Intel Alder Lake

The company has made the biggest change to its products in the last decade.

The Alder Lake-S microarchitecture is one of the biggest changes Intel has made to its desktop processors.

The company is introducing a concept already widely used in platforms based on ARM, and now uses a hybrid architecture in its processors, mixing cores with different characteristics and trying to extract the strengths and minimize the weaknesses of each one.

Intel Alder Lake – Two microarchitectures, one processor

This leaves the cores identical and two different types come into play. Focusing on performance we have the P-Cores, based on Golden Cove, an evolution of the Cypress Cove 11th generation desktop processor, with an improvement of the 19% in CPI compared to its predecessor operating at the same clock rate.

This is the result of modifications to several structures, increasing decoding from 4-width to 6-width, the reorder buffer (ROB) increased from 352 to 512 entries, and execution increased from 10 to 12 ports.

All of these changes represent the biggest shift in Core architecture in decades, equivalent to what the company brought with the introduction of Skylake, on Intel Core 6000 models like the 6700K.

The 12th Gen Intel Core is the most profound modification to the line in decades.

But the big news is the presence of efficiency cores, the E-Cores, based on the microarchitecture Gracemont, an evolution of the Tremont present in products such as low voltage CPUs.

Despite the focus on low power and heat, Intel claims these cores can deliver 40% more performance than a Skylake core while consuming 40% less power and scale up to 80% more performance with 80% less power compared to dual cores and 4-thread Skylake vs. 4 cores and Gracemont 4-thread. According to Intel, the efficiency cores have comparable performance to the 10th generation Core cores, codenamed Comet Lake-S.

Intel Alder Lake brings performance cores and efficiency cores combined

To cope with this variation in cores and their different capabilities, Intel needed to develop a new solution to more efficiently distribute the different operations required by the system.

This is how Intel Thread Director comes into play, giving more information to the operating system about the performance it can get from each core. So, the system will prioritize the use of high-performance cores, then high-efficiency ones, and finally enable Hyperthreading in P-Cores and thus will bring more threads available for high performance parallelism scenarios.

Therefore, Alder Lake processors can use any core for an operation, and the system will intelligently choose which core is best to dedicate a job to.

But for this it was necessary to unite the two architectures and make them capable of handling the same functions, which makes the cores Gracemont receive an update to support AVX2, but in exchange there was a downgrade in Golden Cove, which dropped AVX-512 support.

The structures will even be there, after all, they use a structure similar to the server cores codenamed Sapphire Rapid, but they will be disabled, limiting the use of this type of operation only to the server and HPC market.

The Z690 platform

Another major change to the processors codenamed Alder Lake-S is the technology upgrade, putting Intel back in the forefront after AMD platforms Ryzen take advantage of the adoption of new technologies.

The 12th Gen Intel Core introduces new memory, DDR5 and LPDDR5, while maintaining support for DDR4/LPDDR4 simultaneously.

It is up to the motherboard to determine which memory will be compatible, as processors have controllers capable of handling both formats.

In the PCI Express slot we have the introduction of version 5.0 technology, doubling the bandwidth compared to PCIe 4.0.

Things are more relaxed here, after all we are talking about a technology that is backwards compatible with previous summers, and which involves less mandatory updates compared to DDR.

The products

Intel unveiled its new 12th Gen Intel Core products, codenamed Alder Lake, which hit the market on November 4.

The company will start sales with its high-end products, starting with the Intel Core i5, going through the Intel Core i7 and Intel Core i9 models, all with the K suffix, that is, with unlocked overclocking.

The big news in these products is the increase in core count, adding the two available color styles.

Core i5 models now come with a total of 10 cores and 16 threads, the Core i7 comes with 12 cores and 20 threads, and the top-of-the-line Core i9 comes with 16 cores and 24 threads.

The count may seem curious, since we do not have twice as many threads compared to the number of physical cores, and this asymmetry is the highlight of this generation: Alder Lakes introduces two types of cores, the high-performance ones, codenamed Gracemont, and the efficiency ones, Golden Cove.

Both Core i9 and Core i7 provide the maximum count of high-performance cores (P-Cores) with eight in total, while Core i5 provides six performance cores.

In the efficiency cores (E-Cores) only the Core i9 has eight cores, and both the Core i5 and Core i7 have 4 E-Cores.

This causes the thread count to change. Only the P-Cores, the performance cores codenamed Gracemont, they bring Hyperthread, a technology that makes it possible to use unused portions of the core and generate a new logical core, generating two threads per physical core.

That's why we have these curious values of 10 cores and 16 threads of the Core i5, with only the 6 P-Cores delivering the Intel Hyperthread and thus reaching 12 threads of the performance cores plus four of the efficiency cores.

The complete product table includes:

In direct competition, we have different scenarios.

The high-end Core i9 came in just over $ 549 for the Ryzen 9 5900X.

On the other hand, the Core i7 and Core i5 models are slightly cheaper than direct competitors like the Ryzen 7 5800X (US $ 449) and Ryzen 5 5600X (US $ 299), respectively.

The products are already available for pre-order, including at some Brazilian retailers, and will reach consumers starting November 4.

Overclocking, consumption and XMP

One notable change is that Intel is moving away from using TDP as a power consumption benchmark.

This makes sense as this is not the most suitable unit, after all, although the heat dissipation of the chip is an indication of consumption, it is not the consumption itself, and secondly because the TDP used the base frequency value as a reference, and the frequency increases.

The boost route caused consumption much higher than that indicated in the TDP.

Now comes the TDP and enters the Turbo and Base Power, which indicate the maximum consumption at the maximum turbo of the processor specification and the maximum consumption at the base clock, respectively.

With the introduction of DDR5, memory support also changed. In DDR4, the official support is 3200MHz at CL22, while DDR5… is complicated.

The company has different specifications for different configurations.

If you only have two slots on your motherboard, the support is 4800MHz.

But if you have four slots and occupy two, support drops to 4400MT/s, dropping to 4000MT/s if there are four single-rank modules and reaching 3600MT/s if there are four slots fully populated by four memories and these are dual-rank.

It's not good at all considering that in some slides the company claims that Alder Lake-S supports DDR5-4800, something that, frankly, there are few who will actually align all the conditions for this specification to be supported.

But at the end of the day, people will turn on XMP and operate outside of those specs, as has been the case for years.

Speaking of XMP, Alder Lake-S introduces the next generation of Extreme Memory Profile. XMP 3.0 aims to expand the flexibility of profiles, with several new features.

The first is that memories can now have up to 5 profiles, some customizable by the consumer.

Until XMP 2.0, the consumer was limited to two profiles, both defined by the manufacturer.

Thus, when opening the BIOS, the consumer is faced with an option with the highest frequency supported by the kit and an intermediate option, sometimes with adjustments to the timings, sometimes with some other approach from the memory manufacturer.

The consumer could make modifications to these values on supported platforms, but the system became unstable, failing to boot only to see everything crash and the machine powering on at the traditional 2133MHz of the DDR4 standard.

XMP 3.0 now expands the number of profiles a manufacturer can include in their product to three, and two more slots have been created with settings that the consumer can save, totaling 5 profiles. In addition to fine-tuning, the consumer can also define a name for each profile, making it easier to identify the settings they have created.

But another relevant news is that turbo boost has gained a memory equivalent: Dynamic Memory Boost. This technology makes it possible, on platforms that support memory overclocking, to switch between an efficiency mode, using the JEDEC baseline, and a more advanced XMP profile under high load, increasing memory bandwidth when the system needs more performance.

This is a feature that should have a bigger impact on laptops, as gaming models can deliver high performance while preserving battery life, simply by switching between these profiles.

Overclocking on the Z690 platform is unlocked when paired with K- and KF-end processors, and we have a lot of new things in this field.

The first is that Intel kept the solder (STIM) as the heat dissipation interface between the die and the outer metal frame (IHS), but changed the dimensions, with a thinner die, which makes heat dissipation easier, and also the STIM is now a smaller layer, another factor that will help the platform's ability to control heating.

To maintain a compatible overall height, the IHS has been increased.

Alder Lake-S is now available for pre-order and will be available to consumers starting November 4.

We hope to have our tests with these models soon, bringing our impressions with the new processors and also the platform as a whole.

Read also : How to build a gaming PC.

Fountain: Anandtech