The chips of the Intels Meteor Lake will certainly be used as a driving engine for AI specific tasks on the PC. But Intel also uses AI to be applied, and for the chip, and the specific AI is applied to how it moves the power and transitions between entire and low-power states.
In 2008, the Intels Centrino platform recited a phrase describing the company’s power philosophy : HUGI, or Hurry Up and Get Idle. It was an acknowledgment that meeting the need for a low-power processor forced it to get all the work needed to be done as quickly as possible. That printer could return to sleep in low-power position.
Everything has changed. AI coincidentally or not, often called Intels Centrino moment, is also factoring heavily into Meteor Lakes power management, the Intel executives said at the Hot Chips conference at Stanford University. Originally, Intel referred specifically to Meteor Lake in its program synopsis, but instead drew a more generic talk called Intel Energy Efficiency Architecture.
In any case, the new AI power plan will be used for future products according to Efraim Rotem, a client of Intels Design Engineering Group. In two months, Intel will launch its new client processors, which will use these new features, he said.
Intel demonstrated how the new AI(a) compared to the old algorithm in its Hot Chips presentation. Remember that power is falling in power.
The problem is simple. Rotem said, we really want to understand this subject. We want to take immediate action, so we won’t want to wait too long.
The typical solution is to turn the processor’s power to the speed and get it done faster. The CPU then must figure out when the job is finished, and the processor can switch to low-power state. This is called Dynamic Voltage and Frequency Scaling, or DVFS. The question of power management is how to determine the right frequency, says Rotem.
Intel first developed the foundation of this decision-making in the 6th-generation Skylake system with a speedy switch called Speed Shift. In intelligent fashion, the technology shifted from the power supply to the idle speed. Speed Shift used a standardized estimate of how humans opened and closed a web page, for example.
When Meteor Lake was introduced to AI, Intel has now shifted again to AI. The algorithm knows and predicts how a user will open a web page, scan and close it, and move on. The same algorithm was applied to many other tasks. What’s different is that the algorithm taught itself, extracting patterns of behavior that are more detailed than what Intel originally programmed in.
That will improve Meteor Lake, bringing up to 35 percent more responsiveness the reaction time in which the CPU can rev up to a high-power state. Rotem made a distinction between the work from energy over time. The energy divided into the power consumption of this work, just the general power consumption.
The idea is to give the processor the energy budget that it needs for the time it needs it, and no more. From hearing the questions from the audience, Rotem revealed that there’s room for improvement: The AI is training on specific scenarios. And offline: It has already been trained and won’t react dynamically to individual user preferences. To the best of it, your PC doesn’t learn to act fine, or at least not in this generation. Rotem suggested that different AI models can apply to different scenarios.
Intel s Rotem also suggested that the popular performance per watt metric might be outdated.
Most laptops have a maximum day’s use, according to Rotem. The desktops spend nearly 100 minutes in the same state. As the processors become more efficient, he said, the ratio between the thermal design power and the energy consumed over time will drop in the manner that the chips’s processing are more effective.
We know Intel plans on discussing its latest clients at its Intel Innovation conference in San Jose on September 19. It seems that energy efficiency may be one of the components of Meteor Lake.