Intel’s first 12th-gen processors based on Alder Lake architecture are available starting today and, as expected, the biggest difference between the new chips and most of Intel’s previous-gen processors is the ability to combine two different types of CPU cores to provide a mix of high-performance and energy-efficient cores.
For example, the new top-of-the-line Intel Core i9-12900K processor is a 16-core, 24-thread chip that has 8 Performance (P) cores with support for hyperthreading and 8 Efficient (E) cores that handle basic duties when you don’t need the extra processing power. The efficient cores don’t support hyperthreading, which is why you get the odd thread-counts.
Thanks to heterogenous computing, a feature we’ve seen used in ARM-based smartphone chips for years, Intel’s new chips can use some or all cores at once.
Intel says it plans to launch 60 different processors as part of the 12th-gen Intel Core lineup, with chips ranging from mobile processors for thin and light laptops to higher-power chips for desktop computers… and that’s where the company is starting this time around.
According to Intel, there are 28 different chips currently shipping to OEM processors, but today only the first 6 Alder Lake desktop processors are available.
All of these chips are unlocked (overclockable) processors with a base TDP of 125 watts, although they can also consume more power when running at “turbo” speeds. The new chips also all support up to 128GB of DDR4-3200 or DDR5-4800 dual-channel memory, and 20 PCIe Lanes (up to 16 PCIe 5.0 and 4 PCIe 4.0).
But K-series chips feature Intel UHD 770 integrated graphics, while the KF-series models are designed for systems with discrete graphics (or, I suppose, applications where a GPU may not be needed), and don’t have any integrated graphics at all.
Here’s a run-down of Intel’s Core i5, Core i7, and Core i9 Alder Lake desktop chips:
|Chip||Cores (P + E)||Threads||L3 Cache||L2 Cache||Turbo Max Freq||P-core base / turbo freq||E-core base / turbo freq||GPU||Max Turbo Power|
|i9-12900K||16 (8P + 8E)||24||30MB||14MB||Up to 5.2 GHz||3.2 GHz / 5.1 GHz||2.4 GHz / 3.9 GHz||UHD 770||241 W|
|i9-12900KF||16 (8P + 8E)||24||30MB||14MB||Up to 5.2 GHz||3.2 GHz / 5.1 GHz||2.4 GHz / 3.9 GHz||N/A||241 W|
|i7-12700K||12 (8P + 4E)||20||25MB||12MB||Up to 5 GHz||3.6 GHz / 4.9 GHz||2.7 GHz / 3.8 GHz||UHD 770||190 W|
|i7-12700KF||12 (8P + 4E)||20||25MB||12MB||Up to 5 GHz||3.6 GHz / 4.9 GHz||2.7 GHz / 3.8 GHz||N/A||190 W|
|i5-12600K||10 (6P + 4E)||16||20MB||9.5MB||N/A||3.7 GHz / 4.9 GHz||2.8 GHz / 3.6 GHz||UHD 770||150 W|
|i5-12600KF||10 (6P + 4E)||16||20MB||9.5MB||N/A||3.7 GHz / 4.9 GHz||2.8 GHz / 3.6 GHz||N/A||150 W|
While the move to a new heterogenous architecture that pairs Performance and Efficient CPU cores on the same chip is one of the most noteworthy changes in the new processors, Intel is also promising significant performance gains over 11th-gen chips.
For example, the company says the new Core i9-12900K is about 19% faster than a Core i9-11900K when it comes to single-core performance, offers an 84% improvement in frame rates while gaming and streaming simultaneously, and brings a 50% improvement in Blender graphics rendering, while consuming less power.
Folks in the market for a new desktop should be able to pick up one with a 12th-gen Intel Core processor soon, but if you’re looking to build your own PC or upgrade an older one, note that the chips listed above are designed for a new LGA1700 socket, which means you’ll need to buy a new motherboard to go with the new chips.
This is the same philosophy/architecture as ARM’s big.LITTLE computing.
Intel’s Core-i cpu = large, clocker, fast, designed with hyperthreading
Intel’s Atom cores = small, efficient, slow, no hyperthreading
ARMs small cores = in-order, efficient, slow, for background tasks
ARMs medium cores = out-of-order, efficient, decent performance
ARMs large cores = large, clocks high, fast, inefficient
The way I kind of look at it:
ARM’s Cortex-A53 = (roughly) Intel Atom (Z8350)
ARM’s Cortex-A73 = (roughly) Intel Core-M (5Y10)
ARM’s Cortex-A78 = (roughly) Intel Core-i7 (8557u)
Oh, with that said: I don’t have any faith in Intel.
Their efficient cores may not be that efficient, and their performance cores may not be that fast. And their implementation may be wacky. And Windows may have issues with such a new architecture, that it may never fix the issue, and break proper backwards-compatibility with so many legacy Programs.
I’ll wait to see this in a 14inch laptop, and compare it to the new MacBook Pro with the M1P and M1X chipsets. And also see direct comparisons to some previous-gen Intel and AMD Windows 10 Pro laptops. Either with an i7-10500H and r7-5800H processors (30W range). Otherwise running an i7-10810U and r7-5800U chipsets (15W range).
Not sure what “Total L3 Cache” means… it’s less that the “L3 Cache” number. Is this the most L3 Cache that a single thread can use, is it on-die L3 Cache size?
On Zen 3, any single thread can use the entire L3, and the entire L3 is on-die.
Sorry, I goofed when translating Intel’s table into something that would fit on this page. They have “Intel Smart Cache L3” and “Total L2 Cache.” I accidentally wrote Total L3 instead of L2, but I’ve update the table so it should be accurate now.
Sorry, Zen 3 with 8 cores or less are on a single CCD. 12 and 16 core versions are on two CCDs, where each CCD has its own L3 cache. My 5600x is 6 cores with 32MB L3 on a single CCD.
Comments are closed.