How mobile processors made the notebook work
In the early 1980s, the number of computers was still so small that it was initially not worth combining functions that could be implemented using transistor-to-transistor (TTL) logic into more heavily integrated chips. . And that’s why early laptops, as well as IBM’s first desktop PC, were crammed with generic TTL circuitry. It not only took up a lot of space, but also a lot of energy.
Personal computers of this era, including those of Sinclair and Commodore, however, showed that the design of specific (custom) components (Application Specific Integrated Circuit, ASIC) can be of interest. Additionally, Intel introduced CMOS construction for its processors in 1983. This “complementary metal oxide semiconductor” resulted in a higher level of integration at the manufacturing level because the two channels of a field effect transistor can be combined into a single layer of the component. .
For laptops, however, it was much more important that the power loss and switching voltage be reduced by at least an order of magnitude. Since then, the power consumption has also evolved with the clock frequency. So a device could be made more economical or faster by changing the clock frequency of the processor, which is ideal for laptops.
The predecessor of the PC processor, the 8085, already had a CMOS version called 80C85; that of the 8086 was called the 80C86. As Intel was still licensing liberally at the time, a number of other companies copied the economy chip – including NEC: in 1988, the Japanese developed the 2kg UltraLite laptop based on their V30 chip, which was compatible 8086.
The first real energy saver
Until the generation of the 80386, scaling power by reducing the clock frequency was enough to save power, although there were special CMOS designs from AMD, among others. The next major advancement for the mobile processor was Intel’s 80386SL in 1990, when System Management Mode (SMM) was introduced as an extension of the architecture. For the first time, almost the entire processor was able to go to sleep, all states were saved in a separate area (SMRAM) in main memory. The registers and other data of the processor were kept even without a clock applied, because the chip was built from static transistors (SRAM).
Unlike today’s procedures, the operating system didn’t notice anything when powering up the SMM – it was controlled exclusively by the firmware and the then external interrupt controller. It led to all kinds of issues like a laptop that woke up slowly and was left alone for a few minutes. Such startup problems were quickly resolved by software changes, and until the Pentium generations, SMM was the most important power-saving technology – at least for the processor itself: Intel and Microsoft had a standard for the rest of them with Advanced Power Management. (APM) from 1992 of a personal computer. It later became the Advanced Configuration and Power Interface (ACPI), which is still in use today.
The supply voltages were then separated with the 486SL. The CPU, bus, and RAM were 3.3, 5, and 3.3, or 5 volts, in that order. That alone, according to Intel at the time, should add an hour more with 3.3-volt RAM. Since then, different voltages for different chip areas are also common in the desktop world. With the early Pentiums, which brought a radically new x86 architecture via MicroOps and Pipeline, saving electricity was no longer so easy. Mobile chips without L2 cache have been on the market for a few years now and were therefore slower than desktop processors with the same clock frequency despite noisy fans. As a result, various versions of the 486SL were offered long after the early days of the Pentium, including by AMD, Cyrix, IBM, and others.
Pentiums are barely mobile
The relationship was not reversed until 1997: Power PCs became faster than the Pentium MMX derivative with the codename “Tillamook”, which was specially developed for laptops. After all: the Pentium could be cooled with little effort; depending on the cycle, it burned from 2 to 5 watts. The game was repeated with the Pentium II and III, but these processors were first throttled due to higher waste heat depending on the cooling system. They were also offered in different models, some of them almost bare to solder the silicon directly to the board. This “Tape Carrier Package” made the first particularly thin laptop possible – at least by the standards of the time.
The brash Pentium 4 was finally over, as it shouldn’t go any further in the direction of thicker, heavier laptops :. While the hunt for the 5 GHz bar in the desktop world – which the Pentium 4 never officially reached – first continued, in 2003 Intel pulled the Pentium M out of its hat. , especially for laptops. Despite the name, it had virtually nothing to do with older Pentium architectures. Of course, the success of the Pentium M has also contributed to the improvement of batteries, WLAN adapters and the spread of wireless networks: “Cable off! Was Intel’s battle cry for the Centrino campaign, but it also reflected the wishes of users.
The big turning point
The Pentium M was a design from the Israeli development department of Intel that had been pulled out of the drawer, previously designed under the code name “Timna” and initially pulped in 2000. The idea was revived by the processor core “Banias” – the basis of the first Pentium M. Important for saving energy: Almost all elements can be switched off specifically and dynamically via separate voltage islands, down to the individual parts of the L2 cache. In addition, for the first time there were variable clock frequencies for the core itself, which was marketed as SpeedStep.
Speaking of the core: the brand of the same name was introduced in 2006. To date, all core processors are reverting to the basics of the Pentium M and have also made desktops and servers more economical when doing nothing. For laptops, this meant that multi-core processors had also become possible: first two cores, then long four cores, and now there are up to eight cores – also for thin and light devices.
Even the prospect was so compelling that in 2006, Apple began converting its products from IBM’s PowerPC processors to Intel’s x86 processors. Similar to the Pentium 4, Power PCs have barely made any headway in terms of speed and efficiency, while Intel’s Pentium M roadmap has also promised (and kept) a lot for more powerful processors. It was not the least of the watts) depending on the intended use. have been sold. Fragmentation into such performance classes still exists today. In 2010, the previously separated chipset moved to the processor package, which made smaller motherboards possible. Today, the SoC (System-on-Chip) package is standard everywhere.
Cards are remixed
The fact that Apple has preferred its internal M1 chips with ARM architecture since the end of 2020 – always with a different instruction set – is linked, as with the move from IBM to Intel, to the negligence of the supplier over the years. before. The original version of the lake cores of today’s Intel chips hit the market in 2015, and lingering 10-nanometer manufacturing issues made matters worse.
AMD, on the other hand, hasn’t been at the forefront of laptops for a long time since the 486SL and only caught up with the Ryzen 3000 mobile in 2019. At this point, the change was already decided within Apple, although the rest of the laptop world was happy. Since the Ryzen 4000U, AMD has overtaken Intel’s mobile processors – and that’s still true for lightweight devices.
The picture could change again in the near future: Intel’s upcoming “Alder Lake” processor design combines thicker cores with less powerful cores to further increase efficiency, which returns as a separate design to the chip. of the Atom netbook – with much higher performance these days. Such combinations have been common with ARM processors for decades, but are only now entering the x86 mass market.
And given the ever-growing number of laptops, other home designs specializing in the Apple M1 could emerge. Google is rumored to be working on its own ARM processors for its Chromebooks.
C’t Retro 2021 offers reading material for long winter evenings. We’re paving the way for laptops, from heavy monsters to super slim all-rounders, and shedding light on the early days of the internet. Nostalgic fans will find out how to put old hardware back on its feet, and fans of old games will find out how classics can be carried over to today’s PCs. We have created Python programming instructions for the legendary Enigma. You will find the October 18 c’t retro issue in the Heise boutique and at the well-stocked newsstand.
Disclaimer: This article is generated from the feed and is not edited by our team.