Most enterprise laptops reduce performance to conserve battery life when unplugged — a design decision that has been a long-standing point of frustration for transient workers. The practice of power throttling can also create headaches for IT teams tasked with improving the user experience. Performance challenges will only increase as organizations deploy more compute- and data-intensive AI applications and tools.
It’s easy to blame growing frustrations with power throttling on the smartphone, which has conditioned people to expect consistent performance regardless of whether the device is plugged in or roaming free on battery.
“People are living on their phones, and when they pivot to a PC and get a bad experience, they start to lose hope,” says Craig Tellalian, director, Field Applications Engineering at Qualcomm Technologies, Inc. “It’s like being forced to use technology that’s four or five years old.”
That’s a problem for IT teams dealing with an increasingly mobile workforce — and a mandate to deploy and scale AI across the enterprise. AI holds much promise to improve worker productivity, but not if the devices they’re using can’t keep up.
“A lot of organizations are wrestling with where to run workloads as they push genAI [generative artificial intelligence] out to more users,” says Tom Mainelli, group vice president for Device & Consumer Research at IDC. Concerns about data privacy and costs could drive many organizations to transition away from large language models (LLMs) and focus on small language models (SLMs) that can run locally on edge devices and PCs, he says. IDC predicts that by 2026, 90% of enterprise use cases for LLMs will be dedicated to training SLMs, because of cost, performance, and expanded deployment options.1
This shift will dramatically increase the pressure on IT teams to deploy PCs and laptops that can handle those complex workloads. That’s where the new generation of AI PCs comes in. The addition of a neural processing unit (NPU) dedicated to AI workloads frees up the CPU and graphics processing unit (GPU), improving system performance for unplugged devices.
With highly optimized CPU, GPU, and NPU engines working together, power is designated where and when it’s needed, improving performance not just for AI tasks but for all the workloads and applications running on the device, including videoconferencing and other virtual collaboration tools.
“The NPU runs at a significantly lower power level than your traditional CPU or GPU. And it’s been trained and optimized to handle certain types of tasks more efficiently,” says Tellalian. “You’re never going to run everything on the NPU — it’s meant to handle specific workloads. And that frees up the CPU or the GPU to handle other tasks more efficiently as well.”
Snapdragon X Series processors, which power new CoPilot+ PCs, don’t degrade performance for portability’s sake. Geekbench tests show that whereas traditional processors throttle down the performance of an unplugged device between 29% and 46%, a PC running the Snapdragon X Elite processor showed just a 1% performance drop when running on battery.2
With AI PCs, the elusive promise of laptop performance consistency — regardless of power source — is finally becoming a reality. That’s good news for end users and IT teams alike.
Click here to learn more.
Sponsored by Qualcomm Technologies, Inc.
1 https://my.idc.com/research/viewtoc.jsp?containerId=US51666724
2Performance is based on the Geekbench Single-Core and Geekbench Multi-Core test run in Windows 11 in October 2024. Snapdragon X Elite (X1E-80-100) was tested with a Dell XPS 13 (9345). The Intel Core Ultra 7 256V was tested with a Dell XPS 13 (9350). On-battery performance was measured in “Balanced” Power Mode in Windows and “Optimized” in Dell Power Manager for both devices. Power and performance comparison reflects results based on measurements and hardware instrumentation of given devices.