The history of the Rocket Lake-S processors is tied hand in glove with that of Intel's 10nm manufacturing node. The new CPUs are in fact based on Cypress Cove architecture , which takes the place of the SkyLake used previously, a 14 nm adaptation of the Sunny Cove seen last year on notebooks, made precisely at 10 nm.
Intel's problems in producing large volumes of 10nm
processors have prompted the company to make a drastic but logical choice: to
keep the best manufacturing process for notebooks, which need more energy
efficiency, and to adapt the Sunny Cove architecture to suit 14 nm , thus
creating Cypress Cove processors, like this i9-11900K we're going to talk about
today.
Beyond the production process, which remains the same as the
previous desktop class processors, the news are many, which is why we have
collected them all in a special dedicated to the eleventh generation of Intel
CPUs . In this object, however, we will focus on the performance of the top of
the range i9-11900K, which reserve some surprises, especially when it comes to
single core performance.
From Sunny Cove To Cypress Cove
Adapting the Sunny Cove architecture to the 14nm
manufacturing process was not painless . The 10 nm in fact allow to condense a
greater amount of transistors inside the chip, being smaller in size. With the
transition to the 14 nm production process, however, sacrifices had to be made
in terms of calculation units, so much so that the top of the range i9-11900K
has "only" 8, with Hyper-Threading technology, which brings the
number of logical units at 16. The previous top of the range i9-10900K instead
had 10 Cores and 20 Threads: the consequence is that in highly parallelized
tasks the performance is obviously lower , we will see it even later when we
analyze the benchmark results.
This is why Intel has focused heavily on gaming with these
CPUs, and in general on all those tasks that do not require a very high number
of processor computing units. After all, there are no alternatives, in the
absence of an adequate yield on the 10 nm node, which is sufficient for both
desktop processors and for those intended for notebooks, it is the 14 nm node
to offer the greatest guarantees.
On this front, AMD has an undeniable advantage, being ready
for some time with the 7 nm production process , made by TSMC, but also the 7
nm node has a high price to pay, that is, the availability of processors is not
always capable. to satisfy requests .
Intel instead relies on its own factories for the
construction of processors, with a decent availability on almost the entire
range of 10th generation CPUs. In a market where scarcity and high prices are
now the norm, Intel has an advantage over AMD at the production level and even
with the new generation there should be no problems of availability, except for
a request that exceeds expectations.
Cypress Cove And The New I9-11900k
As we have anticipated, the novelties of the new
architecture are not few and you can read them in our dedicated special . In
summary, with Rocket Lake-S processors Intel claims to have brought an
improvement in IPC of up to 19%, along with a revamped Iris Xe graphics
section, with up to 50% higher performance. Positive news is the use of the
LGA1200 socket, the same as the last generation, which therefore remains
compatible and will benefit from the Resizable BAR, 3200 MHz RAM memories and
20 PCIe 4.0 lanes., obviously after updating the BIOS. With the new Z590
motherboards you get something more, namely support for USB 3.2, Thundebolt 4,
Wi-Fi 6E standards, as well as Intel DMI Gen 3.0 8x, capable of providing
double the bandwidth between CPU and chipset .
Our test was carried out with a ROG Maximus Hero XIII
motherboard, equipped with the Z590 chipset, along with 16 GB of RAM at 3200
MHz and an RTX 3080. On this basis we installed the Intel i9-11900K processor,
with 8 cores and 16 Threads , base clock of 3.5 GHz and capable of reaching 5.3
GHz peak on a single core.
The TDP is 125 W, a value that indicates the ideal
dissipating capacity for the operation of the processor, so we are not talking
about consumption. These vary according to the type of use that is made of the
CPU. To understand the consumption of the i9-11900K it is necessary to look at
the PL1, PL2 and Tau values . PL stands for Power Limit, or the power limit, and
it works through two stages, PL1 and PL2 precisely.
PL1 corresponds to the TDP, with the power limit set to 125W
for this processor, while in the PL2 state it is possible to reach 250 W.
Thanks to PL2, the highest operating frequencies are obtained, however
maintaining performance of this type for long periods of time will not it would
be possible, that's why the Tau (Turbo Time Parameter) intervenes, which
defines how long the CPU can run at maximum (in the case of the i9 this time is
56 seconds). Based on these parameters , Intel's optimization technologies work
, such as Turbo Boost Technology 2, which pushes the CPU up to 5.1 GHz, Turbo
Boost Max Technology 3.0, with which it rises to 5.2 GHz, and finally the
Thermal Velocity Boost, with which reach 5.3 GHz on a solitary core and 4.8 GHz
on all cores.
This is the theory, in practice it all depends on the type
of motherboard used, in fact each manufacturer can customize the energy
thresholds of the processor operating states, but above all from the dissipation
system : the more it is able to dissipate heat, the higher the performance. .
We used a NZXT Kraken X63, if you plan on getting a
processor like this, don't save on the heatsink.
In terms of consumption, a further distinction must then be
made based on use, because 250 W is not easily reached; in addition to the
power limits, the motherboard and the heatsink used, the applications also
influence this parameter. During the AIDA64 stress test, the i9-11900K
immediately reaches the peak of 250 W, but keeping it only for a few seconds,
then settling at around 230 W: our heatsink was not enough to maintain that
power level for longer and not as soon as it reached 100 degrees, thermal
throttling came into play, thus lowering performance and consumption.
However, this test is not representative of actual use,
where the processor is not stressed in the same way. During the game, for
example, temperatures and consumption are lower, after 5 benchmark cycles with
Metro Exodus we detected a temperature of 65 degrees peak (29 in idle) and a
consumption of about 75 W in 1080p, which rise to 88 W in 720p, a resolution
that takes the processor more by generating a greater number of frames. On the
thermal and energy front, the i9-11900K is certainly a strong processor, after
all, the fact that this architecture was designed for 10 nm speaks volumes, but
better not to confuse consumption and peak temperatures during a benchmark with
the real ones. .
Turning instead to performance, the synthetic benchmarks
show a clear step forward in single core performance compared to the i9-10900K
, so much so as to deliver the throne in the CPUZ test to the 11900K, with a
score of 704, the first time we observe a CPU exceed the threshold of seven
hundred points. Even with Cinebench R20 the gap compared to the previous model
is noticeable, but not enough to pass in front of AMD's proposals. Indigo
Benchmark, which analyzes performance in rendering, sees the i9-11900K
outperform the former top of the Intel range, despite a lower number of cores,
also placing itself above the Ryzen 7 5800X but behind the 5900X, which however
has a number of higher computing unit.
Entering the field of multi-threaded benchmarks, the real
rival is the Ryzen 5800X, the superior models in fact win easily, thanks to the
greater number of cores available. With CPUZ and Cinebench R20 the 5800X is
slightly faster, while with Indigo Benchmark it is Intel who prevails, a not
obvious result considering that AMD's solution is made at 7 nm, against Intel's
14 nm, which is squeezing every single drop of power from this productive node.
On the gaming front, use indicated by Intel as primary for
this model, the gap with the previous top of the range varies from game to game
and based on the resolution, in some cases it is palpable, in others the
i9-10900K remains even in front. In this context, the gap with the previous
generation is not so evident , but the fact remains that the comparison with
the Ryzen solutions sees the 11900K slightly below in Full HD, while going up
with the resolution the Intel proposal is often found in front, without however
clearly detach the competitors. Performance practically in line with AMD's
higher-end here, a not just achievement considering the less advanced
manufacturing process.