WHAT CAN AN AI PC DO NOW?
If you’re in the market for a new PC and are weighing up whether to invest in a so-called AI PC, you might reasonably wonder what advantages it will bring over an “ordinary” PC.
We’ll come to the benefits afforded by the piece of hardware that pretty much defines an AI PC – the neural processing unit or NPU – shortly, but there’s another distinguishing piece of hardware found on many AI PCs: the Copilot key.
Calling for a Copilot
The Copilot key was billed as the first major change to PC keyboard layouts in almost 30 years, replacing the menu key to the right-hand side of the spacebar. It’s basically a shortcut for opening the Copilot assistant in Windows 11, sparing you the enormous effort of using the current Windows+C shortcut. “We’re excited about having this key on our systems,” said Kevin Terwilliger, vice president and general manager of Dell’s Latitude and docking business on a recent AI webinar. Indeed, it was the first thing Dell’s guide showed us when we visited the company’s stand at MWC ( see our highlights on p14).
While enthusiasm levels among PC Pro’s professional readership might be a little more muted, it’s worth remembering that 99% of users aren’t familiar with any but the most basic keyboard shortcuts and that such initiatives – while appearing cosmetic and trivial – could make a difference.
“We really think of Copilot as the new search,” said Terwilliger. “And people are going to be interacting with Copilot throughout their day, engaging with it, asking questions, working with it to create content.
And so we think the fact that having that key on the keyboard to jumpstart access into Copilot will be very valuable for the end user.”
Microsoft is driving Copilot integration deeper into the operating system, too. At the time of writing, Copilot’s ability to control the PC is limited. It can help you flick into dark mode, take a screenshot and open File Explorer, but it’s superficial stuff.
However, Windows Insider releases launched at the time of writing show Microsoft continuing to build out Copilot’s system-level capabilities, allowing Copilot to show available Wi-Fi networks, clean storage, show which apps are set to run on startup and more.
Again, enthusiasts and IT pros will barely raise an eyebrow, not needing Copilot’s help to perform such tasks. But it could (and we stress “could”) make a big impact with mainstream users who aren’t as familiar with the workings of Windows. And if it spares IT departments from dealing with these more low-level queries, allowing more time to focus on the business-changing stuff, that can’t be a bad thing, surely?
What does the NPU do?
The NPU is the core component of an AI PC, the factor that really sets it apart from a PC with only CPU and GPU units to call upon. Over time, it’s likely all PCs will ship with an NPU ( see box, p30), but for now they’re the key differentiating factor.
But if you’re an early adopter, one who rushed out to buy a laptop or PC with an NPU-blessed Intel Core Ultra or AMD Ryzen chip inside, what benefits will you feel right now, today? Even the PC manufacturers quietly concede the advantages are slim pickings at the moment.
One that crops up often in PC manufacturer demos is the NPU’s ability to blur the background on video calls. Most users won’t care if it’s the CPU, GPU or NPU that performs something as mundane as background blurring, but it can make a difference to power consumption.
Dell ran a demo showing background blur being applied on a Zoom call on an ordinary PC, where CPU utilisation shot up to more than 8% when the background blur was applied. By applying the same effect locally on the NPU, using Windows Studio Effects, the CPU was barely troubled at 1% utilisation. “This actually translates into a 38% power improvement when you’re doing
People are going to be interacting with Copilot throughout their day, engaging with it, asking questions, working with it to create content
these Zoom calls,” claimed Kevin Terwilliger. “So collaboration is a great area where we’ll take advantage of the NPU to deliver energy efficiency and much longer battery life.”
That’s fine, but it’s hard to believe anyone’s going to buy a new laptop to save a slice of battery life when using background blur on battery power. Is anything more substantial on offer?
Dell points to security benefits, claiming the NPU is better placed to deal with threat detection. The firm ran a demo of an unnamed security package, but instead of running threat detection in the cloud, it used a local engine on the NPU. Dell claims the threat detection on the AI PC kicked in within 20 milliseconds, a 70% improvement in latency compared to the cloud. “Our goal is to not talk about the hype of AI, but show examples just like this to our customers so that they can understand the value of the NPU,” said Terwilliger.
You’re not racing off to the Dell website to order an AI PC yet? Maybe a more creative application will convince you, and here the benefits are arguably more substantial.
Instead of relying on expensive cloud services such as OpenAI’s DALL-E or Midjourney to generate images, you could install open-source models such as Stable Diffusion and run them locally instead.
You can, of course, do that already on a PC without a NPU. But Charlie Walker, Dell’s senior director and GM for Precision workstations, claims it’s much faster with the NPU involved. “If I tried to run Stable Diffusion on my CPU, previously it would have taken two to three minutes [to generate an image],” he said. “Now with the NPU, by offloading that from the CPU into that more efficient architecture... you can do that now in 30 to 50 seconds. So again, significant savings.”
The obvious downside? That only more technical users will feel comfortable downloading opensource projects such as Stable Diffusion, despite many in-depth guides online to step you through the process. For local AI to break through, it needs to be simple.