Since the launch of the first IBM PC, the Model 5150, the Intel CPUs in our PCs have continually evolved generation after generation. Until we reached the current Rocket Lake-S we have seen how the processors based on the ISA x86 have evolved, we explain what that evolution has been generation after generation.
Intel processors have been accompanying us in our PCs since the first model that went on the market and the company that Gordon Moore founded in his day continues to create new CPUs, increasingly faster and more efficient. This is why we decided to take a quick trip to the entire history of Intel CPUs.
Once upon a time there was a PC with an Intel 8088/8086
The first PC was sold by IBM with the 8088 processor, which was a shortened version of Intel’s 8086 that only differed by having only an 8-bit bus. The processor was not originally designed for the PC, since when Intel created it, the PC as such did not exist but as an evolution of the Intel 8080 without backward compatibility with it, but from which it was easy to port the programs designed for the 8080.
It is a 16-bit CPU with the particularity of having a redirection of up to 20 bits, which allowed having up to 1 MB of memory in total, something that in 1981 was outrageous. Although the 8086 is the best known processor in the architecture, the first PC had an 8088 and it was not until more advanced versions like the XT PC that the 8086 began to be used in full with its 16-bit data bus.
The Intel 80286, the biggest jump in CPU of the x86
At the beginning of the 80s the processor par excellence in terms of power was the Motorola 68K, which ended up in several workstations as well as the Apple Macintosh, Commodore Amiga, Atari ST and a multitude of other systems. Given the enormous success of the IBM PC and the large amount of software from Intel, they devised a new 16-bit processor that was released by the IBM PC AT in 1984, although Intel finished the design in 1982.
The 80286 was the largest jump in IPC that has been made in the history of Intel CPUs, achieving a performance of more than double per clock cycle compared to the 8086 and 8088. Their buses were improved and they stopped being multiplexed, fetching instructions from memory was sped up, multiple buffers were created within the processor, and the capabilities of jump instructions were improved.
The first 32-bit x86 CPU, the Intel 80386
One problem that the PC had is that it could not run complex operating systems, the reason was that it had no modes that allowed the operating system to operate in isolation. These were added in the 80386 and thanks to this processor we were able to have much more complex operating systems than MS-DOS, making things like Windows 3 onwards and Linux possible on a PC.
It was also the first 32-bit x86 processor and did so by extending the length of the different registers, while maintaining backwards compatibility. It was also the first CPU with a segmented pipeline, which was in three stages. Without a doubt one of the most important processors in the history of computing.
The 80386 was not licensed to AMD, who had to reverse engineer and release a rough clone of the 80386, the am386 that ran at higher clock speeds and consumed less by using a more advanced manufacturing node than the original 80386.
Integrating the FPU into the CPU with the 80486
The floating point unit, formerly known as the mathematical coprocessor or 80 × 87, used to be sold as a component part, but starting with the 80486 at Intel they decided to integrate it into the CPU. But the changes based on the 80386 were not limited there, but also almost doubled the CPI, with the jump to the 80486 being the second largest jump in the history of Intel processors.
Other improvements included in this processor was the inclusion for the first time of the first level cache, divided between data and instructions since then, as well as improvements in the interface with memory and in pipelining, which increased from 3 to 5 cycles .
As with the am386, AMD went back to reverse engineering, again outpacing Intel’s design. Of course, again with the advantage of a newer and more efficient manufacturing node.
Intel Pentium, the first superscalar CPU for PC
The Intel Pentium was the first superscalar processor under ISA x86, which brought with it the ability to execute 2 instructions in parallel and at the same time. Thanks to this, the CPI from 80486 improved by 40% in the process.
Years later, Intel launched the Pentium MMX, which made a change in the FPU, since it implemented the SIMD over registration in it to accelerate the incipient multimedia programs of the time. But the life of the Pentium MMX was short.
Intel’s P6 architecture
The Pentium Pro was released in 1995 and brought with it out-of-order execution and speculative execution. Apart from increasing the number of pipeline stages from 5 to 14, which allowed Intel to launch several CPUs with this architecture.
- The Pentium Pro was the first CPU to have the L2 cache built into the processor, until then it used to be included on the board near the processor.
- The Pentium II was based on the Pentium Pro, but moved the L2 cache out, although I left it in the same package.
- The Pentium III on the other hand included the SSE instructions and eventually I ended up integrating the L2 cache back into the processor.
Since Intel’s P6 generation, all PC processors work by using out-of-order execution.
Pentium 4, the end of a road for Intel CPUs
For the Pentium 4, Intel created a new architecture called Netburst, which followed the trend of the day of adding a large number of stages to achieve a high clock speed. It was with this processor that Intel hit the ceiling of clock speeds and it was discovered that the race based on that metric had no future, due to the high consumption of the processors and the temperature they had.
It was due to the experience of the Pentium 4 that the power per watt metric became important and they began to design
Intel Pentium M, a CPU designed for laptops
The Pentium 4 were so hot and had so much consumption that they were unviable in a laptop, hence the creation of the first Intel Core, which were based on the philosophy of the P6, but with a shorter pipeline. They added improvements that would later be adopted by other processors such as the fusion of micro-ops, a stack dedicated to handling registers and new SSE instructions.
Intel assembled the Centrino platform with them and thanks to their specifications, laptops began to stop being niche computers with bad specifications to be able to start competing with desktop PCs from you to you.
Intel Core 2, the rebirth of Intel CPUs
The Core 2s are one of the milestones in Intel’s history and a paradigm shift, since they were the first multi-core CPUs from Intel where they also adopted for the first time the 64-bit extension of the ISA that AMD had previously proposed.
They managed to almost double the IPC compared to the Pentium 4, improved out-of-order execution, made the CPU capable of handling more instructions in parallel and added the Smart Cache for the first time on Intel processors.
The saga of Intel Core i3, i5 and i7 CPUs
Although the Intel Core 2 could be considered the first Intel Core for desktop, Intel began to use the Intel Core i3, i5 and i7 brands from Nehalem, which was an improvement on the design of the Core 2. Since in Nehalem Intel adopted AMD’s idea of integrating the Northbridge into the processor and with it the memory controller. In addition, they improved the cache, the jump prediction unit and brought back Hyperthreading after years of being absent.
From this point on, the changes in Intel CPUs have been quite modest compared to what happened before, little by little there are less things to improve.
- Sandy / Ivy Bridge: Intel again improved the jump prediction unit, apart from this improved elements such as the micro-op cache, the integer and floating point units and the performance of certain instructions for fetching data from memory.
- Haswell / Broadwell: Intel again expanded the number of instructions that the CPU can execute per cycle, in addition to increasing the bandwidth of the internal processor caches and improving the memory controller.
- SkyLake generation: Intel improved the number of instructions that the CPU can decode, but did not increase the number of instructions it can execute in parallel. The changes from previous generations are very small.
- Current Generation: It is the current Rocket Lake-S and Tiger Lake, After years with slight improvements in the IPC Intel has decided to follow the path of AMD not to be left behind.
This year we will see Alder Lake, the twelfth generation of the Intel Core i3, i5 and i7. Will it be a minor change like what we’ve had so far or will Intel rebrand itself a Core 2?