A studio's worth of compute power in an ultra-sleek notebook
Content creators and mobile productivity power users are spoiled in 2023. There, it’s been said, and we can’t take it back. That’s probably going too far, but the truth is that premium notebooks today pack an incredible amount of horsepower into ultra-thin, sleek machines that are as beautiful as they are portable.
It’s not hyperbole to state that premium notebooks today weighing in around 5lbs are literally able to outperform massive desktop workstations of just a few years ago at many of the same compute-intensive tasks such as rendering, CAD, and even the occasional gaming session.
In this issue of Notebook Thermodynamics, we are testing 4 mobile powerhouses: the Dell XPS 15 (9530), Dell XPS 17 (9730), Samsung Galaxy Book3 Ultra 16, and the ASUS Zenbook Pro 14 through the user-experience lenses of application performance, skin temperature, and sound output.
If you prefer to skip the fun and just get to the bottom line on these systems, feel free to jump ahead to our Results Discussion page!
Introduction and Background
Most computer users have probably noticed that computers have gotten faster and more compact over time. Similarly, most computer users can probably identify with the sensation of their notebook or cell phone feeling hot to the touch after hard usage. This heat you occasionally experience when touching your notebook is the primary reason that notebooks have, historically, trailed behind their desktop PC counterparts in application performance. The total amount of heat that a computer can dissipate (and thereby CPU performance that can be supported), generally increases with the size of the heatsink attached to the CPU (we will discuss later how thermal control algorithms are actually more important than size alone). Not surprisingly, desktop PCs are capable of supporting much larger heatsinks than what can be crammed into a sleek, portable, laptop form factor. This gives desktop PCs a heat transfer advantage over smaller notebooks.
In the previous installments of this series on Notebook Thermodynamics, we coined the facetious expression “Laptop Thermodynamics Cycle of Shame”. We recommend checking out that article here. While that phrasing was made in jest, the bare truth is that the very laws of nature work against the ambitions of notebook users and designers alike. Everyone wants their notebook to be faster and capable of running all of the same applications as their bulky desktop systems. Running a slim notebook with a heavy workload creates more heat. More heat leads to hot components and surface temperatures. Mitigating hot component and surface temperatures requires loud annoying fans.
CPUs (the brains of our computers) provide us with compute horsepower by converting electricity into heat in a 1 to 1 ratio. This means that if your notebook is consuming 100 Watts of electricity from your wall outlet or battery, it is creating 100 Watts (or Joules per second) of heat. By this logic, you can start to think of your notebook as being a glorified space heater where the more intense your software application, the more heat your notebook is going to produce.
You mentioned something about these new laptops catching up to the performance of desktop workstations?
Yes! While notebooks will always trail the performance of larger desktops due to the size limitations and heat problem we just learned about, recent advances in notebook heat removal technology have dramatically improved the amount of performance we can obtain from the premium systems we are looking at in this study.
The Dell XPS 17 9730 is a beautiful example of the most advanced notebook cooling technologies on the market today. In the figure below, we have attempted to call out some of the most cutting-edge cooling tech and explain what they mean to users.
By leveraging various combinations of these cooling technologies, all 4 of the notebooks included in today’s thermal review can outperform workstations built using AMD’s Threadripper 1950X workstation CPU – considered to be one of the most potent desktop workstation platforms just a few years ago – in CPU intensive tasks such as Cinebench photo rendering. To get that kind of CPU horsepower in an ultra-sleek notebook is absolutely amazing.
But it’s not all roses for these premium notebooks. There are still physics limitations that need to be addressed.
First, while these notebooks are capable of dissipating more heat than previous generations, they still get warm when running intense applications. Consequently, they are forced to rely on spinning up fans to cool them off. Boosting fan speeds results in improved system cooling and accelerated application performance, however, users trade this performance increase for heightened fan noise. Although reducing fan speeds is a viable choice, it might prompt concerns about heightened laptop heat, or alternatively, it could result in CPU performance throttling, thereby slowing down the system.
For this reason, we analyze the cooling solutions of notebooks using the following 3 Key Performance Indicators (KPI):
· Application Performance: how “fast” the system feels
· Skin Temperature: how hot the surface you touch is getting
· Sound Output: how loud is the system under various conditions (aka fan noise)
System designers do their best to try and predict which balance of these 3 competing factors is best for their intended customers, but it is a difficult job and a constantly moving target. Even a single user might find themselves ranking the importance of those 3 factors differently depending upon the setting they are operating in on any given day. Working in a library or conference room? You’ll probably value a quiet system. In a rush to get a big project done? You’ll value speed. Sitting on an airplane with your notebook in your lap? You’ll want to prioritize coolness.
What we like about the Dell notebooks is that they give users the opportunity to make meaningful choices about which of those competing factors is most important to them, individually, through User Selectable Thermal Tables (USTTs). This topic has been covered pretty exhaustively in the past but here’s a quick summary of what you need to know on Dell’s USTTs:
There are 4 selectable thermal modes:
· Optimized: The factory default which balances all 3 competing thermal performance influencers.
· Cool: Choose this mode if a warm palm rest, keyboard, or notebook on your lap really annoys you.
· Quiet: Prefer for your notebook to be seen and not heard? Quiet mode attempts to drop notebook fan noise.
· Ultra Performance: Ultra Performance mode is designed to relax some of the physics constraints around noise and skin temperature to provide maximum application performance.
Whenever we perform an engineering deep dive on notebook systems, we can typically pinpoint the KPI that was deemed most important during system design. Some systems will be faster, some quieter, and some cooler depending upon what that product development team decided was most important for their customers. What makes USTT’s so valuable for Dell XPS systems is that, in most cases, XPS notebooks can achieve superiority in any of the 3 KPI’s with just the click of a button.
Table 1 below includes the system specifications for each notebook included in this study. Special care was taken to configure these systems with as close to identical component selection as allowable by each of their manufacturers.
Table 1. System configuration details for the four premium notebooks included in this study
Application performance is where we get to test how fast each notebook system is for a wide spectrum of software types. We choose a variety of software applications because every software app that runs on your computer utilizes the underlying hardware in its own unique way. Some applications need lots of Random Access Memory (RAM) while others demand more CPU utilization. There are applications that demand the CPU’s full attention for prolonged periods of time and others that only require the CPU to do work sporadically for short bursts. Similarly, there are applications that benefit from having an onboard Graphics Processing Unit (GPU) and others that completely ignore it.
Whether we are talking about CPUs, GPUs, RAM, or storage, every critical component on a notebook produces heat as it is utilized by software applications. When the component gets too hot (or when the overall system gets too hot for user comfort), that component will be throttled back to a lower performance state, and/or the system fans will be boosted to help cool the component and the laptop. Though all 4 of these notebooks are using the exact same models for CPU and GPU, they each have unique cooling solution schemes that results in different behaviors depending upon which hardware component is stressed by the software application.
The bottom line is that the applications that YOU use most regularly might behave very differently on a notebook from the applications that I use daily. Our goal is to try to pick representative applications that align with the targeted audience for a given notebook product. For these systems, the targeted audience is focused on performance users who specialize in content creation or similar applications.
Application: Cinebench R23
Cinebench R23 has become the standard benchmark ‘flex’ for most CPU performance enthusiasts because it does a fantastic job of pushing CPUs to their limits. The benchmark operates using a single, high-resolution, photo that is fully rendered and timed for speed of completion. Cinebench can be operated with just a single core or using multiple simultaneous threads on multi-core CPUs. As most modern software can take advantage of multi-core CPUs, we chose to execute this benchmark in multi-thread mode. This is a great benchmark to focus on if you are a content creator who spends a significant amount of time performing rendering operations.
Cooling Challenge: Short-Duration, Maximum CPU heat
Winner for Photo Rendering: Dell XPS 17 9730
As previously mentioned, Cinebench R23 has, in many ways, become a critical landmark for computer performance comparison. Given its importance for computational performance bragging rights, you can easily find the Cinebench scores for computers throughout history.
Below is a snapshot of the Cinebench R23 scores for some of the most prolific desktop server and workstation CPUs from Intel and AMD just a few short years ago. that the incredible performance previously only attainable with a giant desktop workstation is actually being beaten by the ultra-sleek premium notebooks we can buy today!
Application: UL Procyon
UL’s Procyon benchmark has been gaining massive popularity within the content creation world for good reason: it tests hardware performance using the applications that content creators actually use like Adobe Premiere Pro and Adobe Lightroom. If your world consists of editing photos and videos using Adobe software, Procyon is the best way to compare how effective a given computer will be for your needs.
Cooling Challenge: Long-Duration, Maximum CPU heat
Winner Video Editing: Dell XPS 17 9730
Cooling Challenge: Long-Duration, sporadic mix of CPU and GPU utilization
Winner Photo Editing: Dell XPS 15 9530
Application: BAPCO SYSMark 30
SYSMark 30 is a fantastic benchmark for characterizing how a system will perform using common office productivity applications such as Microsoft’s Office suite (Word, Excel, Outlook, etc.) along with a handful of other Web applications and common tasks like Google Chrome browsing and video conferencing.
Cooling Challenge: Long-Duration, sporadic mix of CPU and GPU utilization
Winner Office Productivity: ASUS Zenbook 14 Pro
Application: 3DMark “Speed Way”
After a long day of CAD, CFD, or creative content production, some of us like to unwind with some video gaming! UL’s 3DMark has been the ubiquitous name in gaming benchmark software since the earliest days of the discrete graphics accelerator (GPU). While it is important to call out that these systems are not officially considered “gaming notebooks,” they pack plenty of horsepower to enable gaming on the go.
Cooling Challenge: Long-Duration, heavy GPU utilization with some CPU mixed in
Winner Gaming: Dell XPS 17 9730
Performance Results Discussion
Let’s unpack what these performance results indicate about the thermal design of the premium notebooks included in this study.
First, Cinebench, Procyon Video Editing, and 3DMark tend to be the most cooling-intensive applications overall. These 3 applications will push the limits of the cooling solution on any notebook. In most cases, the notebook that designs the best (often biggest) CPU and GPU heatsink, will win those 3 benchmarks but not always. As you will soon see, thermal control algorithms that set the runtime policies for maximum component power and fan speeds can be just as critical to benchmark performance as the heatsink size itself.
In the figure below, we compare the size of the combined CPU and GPU heatsinks for each of the 4 notebooks in this study. If we ignore the Samsung Galaxy Book3 for a moment, you will notice that the success in Cinebench, Procyon Video Editing, and 3DMark very much follow the trend for heatsink size with the Dell XPS 17 9730 taking home clear victories in all 3 cooling-intensive benchmarks owed, in large part, to its superior large heatsink.
As you can see in the figure above, the XPS 17 9730 shows off the largest heatsink in this test suite while the XPS 15 9530 has the smallest, giving up roughly 70% heatsink volume to the next closest competitor, the ASUS ZenBook 14 Pro. By trading off heatsink volume the XPS 15 9530 manages to squeeze in about 13% more battery capacity compared to the ZenBook and Galax Book3 Ultra. The XPS 15 9530 winds up being within about 5% of competitive systems for performance benchmarks so users must decide if the extra battery capacity justifies the relatively modest performance variation.
So why does the Samsung Galaxy Book3 Ultra seem to underperform most benchmarks given its relatively large heatsink?
Earlier in the article we discussed that, due to their size limitations, notebook CPUs and GPUs will eventually experience a performance limitation based upon their ability to be cooled. We previously noted that the size of a laptop’s heatsink is among the strongest indicators for the total cooling power of a notebook. But having a large, space-consuming, heatsink is useless without a well-developed thermal control algorithm to maximize performance.
This brings us to a brief discussion on transient or “temporary” power (or heat) vs sustained power (or heat). Remember: performance = power = heat! Performance is the thing you want, power is what it takes to get it, and heat is the resulting waste product that causes problems. A good way to think about transient heating (or power) is when you grab a hot plate out of a microwave with your bare hands. Depending upon how hot the plate is, you might be able to hold that plate of food for a few minutes or just a few seconds before you hurt yourself. This is called thermal capacitance. You can think of the skin on your hand as a battery and as you hold the hot plate, your thermal battery is charging up. When it hits full charge, you will start to feel pain in your hand from the temperature of the plate. This is true in laptops too.
Often, CPU and GPUs can run at high power (heat) levels for a very short time bursts (seconds) before they hit temperatures that will harm either the user or the components within the laptop. .CPU and notebook designers take advantage of this transient (temporary) power capacity and turn it into short-term performance enhancement. Intel calls this temporary boost in performance “TurboMode” or “dynamic power.” One way to think about dynamic power is like the top-speed in your car. The maximum speed limit for most highways in the US is 80 MPH or less. There might be scenarios where your car needs to exceed that speed limit for short periods of time. When that time comes, you will be glad that you have the extra performance on demand provided by your gas pedal.
In a notebook, the “governor” that limits your “top-speed” on Intel-based platforms is called the “PL1 Limit” which sets the maximum speed your notebook CPU can operate at for seconds at a time before being throttled back to its more comfortable operating cooling condition or “speed limit” before the cops (or physics) catches up.
Remember: the heatsink size sets the “speed limit of the road”; the PL1 Limit sets the maximum speed to which the car can accelerate.
The figure below showcases the PL1 Limits observed for all 4 notebooks. The PL1 Limits are configured by each notebook manufacturer according to what they believe their product is capable of and how aggressive they want to be in pursuit of application performance objectives. What we found is that, while the Samsung Galaxy Book3 Ultra has a respectable 2nd-best heatsink size, it limits performance with an ultra-low 45-Watt governor on maximum CPU power. Unfortunately for Samsung, this ultra conservative PL1 limit causes them to waste an otherwise impressive cooling solution.
In stark contrast to the Samsung Galaxy Book3 Ultra, the ASUS Zenbook 14 Pro has its PL1 TurboMode Limit set to a seatbelt clutching 115 Watts! This provides ASUS with a distinct advantage in very “bursty” application workloads such as opening and closing Microsoft Office applications like Word, Excel, and Outlook. With these types of applications, most of the CPU work is done just by opening the program. SYSMark 30 tests this functionality by performing a series of open and close operations with the most commonly used Office applications and stop-watch timing how long it takes. The ASUS Zenbook 14 Pro posts great SYSMark 30 results because they set their PL1 Limit so aggressively high.
The final oddball test to discuss is the Procyon Photo Editing benchmark where the Dell XPS 15 9530 appears to defy its designers with an impressively high score that exceeds even its big brother, the XPS 17 9730 and the ASUS Zenbook 14 Pro, despite both latter systems having much larger heatsinks. The reasoning for this actually goes outside of the conventional realm of thermal engineering into the world of Driver development. Component “drivers”, in the context of computers, are pieces of software, written by the computer manufacturers, that tell your operating system (Windows 11 in this case) how to communicate with your hardware (such as your Graphics Processing Unit or GPU). Drivers help inform the operating system what functionalities are best provided by each onboard component.
In the case of the premium notebooks included in this study, each system has a “discrete” Graphics Processing Unit (GPU) provided by Nvidia and an “integrated” GPU that is built into the Intel CPU. In a perfect world, the integrated GPU handles all light-duty graphics work like your desktop background, Chrome browser windows, and Office applications to save power while graphic intensive applications like gaming are pushed to the powerful discrete GPU from Nvidia. Fundamentally, the integrated GPU consumes less power and is better for battery life and overall energy efficiency while the discrete GPU is far more powerful and provides significantly more performance.
Where the “driver” comes into play is making the decision between which applications should be given to the energy efficient integrated GPU and which applications deserve to be passed off to the powerful discrete GPU. The Procyon Photo Editing benchmark utilizes software that sits somewhere between the demand of an integrated GPU and a discrete GPU. If we look at the GPU Power (where GPU Power, in this case, references the discrete Nvidia GPU) comparison between the Dell XPS 15 9530 and the ASUS Zenbook 14 Pro in the figure below, we can pretty quickly see that the Dell XPS 15 graphics driver seems to preferentially hand more workload to the discrete GPU than the ASUS Zenbook 14 Pro driver. In the dynamic workload of Procyon Photo Editing, the XPS 15 seems to send almost 2x the “work” to the powerful Nvidia GPU which gives the XPS 15 9530 a huge advantage in photo editing software tested by that benchmark suite.
Skin Temperature is an attribute of notebook cooling that most users don’t think about until it becomes a problem. If you are always at a desk using a docking station with an external keyboard and mouse, you probably never think about the skin temperature of that notebook sitting beside you. It’s only when sitting on a crowded airplane that you might begin to take notice of your laptop’s skin temperature. There are two important skin temperature questions to ask about our notebooks:
1. Where are the hot spots located?
2. What can I do about them?
Where are the hot spots located?
Most users won’t notice or care if regions of their laptop that they never touch get warm. A good example of this is the thin strip of material between the keyboard and the display, sometimes called the “strip cover.” There isn’t really a good reason to touch that part of your notebook for a long period of time. Palm rests, on the other hand, are constantly in contact with your skin during laptop usage. Significant heat through the palm rest region of a notebook will be noticed and most people won’t like it. Understanding this logic, most notebook designers began placing the batteries (which rarely get warm under usage) beneath the palm rests and the “hot stuff” like CPUs up near the display hinge and C-strip. The bottom side of your laptop is another tricky spot for skin temperature because it can vary drastically.
At STL, we use IR cameras to first map the hottest locations of each notebook while operating under an extreme thermal stress application and then we physically instrument the notebooks with Type-T thermocouples to provide maximum accuracy as we log temperatures during benchmarks.
The figures below are IR images captured of each system illustrating the location for the hottest points on each respective notebook. Both Dell XPS systems have hotspots on the top-surface at the “C-strip” location that was previously discussed. The ASUS Zenbook and the Samsung Galaxy Book3 both have their hottest surfaces on the bottom (lap-side) of the notebook.
What can I do about the hot spots?
As mentioned, with notebooks, warm skin temperatures really only matter when someone feels them. That isn’t all the time, however, because you don’t always have that notebook sitting in your lap or cradled in your hands. When skin temperature does matter, Dell’s User Selectable Thermal Tables (USTT) on both XPS 15 and 17 allow you to quickly transition into “Cool Mode” which will rapidly bring down surface temperatures on your notebook to more comfortable levels.
For this study, maximum skin temperature was measured using thermocouples during a 10-minute stress test of the Cinebench R23 CPU workload. This combination of applications creates a worst case cooling condition, pushing internal hardware to its thermal limits.
To put skin temperature in context, the human body temperature is 37°C. Notebook skin temperatures in the 30-40°C range are considered ‘cool’. Skin temperatures in the 40-50°C range are considered ‘noticeably warm’ but not hazardous. Notebook skin temperatures above 50°C start to get uncomfortable for humans. All of the systems tested operate in the upper-end of ‘noticeably warm’ during extreme stress testing, however, only the Dell XPS 17 gives users the ability to drop skin temperature deep into the ‘cool’ range at 37.0°C when switching the USTT into “Cool Mode” making it the coolest system in the pack, followed closely by the XPS 15 in Cool Mode.
Winner for Skin Temperature: Dell XPS 17 9730 in Cool Mode
Notebook sound output, or fan noise, is another priority for many users. Unlike application performance, which users tend to want all the time, your level of annoyance from noisy fans can vary depending upon how quiet your surroundings are, whether or not you are in a meeting, if you have headphones on, etc. One fundamental truth about physics needs to be remembered: increasing processing power (speed) requires faster fan speeds to cool them and faster fan speeds means more fan noise.
In the same way that we tested skin temperatures, we measured sound output during a 10-minute dwell of Cinebench R23. This helps us understand what fan noises look like under the worst possible environment.
Looking at the data below from our acoustic study, we are, once again, impressed by how Dell’s USTT’s work to enable tailored performance depending upon the user’s individual needs. This time, it’s the XPS 15 that really pulls away from the pack dropping down to just 34.0 dBa in “Quiet Mode” even while running the extreme stress applications of this test.
Winner for Sound Output: Dell XPS 15 9530 in Quiet Mode
To recap this study, we took a thermal engineer’s perspective to compare the following 4 premium notebook systems:
· Dell XPS 15 9530
· Dell XPS 17 9730
· ASUS ZenBook 14 Pro
· Samsung Galaxy Book3 Ultra
Each system was configured with identical (or as nearly identical) hardware as possible and quantitatively analyzed for their performance in 3 Key Performance Indicator (KPI) categories:
· Application Performance
· Skin Temperature
· Acoustic Output
The Dell XPS 17 9730 provides the best performance in the widest range of applications. This is due to a larger heatsink and superior cooling solution that allows the XPS 17 to sustain higher heat loads for longer periods of time. If your intention is to run high-intensity applications like image rendering or video editing workloads for long periods of time throughout the day, XPS 17 9730 is the best option for heavy workloads.
The smaller and sleeker Dell XPS 15 9530 provided the best performance in Photo Editing application benchmarks due to a superior GPU driver offloading significantly more work to the more powerful discrete Nvidia GPU than the other systems in this study. It should be noted that component drivers are continuously updated by notebook manufacturers and these updates can have a significant impact on application performance. At STL, we take every test system out of the box and update them to the very latest driver versions from their manufacturers before starting any tests. The XPS 15 9530 has 13% larger battery capacity while managing to stay within about 5% of its competitors in most performance benchmarks and displaying impressive acoustic results in both Quiet Mode and Optimized Mode.
For more traditional Office applications, the ASUS ZenBook 14 Pro is the winner due to very aggressive PL1 TurboMode Limits employed by the ASUS BIOS team. In this case, the aggressiveness paid off allowing that system to win the SYSMark 30 office productivity benchmark.
As for the subjective KPI’s (being skin temperature and sound power), Dell’s USTT prove thatboth the XPS 15 9530 and the XPS 17 9730 can feature cool skin temperatures or quiet fan noise by simply switching into the respective USTT mode through the Dell Power Manager utility.
The Samsung Galaxy Book3 Ultra, while being middle of the pack for skin temperature and sound power, was consistently the lowest performer in most of the application performance benchmarks and failed to capture victory in any category. This was very disappointing for the Samsung system as it ships with the 2nd largest heatsink of the 4 test systems but was crippled by highly conservative CPU power limiting.
For the next generation, we would love for Dell to take some strategy notes from the ASUS ZenBook 14 Pro and be a bit more aggressive on their PL1 limits to match the “brain to the brawn”, so to speak, but amongst the systems included in this study, the Dell XPS 17 9730 is the clear winner for ultra-sleek, premium notebook supremacy.