Navigating design tradeoffs and adding value to business-class notebooks
“Cost, Schedule, and Performance: choose any two” is a famous expression employed by engineers worldwide for many decades and serves as an often needed reminder that, in the world of product development, there are no free lunches. Design engineers and product planners in every industry must continually navigate tradeoffs throughout the product development process to be competitive.
When it comes to notebook thermal design, there is a similar network of competing design attributes each needing to be weighed against a continuously shifting set of customer perception and usage trends. Furthermore, because no two customers are exactly alike in their prioritization of product features, notebook product planners (the individuals charged with making critical product definition decisions) are forced to make predictions about which feature hierarchy best aligns with the largest group of potential customers, all without the benefit of a crystal ball.
“Cool, Quiet, and Productive” are three notebook qualities that are perpetually at odds with one another. Add to that list “thin, light, long-lasting battery, and affordable” and what are you left with is a rather entangled list of competing customer demands.
Application performance produces heat from internal components like CPUs and GPUs. That heat will result in your notebook’s skin getting hot to the touch or you can cool it with a fan that spins faster. Fast spinning fans create noise. You can reduce the fan speed and noise by using a larger heat sink which demands less air velocity. Larger heatsinks require a smaller battery or a larger notebook. Smaller heatsinks can be just as quiet as larger heatsinks if they use better materials like copper fins, vapor chambers, and graphene spreaders but these performance boosters add significant cost. And round the tradeoff merry-go-round we go!
In this paper, three 14-inch business-class notebooks from Dell, Lenovo, and HP were evaluated to understand how each vendor navigated several critical thermal design tradeoffs. The notebooks tested include: • Dell Latitude 7410 • HP EliteBook 840 G7 • Lenovo ThinkPad X1 Carbon Gen8
The results from the study are presented to compare the relative strengths and weaknesses of each product using 3 Key Performance Indicators (KPIs): • Application Performance • Skin Temperature • Sound Output
Spoiler Alert: The Dell Latitude 7410 was able to utilize a highly-optimized CPU power management strategy and choice heat spreading materials to distance itself by as much as 22% from the competitive platforms in performance tests while, at the same time, winning in sound output and remaining competitive in skin temperature.
System Configurations For this test of 14-inch business-class notebooks, every effort was made to select configurations that were as identical as possible in order to create a true apples to apples comparison.
Table 1. System configuration details for the three notebooks under investigation.
Notebook Chassis Materials Both the Dell and HP notebooks utilize an Aluminum chassis whereas the Lenovo ThinkPad employs a combination of Carbon Fiber and Magnesium. The Carbon Fiber + Magnesium combination is substantially lighter than aluminum, however, the tradeoff is that the effective thermal conductivity of Carbon Fiber and Magnesium is lower than Aluminum. The reduced thermal conductivity of Lenovo’s selected chassis material, in this case, makes it more susceptible to localized hot spots on the surface of the notebook and can, potentially, drive uncomfortably warm skin temperatures under certain conditions.
Cooling Solution All three notebook designs utilize a similar CPU cooling strategy whereby the CPU die is thermally connected to a remote heatsink fin stack via copper heat pipe. Each heat sink is directly attached to a thin blower-style fan that draws cool air up from the bottom of the notebook and pushes it through the heatsink fins to dissipate the CPU heat into the air.
Below are a few core principles of thermal engineering that can be useful to grade a notebook’s cooling solution: • Heat sinks dissipate CPU heat into the air by spreading the heat across an array of fins. The total surface area of the fins in a heatsink is a strong indicator for how the heat sink will perform; the larger the total surface area, the more heat that can be dissipated, though, this usually implies that the heatsink takes up more space.
• Heat pipes are used to transport heat from point A to point B. Heat pipes are useless, by themselves, to reject heat into cooling air and must be bonded to a heatsink for that purpose. The heat pipe(s) is(are) typically the costliest single entity within a heatsink assembly. Heat pipes are sized for a maximum amount of heat dissipation typically called “Qmax”. Exceeding the Qmax value for a heatpipe will cause a “dry-out” condition that results in large temperature spikes. Larger heat pipes can support a larger Qmax value than smaller heat pipes. An undersized heat pipe selection can create a power ceiling or thermal limit for an otherwise well designed heatsink.
• Fans and blowers are typically used to push airflow through a heat sink. In general, larger fans can move more airflow and provide better cooling than smaller fans when operating at the same speed though a larger fan will usually emit more noise than a smaller fan at the same speed.
At first glance, it would appear that the HP EliteBook 840 G7has stolen the show with a biggest/best in test heatsink and fan solution. This is partly true; the HP heatsink has
over 2x the surface area compared to the Dell Latitude heatsink, however, the HP cooling solution is somewhat crippled by the lowest capability heat pipe attaching the
CPU to the heatsink. While the estimated 43 Watt Qmax value for the EliteBook’s heatpipe will support the steady state PL1 needs of the Intel i7-10610 CPU, it does make
prolonged dwell at elevated PL2 states a bit of a challenge. This will later be shown to be a bit of a handicap on the 840’s performance overall.
Battery Capacity Last but not least is a quick look at how much space within the very similar 14-inch notebook chassis dimensions each vendor reserved for the internal battery as shown below in Figure 2.
Figure 2. Battery capacity for each notebook included in the test The Dell Latitude 7410 seems to be making up for a relatively petite heat sink volume by offering up an extraordinarily large battery option at 68 Whr. This, again, speaks to the design tradeoffs encountered throughout the product development process and points towards Dell’s prioritization of sustained mobile productivity over raw cooling capability.
Application Performance “Productivity” Application Performance is the first, and possibly most important, KPI to be examined in the business-class notebook product group. After all, very few business customers will purchase a notebook just to have something “cool” to rest their palms upon. Business- class laptops are purchased to do work; they are a tool and the heart of the contemporary office productivity engine. One’s ability to quickly push through stacks of work, in many cases, is limited by how quickly your computer can cruise through applications. Application performance equates to productivity.
The question of importance then becomes: “which notebook allows for me to maximize productivity?” The challenge to answering that question is simply that everyone uses their business notebooks differently with varying amounts of time spent in a host of unique software applications.
In order to meaningfully answer the question of system productivity, three diverse, office and creativity-oriented benchmarks were selected to characterize each notebook using a broad range of software applications. Each benchmark suite attempts to characterize system performance by stressing the notebook with the same or similar applications used by business professionals every day.
The three selected benchmark suites:
• PCMark 10
• SYSMark 25
• Cinebench R20
UL’s PCMark 10 is a comprehensive PC benchmark utility designed to cover a wide variety of tasks common to the workplace environment. The PCMark 10 benchmark is comprised of 8 unique workloads organized into three categories: • Essentials - Video Conferencing - Web Browsing - App Start-up • Productivity - Writing - Spreadsheets • Digital Content Creation - Photo Editing - Video Editing - Rendering and Visualization While there are likely some end-user applications omitted from this selection, in reality, most business users will find at least a few activities listed above that resonate with their daily relationship to their PC. The nice part about PCMark 10 is that it allows for the performance of each workload to be evaluated independently rather than only blending all of the scores together. This means that customers purchasing a system primarily for spreadsheets and web-browsing should not be dissuaded from a product that scores high in those areas but low on rendering and visualization as the latter category will have very little impact on their usage experience. PCMark 10: Essentials The PCMark 10 Essentials category seeks to reflect the most common, everyday usage scenario for PCs. Figure 1 below shows how each system performed in each of the 3 Essential workloads and their combined Essentials score.
Figure 1. PCMark 10 Essentials application performance scores. With the exception of Video Conferencing, the Dell Latitude 7410 outpaces the HP and Lenovo systems in all Essentials workloads besting the Lenovo X1 Carbon by just over 22% for the combined Essentials category benchmark.
Figure 2. PCMark 10 Productivity application performance scores.
The story for Productivity is similar to that of Essentials. The Dell Latitude 7410 decidedly wins both writing and spreadsheet application benchmarks. PCMark10 uses the Libre Office suite rather than Microsoft Office for testing Excel and Word type workloads but they should have similar performance characteristics.
Figure 3. PCMark 10 Digital Content Creation application performance scores. The PCMark 10 Digital Content Creation workload category is where the performance story tends to flip between the three 14-inch notebooks. Where the Dell Latitude 7410 is the clear winner for Essential and Productivity type applications, it lags both the HP and Lenovo products for Digital Content Creation.
Figure 4. PCMark 10 Overall application performance scores.
Figure 4 shows the overall, combined scores for each of the PCMark 10 workload categories where the Dell Latitude 7410 claims victory over the 2 competitive systems.
SYSMark 25, from BAPCo, is an office-centric benchmark application that attempts to quantify system performance by executing predefined scenario snapshots from today’s most popular office applications including Microsoft Office 2019, Adobe Product Suite, Google Chrome, and a long list of other commonly used software. What makes the SYSMark benchmark unique from other popular office benchmark applications such as PCMark 10 is that SYSMark actually executes performance tests using the household-name software itself rather than synthetic/proxy applications intended to behave
similarly to the applications most people use. As an example, SYSMark 25 installs and tests a PC’s performance running the actual Microsoft Excel 2019 application rather than trying to fabricate a custom code to test spreadsheet performance.
Similar to PCMark 10, SYSMark 25 categorizes application tests into three functional categories: • Productivity - Adobe® Acrobat® Pro DC - Audacity 2.3.2 installer - AutoIT 3.3.15.2 - Corel WinZip 24 Enterprise - Google Chrome™ - Microsoft® Excel® 2019 Professional Plus VL - Microsoft® Outlook® 2019 Professional Plus VL - Microsoft® PowerPoint® 2019 Professional Plus VL - Microsoft® Word® 2019 Professional Plus VL - Shotcut v19.09.14
• Creativity - Adobe® Lightroom® Classic CC 2019 - Adobe® Premeire® Pro CC 2019 v20.0.6 - Adobe® Photoshop® CC 2019 v13.1.5
• Responsiveness - Adobe® Acrobat® Pro DC - Adobe® Lightroom® Classic CC 2019 - Adobe® Premeire® Pro CC 2019 v20.0.6 - Adobe® Photoshop® CC 2019 v13.1.5 - Google Chrome™ - Microsoft® Excel® 2019 Professional Plus VL - Microsoft® Outlook® 2019 Professional Plus VL - Microsoft® PowerPoint® 2019 Professional Plus VL - Microsoft® Word® 2019 Professional Plus VL
While SYSMark has the more relevant collection of office-centric software in selected benchmarks, its one glaring deficiency is that it does not provide application-specific results, but, rather, provides blended scores for each of the 3 performance assessments.
Figure 5. SYSMark 25 Overall application performance scores.
Figure 5, above, shows the relative performance of each notebook for the three workload categories of SYSMark 25. Once again, the Dell Latitude 7410 pulls ahead of the other two systems in office-centric benchmarks by scoring 10% higher than the HP EliteBook and 13% higher, overall than the Lenovo ThinkPad X1 Carbon.
The final application performance benchmark for this series of tests is Cinebench. Cinebench uses the popular Cinema 4D engine to render a single, common image. The elapsed time that it takes for the application to complete the rendering job is then translated into a performance score where scoring higher means rendering the image faster. This benchmark is most applicable to those who work in the content creation industry or with media-centric applications daily. It should be noted that, while boasting of Cinebench scores is fairly common amongst overclockers and power users, it is debatable how relevant it is to this segment of business class notebook. Users dependent upon rendering applications for their daily lives are likely to find better value in other product segments. With that said, figure 6 below shows the results for the three
14-inch notebooks running Cinebench R20.
Figure 6. Cinebench R20 rendering performance scores shown for the number of consecutive rendering passes.
Cinebench, like the Digital Content Creation test within PCMark 10, creates a long-dwell, intensive, CPU stress that exposes the very limits of a product’s cooling capability. As the number of photo rendering passes increases, the effects of cooling limitations become more evident. The Dell Latitude 7410, which performs very well in office-centric benchmarks, begins to fall behind the pack as the number of Cinebench rendering passes goes up due to its smaller CPU heatsink and fan.The Lenovo ThinkPad X1 Carbon wins the rendering category though with some sacrifices in the skin temperature department as will be seen in the following set of tests.
Reconciling Performance and Cooling within Notebooks All three tested systems utilize the Intel Core i7-10610 10th generation (Comet Lake) CPU. So why do three systems utilizing the same CPU model seem to perform so differently from one another on the same benchmarks?
There are a few laws of nature that apply to CPUs that can help explain the performance differences: 1. Application performance increases proportionately with increases in CPU frequency (GHz) 2. CPU Power consumption (Watts) increases proportionately with increases in CPU
frequency (GHz) 3. CPU Power consumption (Watts) equals heat dissipation (Watts) 4. Given 1-3, application performance increases require an increase in heat
dissipation for a given CPU.
Starting back in 2009, Intel introduced a feature with their “Nehalem” architecture CPUs known as TurboMode (sometimes called Turbo Boost) whereby the CPU will automatically adjust its own operating frequency based upon application demand to help improve performance. Intel placed limits on how high the CPU frequency could increase based upon the power and cooling limitations of the CPU itself. This power limit (often called thermal design power or TDP) effectively created a cap on CPU frequency and application performance. As Intel’s Core architecture matured over time, TurboMode was improved to allow mobile CPUs to dynamically control not only their frequency, but also their own power limits based upon software demand and hardware capability.
TLDR: Today, notebook designers have the ability to control their own CPU Power Limit (hence forth referred to as PL1) and, thereby, control their ability to maintain elevated CPU frequencies (application performance), based upon how much heat their system can cool. The need to optimize a platform’s CPU power, frequency, fan speed, and application performance goals has never been more pronounced nor as fruitful as it is today.
Despite having the smallest CPU heat sink, the Dell Latitude 7410 has managed to create value and differentiation in the business-class notebook segment through masterful execution of an optimized CPU power limitation policy. Figure 7 below shows the unique power limitation (PL1) policies employed by Dell, HP, and Lenovo throughout the PCMark 10 benchmark. There are a few things to notice in the chart:
• The HP EliteBook has a constant PL1 limitation of 18 Watts. Under no
circumstances does the HP notebook allow itself to exceed this limitation, even for
short, opportunistic bursts.
• The Lenvovo ThinkPad defaults to a very low PL1 of 7.5 Watts but then
immediately bursts to 51 Watts upon application demand with very little granularity
in between.
• The Dell Latitude 7410 maintains a high “resting” PL1 limit of 22 Watts and then
modulates it down as cooling capacity is consumed over time. This high “resting”
PL1 ensures that the CPU can achieve maximum frequencies at the instant an
application demands it. The result is significantly higher performance in virtually all
mixed-use “bursty” workloads common to the office environment.
Figure 7. CPU PL1 Power Limits for the PCMark 10 benchmark.
Figure 7 reveals how Dell, despite having the smallest cooling solution in the test segment, is able to achieve the highest application performance scores through thoughtful PL1 optimization that effectively maximizes the total time spent at elevated power and frequencies or the “area under the curve”. By having a higher “resting” PL1 limitation, the Dell Latitude 7410 feels extremely responsive to opening new applications and kicking off large Excel calculation updates.
Productivity Winner: Dell Latitude 7410
Productivity Loser: Lenovo ThinkPad X1 Carbon Gen 8
Skin Temperature
The surface temperature of a notebook has, for many years, been somewhat of an afterthought in the purchasing decision process for business customers. As enterprise and office applications become more demanding of hardware resources, the amount of heat dissipated by increasingly smaller notebooks has skyrocketed. Too often these high levels of heat dissipation rear their ugly head in the form of elevated skin temperatures making the system uncomfortable for prolonged use. Palms resting upon the notebook, fingertips upon keypads, and even the bottom surface of the notebook resting upon one’s lap during a critical meeting can all be sources of user discomfort and frustration.
Today, notebooks use integrated temperature sensors to monitor how warm the device is getting in critical touch locations. This skin temperature information is then used to control fan speeds and or limit the CPU power (heat production) in order to maintain comfortable touch temperatures during operation.
In order to quantify the relative skin temperature performance of each system, an IR camera was used to map out the hottest locations where user skin might contact the notebook. Once the hottest locations were identified, direct contact thermocouples were attached to each of the following locations:
• Left and Right Palm rests
• Touchpad
• Hottest Keyboard cap
• Keyboard rib
• Hottest location on top surface of notebook
• Hottest location on the bottom surface of notebook
Figure 7. Skin temperature thermocouple instrumentation for HP EliteBook 840 G7
Each test system was operated in a temperature-controlled environment at approximately 23°C (74°F) and skin temperatures logged over time throughout the entirety of each performance benchmark application. Figure 8 presents the results of the skin temperature logging for each of the three notebooks included in the investigation.
From a thermal design perspective, metal temperatures can be loosely assigned to the following categories:
•Less than 45°C: Warm but acceptable
•Between 45°C and 55°C: Uncomfortable or annoyingly warm
•Above 55°C: Potential burn hazard for skin in prolonged contact
Figure 8. Maximum skin temperatures observed during each of the three performance benchmarks. In the category of “Cool”, the HP EliteBook 840 G7 comes out ahead. With the aluminum exterior chassis, the largest cooling fan in the test category, and some slightly more restrictive CPU Power Limits allows for the EliteBook to maintain skin temperatures below the other two competitor systems. Both the HP and Dell notebooks would fall into the “acceptably warm” skin temperature category. Conversely, the Lenovo ThinkPad X1 Carbon consistently had the highest skin temperatures observed in all three testing applications and might cross into the “annoying” category in some applications.
Skin Temperature Winner: HP EliteBook 840 G7
Skin Temperature Loser: Lenovo ThinkPad X1 Carbon Gen8
Sound Output The final performance metric utilized to evaluate the three notebooks is “Sound Level Output”. Sound Output is typically measured in decibels (dBA) or Sones with the primary difference being that the former follows a logarithmic scale and the latter follows a linear scale. Aside from the overall responsiveness (performance) of a computer, the sound it emits while you are sitting in front of it is often the next most noticeable characteristic. Acoustic engineers are constantly tweaking various sound-impacting features to not only reduce the sound level being emitted, but also maximize the quality of the sound being emitted which can heavily influence perceived “annoyance”. In the realm of 14-inch business notebooks, the only noticeable sound emission comes from the CPU fan.
There are a few basic principles about cooling fan sound output that are helpful to keep in mind:
1. As the heat produced by a CPU increases, fan speeds must increase to meet the
cooling demand.
2. A larger fan can move more airflow (better cooling) than a smaller fan at the
same speed
3. Sound level output from a fan will increase as the speed of the fan (RPM)
increases
4. A larger fan spinning at the same RPM as a smaller fan will emit more noise due
to the increased airflow (sound level increases with velocity).
In alignment with the previous KPIs, sound level was captured for each system while running the 3 office-oriented benchmarks: PCMark 10, SYSMark 25, and Cinebench.
Figure 9. Sound capture setup for Lenovo ThinkPad X1 Carbon.
Each system was placed into an isolated sound capture environment and pushed through the three benchmark programs. The results for the sound level testing are shown below in Figure 10.
Figure 10. Sound level output for each system running the three office benchmarks. HP’s larger cooling fan does permit the EliteBook 840 G7 to maintain substantially lower skin temperatures than either the Dell or the Lenovo notebook, however, it costs them in noise production as the EliteBook is the loudest system in two out of three benchmarks.
In addition to peak sound output, CPU fan speeds were logged during the PCMark10 benchmark using the HWinfo64 utility and are shown in Figure 11. While the HP EliteBook and Lenovo ThinkPad hit similar peak fan speed values, the HP system is louder at the same fan speed due to its larger fan diameter.
Figure 11. CPU fan speeds for the PCMark 10 benchmark. The semi erratic nature of the HP EliteBook’s CPU fan speed is a bit of a double-edged sword, from a thermal engineer’s perspective. It is clear that HP has elected to use a predictive fan control algorithm (very likely PID) which is continuously seeking to adjust fan speeds in response to instantaneous cooling demand changes from temperature sensors within the laptop. This is, actually, a far more advanced fan control technique than what is employed by either Dell or Lenovo which seem to use nothing more than fan speed lookup tables. The downside with HP’s fan control algorithm implementation is that it is poorly damped and fan speeds are largely unstable during mixed-use workloads. A fan speed that frequently ramps up and down during operation will be more noticeable than a steady fan speed at a similar peak RPM value.
Sound Output (Acoustic) Winner: Dell Latitude 7410 Sound Output (Acoustic) Loser: HP EliteBook 840 G7 Final Thoughts and Conclusions Comparing the Dell Latitude 7410, HP EliteBook 840, and Lenovo ThinkPad X1 Carbon reveals that all three products are highly capable business-oriented machines with something to offer a broad market of customers with a wide range of product experience expectations. Each vendor made notable tradeoffs to allow for their product to accel in other usage conditions.
For customers who are easily annoyed by warm touch temperatures of their notebook, the HP EliteBook 840 G7 boasts the lowest skin temperatures of all products tested throughout all workloads. The tradeoff that HP made, however, was lower CPU performance caps and elevated sound output making it a bit slower and a bit louder than other notebooks in the segment. The smaller (cheaper) heat pipe selected and lack of PL1 optimization seem to hamstring the EliteBook’s capacity to support opportunistic performance burst on applications like PCMark10 and short rendering jobs in Cinebench.
Customers who are frequently utilizing rendering applications such as Cinema 4D, Unity, 3Ds Max, or Maya may prefer the Lenovo ThinkPad X1 Carbon which has a slight performance edge in those workloads. The X1 Carbon achieves the most consistent performance for longer rendering jobs but does so at the expense of having the highest skin temperatures of all three systems in all workloads. With that said, most users who spend significant time in rendering applications will likely be much happier with a larger form factor notebook or desktop platform which can dissipate substantially more heat demanded by this type of software.
The Dell Latitude 7410 seems to have struck the perfect balance for most business use cases. A smaller heatsink allowed for Dell to cram over 28% more battery capacity than both HP and Lenovo systems. Masterful orchestration of Intel CPU PL1 limitations and fan speed control allows the 7410 to best the competition by an average of 12% in application performance of the most commonly used software in the office environment while emitting 14% less noise.
If you are a business customer looking for a notebook that will provide the best performance in the applications that you use most frequently, do so with moderate skin temperatures and low noise, all while providing the longest battery life, look no further than the Dell Latitude 7410.
Download a copy of this paper HERE
Comments