Although Intel’s Gen 12, known as Alder Lake, will also be released this year, its Gen 11 is a change of purpose as far as Intel is concerned. With the aim of expanding the IPC and with it the performance after generations using the same kernel have changed kernels. With this they can stand up to the AMD Zen 3 and surpass the AMD architecture under certain conditions.
Core i9-11900K has better IPC than Zen 3 according to Geekbench
The Core i9-11900K it has been tested three times under Geekbench. All of them with the same motherboard configuration, Gigabyte Z490 AORUS Master, and with DDR4-3600 MHz as RAM.
The best result in terms of the single thread test has given the score of 1905 points, while the one that checks the performance for several threads has given the result of 10994 points. The conclusion drawn from these results is that the performance of the Core i9-11900K is 13% faster than the Ryzen 5950X, which is why the Intel CPU is faster in single-thread than the Ryzen 9 5950X.
This could be due to the abandonment of a core based on a core derived from Skylake for the eleventh generation of desktop Intel Core. Let’s not forget that the Rocket Lake-S architecture uses a new core, which brings a 19% increase in the IPC compared to the previous generation.
Are you cheating Intel?
One of the peculiarities that the Rocket Lake-S architecture has is the fact that a single core can be placed at 5.3 GHz speed exclusively. That is, Intel uses single-core boost, in which only one of the cores is throttled to a higher clock speed than the usual boost.
In the case at hand, the Intel Core i9-11900K has been accelerated to 5.3 GHz speed, so in the benchmark the CPU is running 500 MHz higher than its Boost speed for all cores, which is 4.8 GHz
The AMD Zen 3 win in multithreading to the Intel Core i9-11900K
The counterpart is in terms of performance with several threads of execution running in parallel, the reason for this is that Intel has decided to limit the maximum number of cores of its Intel Core Gen 11 based on Rocket Lake-S to only 8 cores, while CPUs and AMD have configurations of up to 16 cores, which allows it to reach 14000 points as seen in the graph.
How does this translate to performance? It’s hard to tell and it will depend on how the software code is. So there will be programs that will take advantage of more of the greater number of cores and threads of the AMD Zen and others of having a somewhat higher IPC in the case of Intel.
It is a shame that the Intel Gen 11 have such high prices, because compared to the Ryzen 7 5800X with the same number of cores it is more expensive, but the Intel product usually brings a number of implicit advantages in terms of support than its users know how to appreciate.
What is the IPC of a CPU really?
Due to the use of the term IPC in marketing, a misrepresentation of the term has ended up. If we speak in terms of processor architecture, the IPC is the number of instructions that a processor executes simultaneously, which should not be confused with the CPI which is the number of cycles per instruction.
In its classic definition, the IPC does not refer to the number of instructions that a processor has resolved at the same time, but in the definition that has been given from marketing, the IPC is the number of instructions that go out in parallel. And how do they measure it? They take a program, for example a benchmark, of which they have the instructions counted and they execute it counting the time. Since frequency, clock cycles, is the inverse of time, so they can know an average of the number of instructions per cycle.
But this measure is not entirely correct, since not all programs accelerate in the same way on all CPUs and there are other factors such as communication latency that must be taken into account.