« Thunderbolt 4 and other tech issues | Main | If I were on the board of a rugged device provider... »
February 20, 2022
Intel’s 12th generation “Alder Lake” Hybrid processors
Contemplations about a whole new future of mobile chips in rugged laptops and tablets.
It isn’t easy to keep up with Intel’s prolific introduction of ever more processors, processor types and processor generations. Or to figure out what truly matters and what’s incremental and more marketing than compelling advancement. That’s unfortunate as Intel, from time to time, does introduce milestone products and technologies. With their 12th generation of Core processors the company is introducing such a milestone, but it isn’t quite clear just yet what its impact will be.
So what is so special about the 12th generation, code-named Alder Lake? It’s Intel’s first major venture into hybrid processors. Most are familiar with the hybrid concept from vehicles where the combination of an electric and a combustion engine makes for more economic operation — better gas mileage. But what does “hybrid” mean in the context of processors? Better gas mileage, too, but in a different way. Whereas hybrid vehicles employ two completely different technologies to maximize economy — electricity and gasoline — hybrid processors use the same technology to reach the goal of greater economy.
Specifically, Intel combines complex high-performance cores — p-cores — with much simpler economy cores — e-cores — to get the best of both worlds, high performance AND economical operation. This means that this kind of “hybrid” is a little different from the “hybrid” in cars.
In hybrid vehicles the combustion engine shoulders most of the performance load whereas the battery-driven electric motor helps out and may also power the car all by itself at low speeds. Part of the kinetic energy inherent in a moving vehicle can be recaptured when braking by charging the battery, making overall operation even more economical. This by now very mature technology greatly improves gas mileage — my own hybrid vehicle, a Hyundai Ioniq gets around 60mpg. While mostly employed to boost economy and reduce emissions in the process, hybrid technology can also be used to increase performance. There are sports cars where electric motors optimize and boost peak performance, and still help make overall operation more economical.
Hybrid technology, by the way, is not the only similarity between automotive and computer processor performance. Just like vehicles use turbochargers to boost performance, Intel uses “turbo boost” to achieve higher peak performance in its processors. It’s not really the same, but close enough.
In combustion engines, a turbo charger uses exhaust pressure that would otherwise go to waste to increase intake pressure and thus the engine’s performance. Higher performance requires better cooling and meticulous monitoring so that the engine won’t blow up. Whereas turbo engines initially catered primarily to peak performance, turbos are now commonly used to COMBINE better economy with robust performance on command. Relatively small 2-Liter 4-cylinder turbo motors have become ubiquitous.
In computer processors, “turbo boost” simply means increasing the clock frequency of the processor while very closely monitoring operating temperature. A chip can only run at “turbo” speed for so long until it needs to slow down and cool down.
How does “hybrid” in computer processors come into play? In essence by combining different types of computing cores into one processor. The most common approach is combining performance cores and economy cores, but it could also be combining cores designed for different kinds of work. In Intel’s 12th generation, the company combines high performance cores — the latest and greatest versions of its “Core” branded processing cores — with much simpler economy cores — the kind used in Intel’s low-end “Atom” chips. How much simpler are those “economy” cores? Well, when you look at a magnification of an Alder Lake chip, you see squares that are the performance cores and then the same size squares that includes four economy cores.
Given Intel’s practice of offering high-end Core chips and lower-end Atom chips, why combine the two? After all, while the high-end Core chips clearly use much more electricity when they are under full load, their power conservation measures have become very effective. So much that when systems just idle along, Core-based computers often use LESS power than Atom-based systems. Since many computer idle along most of the time, does a complex hybrid system really make sense? It can, and Intel certainly has its reasons for banking on the hybrid approach.
One such reason undoubtedly is that the ARM processors that power almost all of the world’s billions of smartphones have been using hybrid technology for many years. When you look at the tech specs of your octa-core smartphone you’ll most likely find that it has four economy cores and four performance cores, or perhaps even three types of cores. That’s because in mobile systems battery life matters just as much (or more) than the performance needed to effortlessly drive the latest apps. Intel doesn’t want to fall behind in those optimization technologies. Apart from that, as there are more and more server farms and other mega-processing systems, cooling and power consumption are becoming ever more pressing issues. The era of gas guzzlers is over there, too, and it’s time to optimize both performance AND economic operation.
Hence the Alder Lake hybrid chips.
Forget all about the way things used to be. Four cores no longer automatically means a total of eight threads. The total number of threads now is the number of power cores plus the number of threads (only power cores can have threads) plus the number of economy cores. An Alder Lake chip may have 14 cores and a total of 20 threads, those consisting of six performance cores, which means 12 threads, plus eight economy cores. Or a lower-end chip may have eight cores and 12 threads, consisting of four power cores, their four extra threads, and four economy cores.
But what about the thermal design power (TDP) of those new chips, you may ask — TDP being the maximum heat expressed in watts that the system has to be able to handle, and also sort of a measure of the general performance level of an Intel chip. Well, with Alder Lake there is no more TDP. The single number TDP had already been split into a TDP-up and TDP-down with the “Tiger Lake” 11th generation of Core processors, so that the former Intel U-Series of mobile chips that traditionally had had a 15 watt TDP were now listed as a manufacturer-configurable 12/28 watts. Alder Lake disposes of TDP entirely and lists “Processor Base Power,” “Maximum Turbo Power,” and “Minimum Assured Power” instead, all expressed in watts.
What does that mean? Well, the Processor Base Power is the maximum heat, expressed in watt, that the chip generates running at its base frequency. The Maximum Turbo Power is the heat the chip generates when it’s running at its maximum clock frequency, and that can easily be way more than the heat generated at the base frequency. Another difference is that prior to Alder Lake, Intel specified in a value they named Tau how long a processor could run at top turbo speed before it had to throttle back to cool down. Tau is no longer there. Does that mean it’s now up to manufacturers to make sure systems don’t overheat? I don’t know.
Now let’s move on to figuring out what those performance and economy cores are going to be doing. The overall idea is that the power core handle the heavy loads whereas the economy cores efficiently do routine low-load things that the performance cores shouldn’t be bothered with. Theoretically that makes the chips both faster and more powerful (because the power cores can fully concentrate on the heavy-loads) but also more economical (because it’s those miserly economy core that do the trivial everyday stuff without having to engage the power-hungry performance cores).
But who makes the decision of what ought to run where and when? Chip manufacturers on the hardware side and OS vendors on the software side have had to deal with that issue ever since the advent of multi-core processors long ago. That was relatively easy as long as all processors were the same, and became a bit more difficult when Intel also introduced hyper-threading (which makes one hardware core act like two software cores) a couple of decades ago. A scheduler program decided what should run where for best overall performance. With hybrid chips that contain different types of cores the task becomes considerably more difficult.
The scheduler program must now make sure that everything runs as quickly as possible AND as economically as possible. And for that to happen, Intel and Microsoft worked together on a hardware/software approach. On the hardware side, Intel Alder Lake chips have a special embedded controller — a Thread Director — that keeps track of a myriad of metrics and passes on that information to the operating system scheduler program that then uses all that information to make intelligent decisions. This teamwork between hardware and software, however, requires Windows 11. Alder Lake chips can, of course, run Windows 10, but to take full advantage of Intel’s new hybrid technology, you need Windows 11.
What does all of this mean for rugged computers?
On the mobile side, there’s the potential for longer battery life as well as better overall performance. However, a lot of things need to fall into place and work perfectly for that to happen in real life. We’ll know more once we get a chance to benchmark the first Alder Lake-based systems.
On the embedded systems and industrial/vertical market side, there’s even more intriguing potential. As is, many industrial PCs are built to provide “targeted” performance, i.e. just as much as is needed for the often completely predictable workloads and no more. That works fine and lowers the cost of hardware, but it can be frustrating when workloads increase or unexpectedly peak, and the performance just isn’t there. Other industrial PCs need all the peak performance they can get, but many are underused and just idling along much of the time. Alder Lake theoretically makes it possible to have a new generation of multi-purpose systems that can combine multiple types of workload into one single hardware box, with the Thread Director optimally assigning high-load and low-load assignments to the proper cores.
And for some very special applications that require both real-time and general purpose computing, these new chips could be used in conjunction with special-purpose operating systems that support virtual machines, doing different kinds of work on each, or groups, of the different types of processor cores.
When will all this be available? In stages. As of this writing (February 18, 2022), Intel’s website only shows eight Alder Lake mobile processors (plus a variety of desktop and embedded versions). However, more were announced the beginning of 2022, including a good number of P-Series and U-Series versions. P-Series chips are tagged “For Performance Thin & Light Laptops” whereas U-Series chips go under “Modern Thin & Light Laptops.” And Intel clarified the relationship between the old TDP and the new ways of naming things in its voluminous February 2022 documentation of the Alder Lake platform: “Processor Base Power” is just renamed TDP, TDP Down “Minimum Assured Power,” and TDP Up “Maximum Assured Power.”
This would place the 12th generation P-Series processors with a Processor Base Power/TDP of 28 watts higher than the 11th generation U-Series chips, most of which were listed as 12/28 watt designs with presumably a legacy TDP of 15 watts. P-Series chips have eight economy cores and two to six performance cores, for a total of 12 to 20 threads.
The 12th generation U-Series on the other hand seems to have two power levels, one with 9 watt legacy TDP and the other 15 watt legacy TDP, so the former would be more like the legacy Y-Series chips, and the latter would more or less correspond to what used to be the U-Series processors’ long-standing 15 watt TDP. If we assume that the 15 watt Processor Base Power/TDP will be the ones found in most higher-end mobile rugged devices, then we’re looking at Celeron, Pentium, i3, i5 and i7 chips with four to eight economy cores and one or two performance cores, and six to 12 total threads. That lineup is lacking a quad performance core chip and is not too terribly exciting, more like a hybrid of Atom and a couple of Core cores.
This may mean that P-Series mobile chips may carry the torch for high-performance rugged laptops and tablets, with eight economy cores and two to six performance cores. Only time and benchmark results will tell.
It ain’t business as usual, and it should be interesting to see how it all pans out.
For Intel's explanation of how their 12th generation Core hybrid technology works, see here.
Posted by conradb212 at February 20, 2022 9:38 PM