The arrival on the market of the eleventh generation of Intel Core processors will also unequivocally mark the company's future roadmap , representing a sort of watershed between today and tomorrow, between the now traditional 14 nanometer node and the future ones. lithographs.
So let's find out what lies under the hood of these
intriguing CPUs which, while on the one hand represent the end of an era, are
probably also its best technical expression .
The Return To 8 Cores
The Cypress Cove architecture , adopted by Intel in this new
generation, allows Rocket Lake processors to obtain an increase of up to 19% in
IPC compared to Comet Lake CPUs , according to the American company. The number
of instructions per cycle is a fundamental parameter to be able to understand
the qualitative leap made this year. It is, in a synthetic way, the processing
capacity of the processor in a single clock cycle and, trivially, the higher it
is, the higher the performance. It is no coincidence that the intergenerational
leap made by AMD between Zen 2 and Zen 3 has allowed the latter to be able, for
the first time in years , to excel even in Single Thread activities against
Intel.
The objective of the Cypress Cove architecture is therefore
to fill this gap , through an interesting recipe that mainly uses two factors:
the 14-nanometer transistors and the Sunny Cove mobile architecture present in
the 10th generation Ice Lake processors. This is also the reason why it was not
possible to physically go beyond the 8 physical cores, as the architectural
scheme of Sunny Cove was designed for a 10 nanometer lithography.. Once
transposed to a 14 platform, a purely dimensional problem arose, also thanks to
a totally renewed and improved architecture of the UHD Graphics 750 graphics
accelerator, 50% more performing than the last generation but with more
generous dimensions.
It will therefore be stimulating to observe the behavior of
Rocket Lake processors on the test bench, to discover their real advantages and
any limitations, both in terms of performance and thermal management.
More Overclocking Possibilities
News are also present on the front of the RAM memory
controllers, which has allowed to further increase the range of frequencies
supported, now up to DDR4-3200 banks. In this regard, this year Intel has
focused a lot on memories and their management, expanding the overclocking
possibilities even on motherboards with mid-range chipsets, including the H570
and B560 , thanks to controllers with Gear 1 support. , Gear 2 and wider
timings , but the real pearl lies in the XMP suite, through which it will be
possible to modify the memory profiles and therefore overclock the memory in
real time, without accessing the BIOS.
Energy Management
In recent weeks, there have remained many rumors about the thermal
and energy management of the new generation processors, so it is necessary to
fully understand what the parameters TDP, PL1, PL2 and Tau mean .
The Thermal Design Power , in its most narrow as possible,
it is a parameter that indicates Watt in the dissipation capacity recommended
for optimal use at specific frequencies. This means that for a processor with a
125W TDP you will need a heatsink with that dissipation capacity.
However, this term also expresses power consumption under a
maximum usage load and is above all the definition of TDP that Intel provides
to consumers.but, as explained by the same company, the real consumption is
dictated by the usage scenarios, so this value is purely indicative and the
real CPU consumption can be higher in some cases, but very often lower.
So why do processors often exceed this threshold if it is a value declared
by the company?
This question also has a logical explanation and is to be
found in the Power Level 1 , Power Level 2 and Tau parameters , which mark its
duration. PL1 indicates, by definition, an energy consumption that can be
superimposed on the TDP, therefore the energy necessary to ensure the operation
of the processor under maximum load. The term PL2 instead means a peak
consumption higher than the TDP , which allows the processor to reach higher
frequencies in a more efficient and stable way for a given period of time,
which is quantified by the Tau parameter, i.e. the so-called Turbo Time
Parameter .
Also from this point of view Intel seems to have achieved a
small miracle, managing to maintain values very similar to those of the last
generation . Specifically, the new 8-core i9-11900K has a TDP set at 125W, an
identical PL1 and a PL2 of up to 250W, which is also the only difference to the
10900K with its 177W. In both cases, the Tau time limit is set at 56 seconds.
The existence of these parameters is therefore necessary to
understand Intel Turbo Boost Technology 2.0 , which allows the 11900K to reach
5.1 GHz, Turbo Boost Max Technology 3.0 , which raises this limit to 5.2 GHz
and Intel Thermal Velocity Boost , introduced on the very high end alreadyin
the last generation and which allows to reach a maximum clock of 5.3 GHz in
Single Core and 4.8 GHz in All-Core operations . These technologies translate
into thresholds, which can only be reached by combining the energy
possibilities of the system and its dissipation capacity .
A True All-Rounder
If on the performance front the differences and improvements
compared to the last seem tangible, the same can be said of the features
integrated within the die for multimedia and professional purposes.
We are talking about the support for HDMI 2.0 , HBR3 , as
well as the improvement of encoding technologies , thanks to the support of
technologies such as 12-bit HEVC and 10-bit AV1, in addition to full support
for E2E compression and Resizable BAR technology, which allows the CPU to
access the entire GPU framebuffer .
The icing on the cake, however, concerns the acceleration of
professional workloads mediated by AI, thanks to the implementation of
technologies hitherto reserved for CPUs dedicated to workstation and server
solutions . With Intel Deep Learning Boost and VNNI , in the professional
suites that take advantage of AI acceleration, the new CPUs will benefit from a
performance increase over competitors of up to 38%, depending on the type of
workload .