As AI becomes more commonplace in business software and workflows, investing in AI PCs for your team becomes more of an imperative. That’s because AI PCs contain neural processing units (NPU) as well as the traditional CPU and GPU architecture. NPUs accelerate AI tasks and free up the CPU and GPU so they can focus on what they do best.

We evaluated CPU and NPU performance on three AI PCs:

• HP EliteBook 6 G1a powered by an AMD Ryzen^™ AI 7 PRO 350 processor
• Dell^™ Pro 14 Plus powered by an Intel^® Core^™ Ultra 7 268V with Intel vPro^® processor
• Lenovo^® ThinkPad^® T14 Gen 6 powered by an Intel^® Core^™ Ultra 7 268V with Intel vPro^® processor

We found that the AMD Ryzen^™ AI 7 PRO 350 processor-powered HP EliteBook 6 G1a achieved higher CPU and NPU performance than either Intel Core Ultra processor-based AI PC. Better CPU and NPU performance translate to a more responsive device when working on resource-intensive tasks or incorporating on-device AI features into daily routines.

Our testing approach

We equipped each PC with Windows 11 Pro, 32 GB of memory, and 1 TB of SSD storage:^*

*The results we report reflect the specific configurations we tested. Any difference in the configurations—as well as screen brightness, network traffic, and software additions—can affect these results. For a deeper dive into our testing parameters and procedures, see the science behind the report.

To help you better understand the CPU and NPU performance capabilities of each AI PC, we ran these industry-leading benchmarks:

• Cinebench 2024
• Geekbench 6
• Geekbench AI
• LM Studio
• Procyon^® Computer Vision Benchmark

We also tested multitasking capabilities by running the Procyon Office Productivity Benchmark in the foreground and conducting a Microsoft Teams video-conferencing meeting in the background. To see how the NPUs affected performance, we ran this test twice, once with Windows Studio Effects on during the video-conferencing meeting and again with it off.

Finally, we collected performance scores and measured temperature (or heat) and noise output while the AI PCs ran a sustained Cinebench 2024 workload.

Note: The graphs in this report use different scales to keep a consistent size. Please be mindful of each graph’s data range as you compare.

Powerful CPU and AI Acceleration for Modern Workflows

To evaluate processor performance for resource-intensive tasks, we used Cinebench 2024. Cinebench 2024 renders a 3D scene with CPU-intensive Redshift for Cinema 4D software.¹ Higher CPU multi-core performance scores indicate that investing in AMD Ryzen^™ 7 350 processor-powered HP EliteBook 6 G1a Notebook AI PCs would help users finish resource-intensive tasks in less time than the Intel Core Ultra 7 268V with Intel vPro processor-based Dell and Lenovo AI PCs we tested.

To stress the processors from another angle, we ran the Geekbench 6 benchmark. Geekbench 6 executes multiple CPU-intensive tasks simultaneously.²

To see how each AI PC handled machine learning (ML) workloads, we ran Geekbench AI, which uses real-world ML apps to provide a multidimensional picture of on-device AI performance.⁸ For this test, we used the Open Neural Network Exchange (ONNX) open-source AI framework and DirectML AI backend for ML on Windows.

The precision level scores below reflect different AI model requirements: Full Precision (FP32) is the most accurate and the most resource-intensive, Half Precision (FP16) is less accurate but more efficient, and Quantized (INT8) is the most resource-efficient and least accurate of all.⁹ These machine learning results indicate the AMD Ryzen^™ AI 7 PRO processor-powered HP EliteBook 6 G1a offers high CPU performance for AI apps that don’t yet utilize the NPU.

To measure CPU performance for on-device AI Chat models, we ran LM Studio, which uses local large language models (LLMs) to capture token metrics.¹⁰ According to Microsoft, LLM tokens are “words, character sets, or combinations of words and punctuation.”¹¹ In these tests, we ran the Meta-Llama-3.1-8B-Instruct-Q4_K_M. This LLM model predicts what the person typing the query is going to type next based on what they’ve already input.

Finally, to measure NPU performance for AI inference engine models, we used the UL Procyon AI Computer Vision Benchmark.¹² In our integer-optimized (INT8) testing, we used the API optimized for each AI PC’s NPU: the AMD Ryzen^™ AI API on the HP EliteBook 6 G1a and the Intel OpenVINO^™ inference API for the Dell and Lenovo AI PCs. We tested the following AI inference engines:

• MobileNetV3, ResNet-50, and Inception-v4: Convolutional neural networks (CNNs) widely used for image recognition, object detection, and image classification tasks. Essential for research institutions, tech companies, and individuals.^13,14,15
• YOLOv3: A deep neural network (DNN) that distinguish between different objects and features within images and videos. Used by healthcare providers, manufacturers, and video surveillance companies.¹⁶

Multitasking performance

To see what users juggling many tasks or screens at once could expect from each AI PC, we ran a productivity benchmark that stresses the CPU in the foreground and conducted a 10-person video-conferencing meeting with and without Windows Studio Effect turned on in the background. Procyon^® Office Productivity Benchmark, which we ran in the foreground, is designed to mimic a typical day at work, even leaving Microsoft 365 apps “running in the background as the focus moves from one task to another.”¹⁷ For multitasking with NPU testing, we ran Microsoft Teams and Windows Studio Effects in the background. For multitasking without NPU testing, we ran only Microsoft Teams in the background.

Consistent Power for Demanding Workloads

AI PCs and the high-performance processors that power them can produce heat and noise during sustained workloads. Before we started testing, we turned on HP Smart Sense, which, according to HP, automatically balances performance, power consumption, fan noise, and heat output for a quieter and more comfortable user experience.¹⁸ We also turned on the built-in intelligent features for the other AI PCs: Dell Optimizer for the Dell Pro 14 Plus and Lenovo Intelligent Cooling for the ThinkPad T14 Gen 6. Our engineer then ran all three AI PCs full tilt for 30 minutes under a Cinebench 2024 workload. The performance scores and heat and noise output show you how each AI PC handled the load with built-in AI system optimization technology turned on.

While all three AI PCs put out similar heat and were quieter than a soft whisper (30 dBA),¹⁹ the AMD Ryzen^™ AI 7 PRO 350 processor-powered HP EliteBook 6 G1a beat out its Intel Core Ultra processor-powered competitors in the sustained performance arena.

Conclusion

In our hands-on tests, an HP EliteBook 6 G1a Notebook AI PC powered by an AMD Ryzen^™ AI 7 PRO 350 processor delivered higher CPU and NPU performance compared to two Intel Core Ultra 7 268V processor-based AI PCs: a Dell Pro 14 Plus and a Lenovo ThinkPad T14 Gen 6. Better CPU and NPU performance indicates a more responsive AI PC for employees tackling resource-intensive tasks or incorporating on-device AI features into their daily routines.