« DJI Enterprise -- Drones are becoming serious business | Main | Thunderbolt 4 »
March 24, 2021
Intel generations
With the rapid succession of Intel Core processor generations it’s difficult to keep track of what’s available, what it can do for you, and what really matters when selecting the most suitable processor for your systems.
It seems like only yesterday that Intel introduced the high-end Core processor brand, and now it's a decade later and we’re already seeing the 11th generation. While initially it wasn’t very difficult to keep the generations apart and determine when and what to buy, it’s become quite confusing as of late.
Initially, generations happened at what Intel called the Tick-Tock pace, where Tick meant a new architecture or processing technology and Tock meant refinement, that’s no longer the case. And while initially each new generation brought very tangible technology and performance benefits, that’s no longer necessarily the case. Fundamental technological progress has slowed down as it’s becoming ever more difficult to miniaturize processing technology yet again. There are also overlaps of generations as well as additional complexity due to additions of Intel’s initial i3/i5/i7 “good/better/best” designations.
So let’s take a look at various areas, aspects and issues regarding the processor family that drives the vast majority of mobile Windows computers today.
Overall importance and position — Even more so than on the desktop where Intel has some competition from AMD, Intel totally dominates the market for high-end, high-performance mobile processors. For the purpose of this discussion, the term “mobile” isn’t strictly limited to non-stationary devices, but also to those where thermal considerations mandate use of processors that balance performance with low power consumption (and therefore low heat generation). Intel generally describes such processors as “ultra-low voltage” with thermal design power of 15 watts or less. This includes not just laptops, tablets and handhelds, but also a good number of panel computers and embedded systems.
So many different processors -- The PC revolution began with one single processor, the Intel 8088 than ran at 4.77MHz -- a thousand times slower than a good Corte processor today -- and cost about five bucks. Subsequent early Intel processors, such as the 80386 and 80486 could be had with different clock speeds, and that was that. Today, Intel offers several hundreds of CPUs, with often just the most minute differences. Some of that differentiation is justified, but far too often the proliferation seems entirely marketing-driven.
Consider pricing -- When the first IBM PC came out in 1981 it cost US$1,565, which would be about US$4,500 today. Its 8088 processor cost about $5, or about $15 in today's money. Today, a simple Intel Atom processor still only costs about $15, but a higher-end Intel Core processor can be up to $500. This means that the type of Intel processor used in a system greatly affects the system's price. You may or may not need the top-of-the-line processor and you may be able to save quite a bit by going with a slightly lower processor option.
It's not just the processor -- Intel's simplistic "good/better/best" approach to CPU marketing leads many customers to believe i3/i5/i7 is not only good/better/best but also fast/faster/fastest. There is something to that, of course, given the number of cores, clock speed and turbo modes, but a computer's performance does not only depend on the CPU. It also depends on the type, speed and quantity of RAM, the type and speed and capacity of mass storage, the type and speed of graphics, the type and configuration of the operating system, and a whole bunch more. If there is not enough RAM, the processor doesn't have enough room to do its processing in. If there's not enough storage, the system must constantly seek for room to store stuff. If putting stuff into and getting stuff from mass storage is slow, that becomes a bottle neck. If the OS is sluggish and inefficient, it slows things down. Even drivers and obsolete versions of software can slow things down.
The latter, especially, is frustrating, to the degree where one might get the impression that with computers it's almost a game where software becomes bigger and more bloated and that necessitates buying a more powerful processor. But then the software takes advantage of the new power and the system gets slow again and the cycle repeats. If you have ever "upgraded" an old PC's operating system and software multiple times, you'll have noticed how the computer got slower and slower. I recall an old Toshiba Portege laptop I once had. It was a quick and nimble machine when I got it. 10 years later, though equipped with more memory and a much faster disk, the laptop became so slow as to be completely useless -- a result of the continuing see-saw dynamic of new hardware and software.
The point, however, is that technology never stands still. What’s good today is obsolete tomorrow. Replacement will happen. Your goal must be to get equipment powerful enough to easily get today’s work done on today’s software and operating systems. No need to go for the absolute fastest because it, too, will soon be obsolete. On the other hand, don’t skimp and get gear that can barely keep up with today’s software.
The mighty Cores — For a long time CPUs had just one core. "Mobile" processors -- processors small and frugal enough to run on battery -- had two cores since the beginning of the Core processor line. It stayed that way until the first quad core processors showed up in the 8th generation ("Kaby Lake Refresh"). The 10th generation ("Cannon Lake/Ice Lake") boosted that to six cores.
Why does it matter? Are two cores twice as good/fast as one? And four twice as good and fast as two? Not necessarily. That's because it not only depends on the number of processing cores, but also on how software can use them. If software is written so that everything must be processed serially, one thing after another, then two cores aren't any faster than one single core. The the software is written so that processing can be broken down into parallel "threads," then two processors can be as much as twice as fast as a single one, and four processors as much as twice as fast as just two.
But doesn't it require more energy to run four processors than just two? It does, and that's why the default clock speed of multi-core processors is usually lower than that of a single or dual core chip. The idea is that their default clock speed can be lower to an extent where they don't need more power than a simpler processor and still work faster. As software becomes more and more sophisticated, it can take more and more advantage of multiple cores. And that's why more cores are almost always better.
Do note, however, that not all cores are the same. Very inexpensive Intel Atom processors have four cores, but because those cores are much simpler than those used in the Intel Core processor line (could they not have thought of a better name??), their performance is much lower, usually lower than even mid-range dual core Core processors.
Integrated Graphics — Graphics were not always included in processors, but now they are. This means computers no longer need a separate graphics card. But are graphics integrated into processors powerful enough? That depends. They are certainly good enough for everyday work, but they get bogged down when it comes to graphics-intensive use, such as gaming, rendering or 3D work. In that case, “discrete” graphics can provide a substantial boost that’s reflected in overall benchmark results. Note, though, discrete graphics do not speed up everything across the board; the boost will come in certain areas.
Limits of miniaturization -- Computers live in a world of ones and zeros. Gates are either on or off, and that is the basis of all computation. As long as those gates can signal one or zero, it doesn’t matter how big or how small they are. But since ever more powerful CPUs use ever larger numbers of those one-or-zero gates — many billions by now — size does matter in several ways. Smaller size means less material and less heat to be removed. Smaller size also means less travel distance between gates, and with billions of gates, distance traveled becomes an issue even at the speed of light, as electrical current does.
Advance in CPU miniaturization is often described in how close together transistors can be. That is measured in nanometers. A nanometer is one billionth of a meter or roughly 3.3 billionth of a foot. To put that in perspective, a human hair is about 100,000 nanometers wide, and the corona virus about 100 nanometers. Intel's 11th generation Core processors are based on 10 nm "process" technology, a tenth the width of a single corona virus!
The Intel 8088 chip in the first IBM PC used 3,000 nm technology, but for the modern Core chips it all started with 45-nm technology in the first generation of Core processors ("Nehalem"), and that was seen as a huge improvement over the 65nm and 90nm of older chips. The second generation ("Sandy Bridge") got that down to 32nm, and the third ("Ivy Bridge") to 22nm, but then it got progressively harder. The 5th generation ("Broadwell") used 14nm process technology, and that's the way it stayed all the way to the 10th generation ("Cannon Lake/Ice Lake"). There were tweaks and optimizations in between, but overall current technology is approaching physical limits.
About power conservation -- With processors becoming ever more complex, having ever more transistors, and running at ever higher clock speeds, don't they use more power? The simple answer is, they don't. In fact, due to miniaturization and many other tricks they use less. There are a great number of factors that affect how much power a computer uses.
Power usage hardy matters when you're plugged in; it only becomes a factor when you run on battery. Mobile systems will run on battery some or most of the time, and a variety of components other than the processor use power -- memory, storage, the display and others. The processor, however, is the biggest, or among the biggest, energy user, and so it is important how the CPU handles power conservation.
In mobile computers, power should only be consumed when it's actually needed for work. That's why displays dim or go to sleep when not used, and that's why modern processors have various "sleep" states they go into as soon as there's no work to be done. Designing and building these sleep state mechanisms is complex and the most comprehensive implementations are usually reserved for higher end processors. That’s the reason why we often see machines with very powerful processors being more power-efficient than systems with far less powerful chips. High-end Core processors have all those power conservation technologies and inexpensive Atom chips don't.
Note, however, that as a user you can configure your system for minimal power usage in the display, sleep and power settings. That can have a major impact on battery life.
Core i3 chips — viable alternatives? -- Ever since Intel came out with their Core line, the general approach has been to judge and select processors by their i3/i5/i7 “good/better/best” naming. By and large, the good/better/best thing is an oversimplification and really as much a marketing tool and up-sell ploy as it is a description of real value. There IS generally a relatively big difference between i3 chips on the one side, and i5/i7 chips on the other. i3 chips often don't use "hyperthreading," a technique where one physical core is split into two virtual cores so it can get more things done faster.
An i3 processor may have two cores whereas the i5/i7 cores have also two but it is more like four due to hyperthreading. Likewise, i3 chips often have no "turbo" speed. "Turbo" speed means that a CPU can briefly run faster than its default clock speed, getting things done quicker. That, however, requires careful temperature monitoring at various places throughout the system. As soon as the temperature exceeds a certain limit, turbo speed is throttled back down. i5 and i7 chips can take advantage of temporarily run much faster when maximum processing power is required. Many or most i3 chips can't. That can make a big difference. On the other hand, it also makes i5 and i7-based system performance less predictable.
Those baked-in Intel technologies -- Just like car manufacturers equip their vehicles with various extras and enhancements, Intel does the same with their processors. Special Intel technologies add a variety of functionality to processors. As you'd expect, i3 chips have the least, i5 chips more, and i7 chips the most. Most of these technologies are quite subtle and generally not necessary for the basic running of a system. They MAY come in handy by handling certain things faster and/or better, and there may be programming that requires them. Overall, almost all of those technologies are way too arcane and opaque to mean much, if anything, to the average buyer.
Recommendations -- when you buy a computer, be it a single machine for work or a thousand machines for an enterprise, think things through. It may be tempting to go for the lowest price but that may come back to bite you in terms of quicker obsolescence and higher maintenance and upgrade costs during the life of the machine.
Likewise, it may be tempting to go for the biggest and best, but will you be paying a premium for technology that'll quickly be updated or come down in price anyway? It absolutely pays to assess costs and benefits.
If your workload is predictable, there's little need for massive turbo boosts. If you don't store much data on the machine, there's no need for a huge disk. If you don't play advanced games or do 3D renderings, there's no need for advanced graphics.
On the other hand, going with obsolete technology will inevitably drag you down. An example was hanging on to rotating hard disks when solid state disks became available. That was a bad move, even though the hard disks were cheaper. And now, hanging on to, for example, obsolete SATA storage when much faster PCIe NVMe storage is available is almost masochistic. And, as long as we're talking non-CPU related matters, staying with a lower resolution display when higher resolution options are available is never a good idea. And let's not even get into fans... Nothing is more annoying when you're trying to concentrate and get work done than a loud, whining fan.
So be smart. Think things through. Assessing pros and cons, considering your needs and requirements from all angles, and only then making decisions is something we humans still do better than any computer.
Posted by conradb212 at March 24, 2021 8:34 PM