Chris Patrick, senior vice president and GM of mobile handsets at Qualcomm, said the company does not design mobile chipsets with benchmarking in mind. But we do measure,” Patrick told attendees at this week's Snapdragon Summit event. As for the just-announced Snapdragon 8 Elite system-on-chip, it's clear that Patrick likes the chip, which will be in many of the best Android phones in 2025.

Qualcomm has released some eye-popping benchmarks on its new silicon, with the Snapdragon 8 Elite posting general performance numbers that beat Intel's new Lunar Lake chip in Geekbench tests.

At the same time, Qualcomm is promising power savings of 44% on the Snapdragon 8 Elite's CPU and 40% on the GPU compared to the Snapdragon 8 Gen 3. And the chipset, as featured in several entries on our Best Smartphone Battery Life list, consumes very little energy.

So how did Qualcomm achieve both performance and power efficiency? I was at the Snapdragon Summit as a guest of Qualcomm and had a roundtable discussion with Patrick about the Snapdragon 8 Elite along with other journalists. The following is an edited version of my questions to Patrick, a 28-year veteran of Qualcomm, covering the reasons for Qualcomm's changes to the top-of-the-line silicon and the potential impact users may notice on camera performance and other aspects.

Tom's Guide: What are the benefits of switching to a proprietary custom CPU in Oryon?

Chris Patrick: Most of the technology we use is designed in-house. Ultimately, we are building a fairly complex architecture that has evolved over decades of how we handle low-power operation, how we ramp up ultra-fast to extreme peak performance, how we move data throughout the system, how we manage current spikes, and so on. So we felt that CPUs would be a very powerful part of our product portfolio.

And we had ambitions to do more than we could get from vendor roadmaps alone. We were excited about the opportunity, and I think we were able to do exactly what we wanted to do: to be a desktop-class CPU (at the Snapdragon 8 Elite launch), but with excellent power efficiency for mobile. Peak performance is very high, but can be brought down to the battery-saving levels needed for mobile.

TG: What steps do you need to take to balance power and efficiency? What are the tips?

Patrick: The right microarchitecture that can maintain efficiency in each area. It is very difficult. It's like trying to be an Olympian in two sports at the same time. It's very difficult to train in both at the same time. But again, with the right microarchitecture, you can do it. And the right [system-on-chip] design with the right hooks makes it possible. [If there is one Olympic sport we have trained in, it is this one.

TG: Regarding Adreno GPUs, what are the advantages of the slice architecture you are introducing?

Patrick: We have been working on a certain philosophy of GPU architecture design for many years. It is really ...... not monolithic and has had a very heavy share across GPU architectures. As we discussed (at the Snapdragon 8 Elite launch), the growing needs of smartphone products require more advanced nodes. We have found that a more modular architecture is needed and more dedicated resources are needed for the GPU trunk.

Certainly not multiple GPUs, but still, a more modular architecture. This is the concept of slicing, where you slice pieces of a GPU to make them somewhat independent.

As we moved into more advanced technology nodes and technologies, we realized that this architecture, this partitioning, is very well suited to scale up to the tremendous power that cell phones require while managing power efficiency at the bottom.

TG: With the hexagonal neural processor having more direct access to the image and image signal processor, what will end users notice in the images of Snapdragon 8 Elite-powered phones?

Patrick One very direct way is segmentation, inline segmentation. You take a picture and then the AI says, “This is actually a leaf, not a piece of someone's hair. This is actually a flower, so I want to process it this way.” That can always be done in post-processing. But in the meantime, you lose information that will help you make a better photo.

When you look at a beautiful scene outside with the naked eye, you don't realize it, but you are really processing it differently depending on the context.

The great thing is that if we can use AI resources live, not post-processing, to identify features of a scene and provide real-time feedback, we can get better results. [Autofocus, white balance, color correction - if you can do all of these things live, you get a better balance. Consumers will not say, “This definitely shows that direct linking is enabled.” They will not say that. But in the end, they will see a more favorable photo, one that better captures what their eyes would have captured.

TG: If you had to narrow it down to just one elevator pitch, or one feature that would tell someone that this is why you are excited about this chipset, what would it be?

Patrick: It's hard to over-index a CPU, but that's not the case here. This is really something special: a desktop-class CPU, but with mobile-level power efficiency.

In the pocket of your smartphone, you will likely find something comparable to the PCs you have been using. They enable entirely new use cases: productivity use cases, gaming use cases, AI use cases, and complex applications. There is no hesitation about any of them. Even if you use an incredibly complex web page and the one in the link doesn't work at all, it will all work on your phone without any issues, problems, or lag.