(Credit: Charles Jefferies)
LAS VEGAS—Dell’s new Pro Max line defines cutting-edge mobile performance, and no model embodies that more than the gargantuan Pro Max 18 Plus. I was at the Dell Technologies World conference recently to see how that laptop works with Qualcomm’s AI-100 inference card, a new dedicated AI coprocessor with 64GB of additional memory installed. That's a different AI-crunching implementation than we've seen to date, with the laptop relying on discrete silicon instead of a GPU or a neural processor on the CPU die for machine learning tasks.
This 18-inch laptop is a powerhouse mobile workstation engineered around Intel’s “Arrow Lake” Core HX-class processors, scaling up to the Core Ultra 9 285HX with 24 cores and 5.5GHz of Turbo Boost capacity. The laptop models I saw ran on 96GB of RAM and multiple storage drives, making them a compelling alternative to a workstation tower. Yet, the question remains: Can it replace or complement a server?
(Credit: Charles Jefferies)Dell and Qualcomm: Can They Bridge the Gap Between Desktop and Mobile Workstations?
The answer: It sure can! Breaking new ground in the laptop space, Qualcomm’s AI-100 inference card features two AI-100 neural processing units (NPUs). This cutting-edge configuration delivers 32 AI processing cores and 64GB of LPDDR4x dedicated memory. That last specification is essential for AI engineers, as it enables local execution of large language models (LLMs). Local processing has many performance benefits while ensuring data remains private.
(Credit: Charles Jefferies)Dell demonstrated an AI-powered coding showcase on Ubuntu Linux. In the demo, the Pro Max 18 Plus dynamically generated an animation using Python in response to my straightforward user prompt.
I instructed it to create a heptagon with 10 red balls bouncing inside. The entire process of generating the code took just a few minutes, which I watched materialize in real-time.
(Credit: Charles Jefferies)I also witnessed the Pro Max 18 Plus loading a Llama 3 model with an impressive 70 billion parameters, demanding 45GB of RAM. Remarkably, it utilized 49GB of the NPU’s memory; to put this into perspective, no mobile GPU on the market provides that amount of memory, making this a task uniquely suited for Dell’s powerhouse laptop and the AI 100 inference card.
(Credit: Charles Jefferies)When I entered natural language queries, I saw a spike in NPU usage, while the CPU remained virtually untouched. The laptop's core CPU resources were free for other tasks, since the dedicated NPU handled all the AI processing.
The Qualcomm AI inference card also provides impressive power efficiency, occupying a 75-watt power profile, far less power than a typical high-end dedicated GPU.
(Credit: Charles Jefferies)Those who prioritize GPUs can also order the Pro Max 18 Plus with an Nvidia “Blackwell” RTX Pro 5000 professional GPU, which features 24GB of dedicated memory. Note that the laptop can have either the Qualcomm inference card or the Blackwell GPU, but not both. While the GPU model will be more versatile, the 40GB memory advantage the Qualcomm card holds will likely give it the edge in AI crunching and dealing with large models.
A 280-watt USB-C power adapter feeds the Pro Max 18 Plus juice in either configuration. Dell has also implemented a three-fan cooling solution that promotes improved airflow versus the two-fan configurations of previous Dell mobile workstations. The workstation’s other highlights include two Thunderbolt 5 ports—driving 80Gbps of throughput each—on its left edge.
(Credit: Charles Jefferies)The Takeaway: High-Level AI Development Just Went Mobile
The Dell Pro Max 18 Plus holds immense potential, delivering enough performance to run LLMs locally from different places in the office or even remote locations where a traditional desktop or server is impractical. And the device is on the horizon: Models with RTX Pro graphics cards will arrive this summer, while configurations featuring the Qualcomm inference card should land by the end of 2025.