than 3W, so it’s really hard to imagine a situation where any number of P cores which are probably ~3W themselves combined with a large number of E-cores which are substantially but not magically more efficient is doing much better in power terms,
  • iphonebestgamephone - Friday, January 21, 2022 - link

    Which zen 3 core is this?
  • Da Kat - Sunday, January 23, 2022 - link

    >Bringing up the rear is the Core i3-1220P, which offers only two efficiency cores and eight performance cores.

    I BELIEVE that should be the other way around, 2 performance, 8 efficiency..
  • ptr2ptr - Wednesday, February 2, 2022 - link

    Im new here trying to understand. Could someone explain why Intel makes so many variants of its processors even within i7, or i5 or i3 ?

Log in

Don't have an account? Sign up now