Discussing Lunar Lake’s Low Power Design: Can x86 Achieve Low Power?

There has been a persistent belief that the x86 instruction set is inherently incapable of low power consumption. Is this really the case? This article focuses on how the Core Ultra 2nd generation considers low power consumption and whether it is possible to achieve it…
For many years, there has been a rumor: x86 cannot achieve low power consumption, and Arm cannot achieve high performance. This emphasizes the differences in instruction sets. The oldest argument revolves around the battle between CISC (Complex Instruction Set Computing) and RISC (Reduced Instruction Set Computing), or it extends to the fact that Arm uses fixed-length instructions while x86 uses variable-length instructions, resulting in higher decoding power consumption for the latter, not to mention that x86 has a heavier historical burden.
Two years ago, we provided a detailed explanation in our article ‘The Rumor That x86 Cannot Make Low Power CPUs: Arm Remains Silent.’ If we narrow down ‘low power’ and ‘high performance’ to PC and workstation platforms, low power roughly refers to low-voltage/ultra-low-voltage processors in thin and light laptops, while high performance naturally refers to processors in desktops or workstations.
The release of Apple’s M series chips, including the Mx Ultra workstation chip with over 100 billion transistors, has shattered the bias that Arm cannot achieve high performance. However, on the x86 side, purely from the perspective of the PC platform, Intel and AMD processors seem to consistently lag behind Apple and Qualcomm in terms of device battery life.
Recently, at the IFA Berlin Electronics Show, Intel officially released Lunar Lake, which is the Core Ultra 200V series processor, targeting a power consumption range of 9W-33W for thin and light laptops. During our visits to various OEM manufacturers showcasing new laptops, we found that many advertised battery life figures exceeding 25 hours. Some laptops of the same series, one with Core Ultra and the other with Snapdragon, showed that the former had slightly better battery life than the latter.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
The Acer Swift 14 AI, which claims 26 hours of battery life.
In its promotion, Intel has devoted a lot of space to the battery life and energy efficiency of Lunar Lake laptops, directly comparing it to the previously praised Snapdragon X Elite. During the press conference, Intel representatives repeatedly mentioned that the advent of Lunar Lake has dispelled many erroneous claims that x86 cannot achieve low power consumption.
So let’s discuss how x86 processors achieve low power consumption.
Overview of x86 Low Power Achievements
Once again, it is reiterated that the ‘low power’ discussed in this article is limited to PC processors and does not extend to low power in mobile and embedded applications. Lunar Lake, as a processor aimed at thin and light laptops, continues Intel’s previous generations’ heterogeneous architecture of performance cores (P-core) and efficiency cores (E-core). What differs this generation is that the P-core has eliminated the hyper-threading design.
With a unified design of 4 performance cores and 4 efficiency cores, Lunar Lake is likely one of the products with the least core and thread count Intel has offered for laptops in recent years. However, the selected Xe2 integrated graphics and the fourth-generation NPU have significantly increased computational power.
The core and thread configuration of the CPU essentially determines that Lunar Lake is aimed at thin and light laptops—the target competitors being processors like Snapdragon X Elite and Apple M3. Moreover, in a strict sense, the TDP power setting for the high-end Snapdragon X Elite is generally higher than that of Lunar Lake.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
It seems Intel is taking a serious approach to specifically target low power thin and light laptops, likely recognizing the fierce competition in this market. During the media conference, Intel emphasized that ‘everything about Lunar Lake is related to energy efficiency.’
So how effective is it? Currently, only partial data released by Intel is available. First is the UL Procyon office productivity test, which tests the Microsoft Office suite. The Core Ultra 9 288V (Lunar Lake) performs about 7% better than Snapdragon X1E-80-100 (Snapdragon X Elite), with package power consumption reduced by over 50% compared to the previous generation (Core Ultra 7 165H), and it is also lower than that of Snapdragon X Elite.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
This means that with half the power consumption, Lunar Lake achieves slightly higher performance than Meteor Lake during office tasks. From the perspective of performance per watt, Lunar Lake is 2.29 times that of Meteor Lake and about 20% higher than Snapdragon X1E-80-100.
In addition, comparing battery life data, during the same UL Procyon office productivity test, Lunar Lake outlasted Snapdragon X Elite by 2 hours (20.1 hours vs 18.4 hours); if purely used for Teams meetings, the former falls short by 2 hours (10.7 hours vs 12.7 hours). This reflects a mixed outcome.
Intel also emphasizes that this is based on the same laptop brand and mold design, striving to keep the system design as consistent as possible. Notably, Intel’s official promotion mentions that Lunar Lake can achieve 20 hours of battery life, which should originate from this comparison. After all, one major audience for Lunar Lake thin and light laptops is white-collar workers using Office applications.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
In summary, comparing with Qualcomm Snapdragon X Elite and AMD Ryzen AI laptops, Intel’s office testing and Teams meeting testing results are as shown in the above image.
It is said that in this comparison, Intel may have slightly disadvantaged, as the AMD Ryzen model chosen for comparison was equipped with a 78Whr battery, while the Core Ultra 9 288V laptop had only a 70Whr battery capacity. We believe this set of data may be more valuable as it reflects the actual battery life users can achieve.
Overall data shows that in application scenarios such as office work, web browsing, video conferencing, and streaming 4K video playback, Lunar Lake has achieved up to a 50% overall power consumption reduction compared to the previous generation Meteor Lake. Moreover, this power consumption refers to the packaging power consumption of Lunar Lake that includes on-chip DRAM, while the previous generation did not include DRAM in its packaging.
Another set of data is the gaming power consumption comparison. This comparison mainly examines the iGPU power consumption. Intel chose to compare it with the previous generation Meteor Lake, and in the three games ‘Assassin’s Creed: Valhalla,’ ‘Cyberpunk 2077,’ and ‘Farming Simulator 22,’ Lunar Lake achieved up to a 68% higher gaming frame rate, while GPU power consumption was reduced by up to 35%.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
This comparison did not include Qualcomm Snapdragon because the actual performance of the Snapdragon X integrated graphics is not worth mentioning, and more than half of the Windows ecosystem games cannot run on it. At the IFA event, we saw Lunar Lake running ‘Cyberpunk 2077’ and ‘Shadow of the Tomb Raider’ at 1080p medium quality, achieving frame rates of 60-70 fps with the help of AI upscaling.
Why Can It Achieve Low Power Compared to the Previous Generation?
If the core of choosing Windows on Arm is battery life, then with Snapdragon X Elite representing the Arm camp, the battery life of laptops is no longer better than that of new products from the x86 camp, and one must also consider the currently barren state of the Windows on Arm ecosystem. Why would ordinary users choose Windows laptops with Arm instruction set processors?
This article has not provided some data, such as Intel stating that in the 9-23W power consumption range, the multi-threaded performance of Lunar Lake is 2-3 times higher than the previous generation Meteor Lake (multi-threaded performance ÷ total thread count), reflecting a significant leap in CPU efficiency; it even mentioned that this generation’s 8 threads outperform the previous generation’s 22 threads. Although we believe this comparison may be somewhat biased, as multi-threading performance requires sufficient power consumption.
According to Intel, Lunar Lake has 4 fewer cores than Snapdragon X Elite, but the latter reaches 50W packaging power consumption for SPEC2017 performance, which the former can achieve at around 30W. Therefore, Intel refers to Lunar Lake as ‘8-core magic.’ For more detailed data, it is recommended to refer to last week’s Lunar Lake release report.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
So how does Lunar Lake achieve low power consumption?
First, we must reiterate that the key to low power versus high performance does not lie in the instruction set itself. Jim Keller has publicly stated multiple times, ‘The relationship of instruction sets is not that significant.’ Over the past decade, numerous research papers have reached similar conclusions, indicating that the differences in power consumption and performance between Arm and x86 processors stem from differences in design goals, and the instruction set itself is not the deciding factor; the ‘implementation’ is ultimately the most important.
During media interviews at IFA, we received some valuable summary information regarding Intel’s achievements in low power consumption with Lunar Lake from Feng Dawei (Vice President and General Manager of Intel’s Client Computing Group and Client Segmentation). He stated:
‘After Meteor Lake was launched, the two LP E-cores plus cache can solve many problems. However, in many applications, especially productivity applications or non-performance metric applications, we found that some could run on the two LP E-cores of Meteor Lake, but fitting everything in is still challenging.’
The CPU design of the previous generation Meteor Lake included three core clusters: P-core, E-core, and LP E-core. The last low-power efficiency core was located in a low-power island, originally intended for achieving low power consumption.
Chip and Cheese previously analyzed that Meteor Lake did not fully achieve its low power goals for the LP E-core to some extent, partly because the LP E-core’s performance was too weak (low frequency); on the other hand, the LP E-core lacked L3 cache, and the L2 cache was too small, which greatly impacted the IPC of the LP E-core; leading to situations where running loads on the LP E-core would significantly affect performance and experience.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
The LP E-core of Meteor Lake (low-power efficiency core) is not on the same die as the other two core clusters.
Robert Hallock (Vice President of Intel’s Client Computing Group and General Manager of AI Technology Marketing) also mentioned briefly at the media conference, ‘Common loads such as productivity, video conferencing, and web browsing can waste a lot of power if not managed properly. We found that the LP E-core cluster of Meter Lake was still insufficient to meet performance goals.’
‘For example, with 10 people in a video conference, we found it was challenging to fit everything in,’ Feng Dawei said during the interview. ‘But with Lunar Lake having twice the efficiency cores and double the storage resources, it can fit everything in completely.’ ‘Many productivity applications, like Outlook and browsers, are background applications that only open when needed. So they can be accommodated.’
‘In daily applications, the reduction in power consumption is particularly noticeable.’ Of course, there is also the contribution of ‘Memory on package (on-chip DRAM), which reduces the overall packaging power consumption significantly compared to the previous separate CPU, and the CPU’s own power consumption has also decreased significantly.’ ‘Every part contributes. My own feeling is that the doubling of LP E-cores and the increased cache contribute significantly.’
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
The two dies stacked above are the DRAM memory packaged together with the processor.
Several Reasons Why Lunar Lake Achieves Low Power Consumption
From a high-level architectural design perspective, we believe that what Feng Dawei mentioned is indeed the most important factor for Lunar Lake to achieve low power consumption—the continuation and optimization of the low-power island concept. However, this is not all. We summarize the reasons for Lunar Lake’s low power and high energy efficiency as follows.
First, Intel has not mentioned that the Compute Tile of Lunar Lake is based on TSMC’s 3nm process. This means that the die where the CPU, GPU, and NPU are located is manufactured using a 3nm process. Wikipedia clearly states that this tile/die’s specific process node is TSMC N3B. In the above comparisons, Intel’s competitors generally still use a 4nm process, which is a generation apart.
Although Apple has demonstrated that 3nm may not be significantly better than 4nm, the update of the process node is still crucial for Core Ultra processors. A couple of years ago, Lisa Su summarized key factors for improving chip performance and energy efficiency during a press conference, and manufacturing process is a major factor. Design methodology is less impactful than manufacturing process. This statement may still require specific analysis for specific problems, but the manufacturing process is undoubtedly an important factor.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
Secondly, as Feng Dawei mentioned, based on the low-power island design concept, Lunar Lake pursues more active invocation of E-cores, thereby minimizing the awakening of high-power components under medium and low loads. This point is also tied to three sub-points:
(a) The performance of this generation’s E-core (Skymont) has significantly increased, not only greatly surpassing the previous generation’s LP E-core (Crestmont, reportedly 2 times single-core performance and 4 times multi-core performance) but also having a higher IPC than the previous generation’s Raptor Cove large core; the L2 cache has increased to 4MB; more importantly, the compute tile has added a global 8MB Memory Side Cache—which we believe adds significant value to the low-power island design.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
Of course, the microarchitecture design of Skymont itself is also inseparable from achieving low power for the loads running on it.
(b) More detailed power supply and power management solutions. This should be considered a conventional approach. For power supply, Lunar Lake has a total of 4 PMIC power management controllers, providing as many power rails as possible for different components, allowing the P-core cluster, E-core cluster, graphics, and memory control components to operate ‘independently.’
During the Lunar Lake architecture analysis meeting earlier this year, Intel mentioned that while achieving a ‘fine-grained voltage rail topology,’ it also ‘achieved enhanced telemetry to better discern power usage states for better control.’
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
It is worth mentioning that significant changes in power management include the efficiency optimization of the ITD (Intel Thread Director) thread scheduling assistance mechanism, such as enhancing sleep state power consumption and latency performance, load distribution, and frequency control based on machine learning, etc. ITD, as a key element in achieving low power consumption, is also a focus Intel has emphasized before; we have discussed this in our architecture analysis article and will not elaborate further here.
In this media conference, Intel summarized the four major features of the new ITD as: dynamic scheduling strategies; using a single E-core when appropriate; extending to other E-cores for multi-threading; and scheduling to performance cores as needed. Simply put, it is an overall strategy of ‘E-core priority.’ This still relies on the improvements listed in (a).
(c) Low latency inter-core communication. It is unclear whether this was specifically to mock Qualcomm and AMD, but during the IFA media conference, Intel also specifically released data on the inter-core communication latency of Lunar Lake CPUs and memory access latency, especially the latency from E-core to P-core being 55ns—readers familiar with Ryzen and Snapdragon should know that the inter-core communication latency between different clusters of competing CPUs reaches 3-4 times this number, which is one of the most criticized issues for the latter two.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
Listing this point as part of more active E-core invocation for achieving low power is because high latency in inter-cluster communication can also affect the basic design strategy of ‘E-core priority.’
Thirdly, other factors. This includes the P-core eliminating the hyper-threading design, on-chip DRAM reducing memory access latency and power consumption, Xe2 integrated graphics significantly improving efficiency, and technologies such as the AI-based self-tuning controller for enhanced power management, which allows core frequency adjustments in finer increments of 16.67MHz—all can be considered technologies that contribute to achieving low power consumption.
Among these, Intel states that packaging memory on-chip leads to ‘PHY power consumption reductions of up to 40%’; the efficiency improvements of Xe2 integrated graphics can also be seen in the actual comparisons in the three games…
The removal of hyper-threading design for the P-core is something we greatly appreciate upon learning about the microarchitecture design of Lunar Lake. Hyper-threading, in our view, is more of a technology product of its time. Although Feng Dawei did not disclose whether Intel plans to completely abandon hyper-threading technology, he mentioned: ‘In the era when desktops were dominant over a decade or two ago, our most important goal was to push performance to the extreme.’
‘Laptops equipped with Lunar Lake are focused on balancing performance, power consumption, and all other capabilities.’ ‘After introducing efficiency cores and hybrid architecture, we gradually realized that with hybrid architecture, having more efficiency cores is actually the optimal solution for multi-threaded performance power consumption ratio.’
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
Robert mentioned at the media conference that removing the hyper-threading design compared to hyper-threading can lead to a comprehensive improvement of 15% in perf/power/area, the key being to consider that this design may sacrifice some power and area benefits for overall better performance per unit area and power. ‘In other words, compared to hyper-threading design, achieving smaller and lower power CPU cores results in more performance.’
However, he also mentioned during his speech: ‘We’re not always going to make this same decision, but it makes a ton of sense when you’re fanatical about power like we were in Lunar Lake.’ This indicates that the decision to remove hyper-threading was reasonable considering power consumption requirements for Lunar Lake. But ‘we’re not always going to make such decisions’ may imply that hyper-threading could reappear in future cores.
Looking Ahead to the Thin and Light Laptop Market
Intel summarizes the key innovations of Lunar Lake in terms of power consumption as shown in the image below:
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
This article covers all the components involved. These improvements will bring significantly longer battery life to laptops at a system level, reducing battery anxiety. However, Intel did not directly compare Lunar Lake with Apple’s M series chips, which is somewhat disappointing.
Therefore, many OEM manufacturers showcased new laptops at IFA with claimed battery life exceeding 25 hours—though these figures may not be entirely representative of daily use, they are at least 8 hours longer than the claimed battery life of Meteor Lake laptops. Finally, long battery life, along with ensured compatibility, Windows laptops are finally on the way.
Setting aside the myriad negative news Intel has faced in the past two months, as well as the uncertainty surrounding Arrow Lake to fill the Lunar Lake market gap, we believe that Lunar Lake carries a bit of the boldness Intel displayed when transitioning from the Pentium 4 to the Pentium M architecture and subsequently to the Core platform.
As long as nothing unexpected happens with Lunar Lake, the Core Ultra 200V series thin and light laptops expected to hit the market later this year are indeed something to look forward to; they have made a good start in such a market environment.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
The MSI gaming handheld expected to use Core Ultra 200V series processors is showcased.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
LG Gram seems to be on the list of first adopters of Lunar Lake.
Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?
The Asus Zenbook series is also a key product in this release.

Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?

Recommended Hot Articles

  • Shenzhen Autonomous Driving Ride-Hailing Cars Begin Highway Testing

  • Dissecting Earphone Cases: Overvoltage Protection Circuit Analysis

  • I Found the Meaning of Black Myth Wukong in Disassembling PS3!

  • Xiaomi Loses 60,000 on Each Car Sold? Wouldn’t Not Buying Save You Money?

  • Disassembling the Old IBM ThinkPad Optical Drive: Some Surprises

Discussing Lunar Lake's Low Power Design: Can x86 Achieve Low Power?

Related posts

  1. Exploring Low Power Design in Microcontrollers
  2. Creating a Versatile Retro Mini Console with Frontend and Independent Emulators
  3. Comprehensive Overview of the 18 Major Industry Chains, Including the Latest Aircraft Industry Chain
  4. Low Power Design Strategies – Resource Sharing
  5. Nine Key Techniques for Low-Power Processor Design
  6. Low Power Design of Autonomous Driving Chips
  7. Overview of Low Power Design
  8. Best Programming Model for Low Power Design: Asynchronous Programming
  9. Low-Power Design: Retention Cell
  10. Introduction to Low-Power Design in Digital ICs (Part 5)
  11. Analysis of Common Diagnostic Methods for Profibus Bus and Interference Handling
  12. Practical Issues to Know When Flashing ECU – Engine Management Vol.01
  13. Digital Home System Base: Infrastructure Software and Hardware Solutions Analysis (Part 2)
  14. The Four Essential Tools of ASIC Design
  15. Low Power Design: The Energy-Saving Code for Electronics
  16. ASIC Chip Design Enhances Safety and Reliability of Electronic Detonators
  17. Installing Smart Smoke Detectors Increases Escape Chances by 50%!
  18. Understanding China Unicom’s NB-IoT: Wide Coverage, Low Power Consumption, Massive Connectivity, and Low Cost
  19. The Importance of Screen Calibration: A Look at SoC Mobile Platforms

Leave a Comment