Since the release of the Snapdragon X Elite platform at the end of last year, rumors about its performance have never ceased. Fortunately, we can now personally experience just how powerful and efficient this platform truly is.
A few weeks back, I made a trip to Qualcomm's headquarters in San Diego to get a firsthand experience with the Snapdragon X Elite platform. How does the Snapdragon X Elite stack up against its competitors?
Benchmark Tests During the demo, we were able to run some predefined benchmark tests using reference hardware systems such as 3D Mark and JetStream. The table below outlines partial results from running these tests on the Snapdragon X Elite 23w (system wattage, not package wattage) versus the Intel Core i7 Ultra 7 155h:
| Test | Snapdragon X Elite 23w | Intel Core i7 Ultra 7 155h |
|---|---|---|
| 3D Mark | Result | Result |
| JetStream | Result | Result |
[Note: Results to be filled in with actual benchmark scores.]
From the data above, it's evident that the Snapdragon X Elite performs admirably, outperforming the Intel Core Ultra 7 155h in every benchmark test, while consuming only a fraction of the power. For me, that's the crucial point - the total system power consumption of the Snapdragon X Elite is approximately 23 watts, whereas the CPU power consumption alone for the Intel system exceeds 100 watts.
Snapdragon X Elite Benchmark Tests
Power Consumption and Performance
Qualcomm's approach to providing power consumption information for its platforms is quite intriguing. Unlike traditional chip manufacturers, they don't measure CPU power consumption in terms of TDP (Thermal Design Power)—TDP being a figure that effectively measures the expected heat output of a CPU under load. Though not based on "power consumption" per se, reviewers often use this figure as a reference point for the chip's raw power usage. Instead, Qualcomm employs system-wide power consumption, likely continuing a method they used when measuring power consumption for their third-generation Snapdragon 8 series mobile platform chips. This means that in the example mentioned above, the 23-watt figure refers to the entire system's power consumption, not just the CPU's, indicating that the CPU's power consumption would be lower—something I'm eager to test once I get my hands on the actual hardware products.
Unfortunately, we couldn't measure the temperature or other metrics of these systems, but as someone who uses a relatively recent 13th generation Intel Core i7 Ultra laptop where idle temperatures often reach 60 degrees Celsius or higher, having a product with only a fraction of the power consumption for potentially higher efficiency is very appealing.
Gaming Can these computers run games? The answer is, of course, yes! Although these products aren't marketed as "gaming laptops," it doesn't mean we can't run lightweight games smoothly anytime, anywhere.
We tried out some game demos, including "Baldur's Gate 3," "Control," and "Red Alert 2," and as seen in the video below, these games can run impressively well. I won't get into the debate of 30fps versus 60fps, but in my opinion, if you want to squeeze in a few rounds of lightweight gaming during breaks between meetings, you can get a decent gaming experience.
"Baldur's Gate 3"
I'm not sure what graphics settings were chosen at the time, but the live demo ran at 1080p resolution with around 30fps, without any stuttering.
It's worth noting that these systems aren't positioned as gaming systems, but rather as a testament to the collaboration between Qualcomm and Microsoft to enable Windows 11 ARM (and Snapdragon X Elite GPU) to excel at running modern games.
On the terminal-side AI front, one of the most intriguing aspects of the Snapdragon X Elite platform is its terminal-side AI processing. While both Intel and AMD have released products with dedicated NPUs (Neural Processing Units), the Snapdragon X Elite is the most powerful, at least on paper. For comparison's sake, it should be noted that Intel combines its NPU with its CPU/GPU for terminal-side AI processing to enhance its processing capabilities, which may result in improved battery life but lower performance. In contrast, the dedicated NPU of the Snapdragon X Elite has enough raw processing power to handle all AI tasks without the need for CPU or GPU (though they can run on GPU/CPU if necessary, but that's not the focus).
Currently, AI performance is measured in TOPS (Trillions of Operations Per Second), representing the speed at which AI models are processed at the terminal:
It's evident that compared to the current offerings from Intel and AMD, the Snapdragon X Elite boasts superior AI performance.
So, what does this mean for real-world performance and applications? It means users can process AI models incredibly swiftly. How fast, you ask? Well, here's a video clip for you. I had to shoot it multiple times because it was processing so rapidly. What you're seeing here is Stable Diffusion running on the terminal, generating images based on input text prompts:
With just a blink, you might have missed the whole processing affair. What stood out particularly in this presentation is that the entire Stable Diffusion model is handled on the terminal side, running solely on the Hexagon NPU. While achieving similar functionality on AMD or Intel systems using a combination of NPU/CPU/GPU is possible, it would be less efficient and consume more power. NPUs are designed specifically for AI processing tasks, hence employing a dedicated NPU makes more sense, especially in laptops where battery life is precious.
The Future of Computing
So, after watching all these demos and benchmarks, you might wonder, when can we expect commercial models featuring this chip? The answer: pretty soon. It's expected that in the summer of 2024, OEMs like Dell, Acer, Lenovo, along with Microsoft's Surface series, will unveil a rich array of new machines.
Once we get our hands on these commercial products, we'll definitely revisit the performance of this platform. Meanwhile, we're also eagerly anticipating products featuring the Snapdragon X Elite, which could truly revolutionize the future of computing.
