Snapdragon X Elite Benchmarks: Geekbench, Cinebench, 3DMark & More

Jul. 24, 2024

Qualcomm’s Snapdragon X Elite has been released and laptop makers have finally launched severalCopilot+ PCsin the market. We recently got our hands on the Asus Vivobook S15, which features the new Snapdragon X Elite chipset. To test out its performance, power consumption, clock speeds, and more, I have extensively benchmarked theSnapdragon X Elitechipset (X1E-78-100) using a variety of tools. On that note, let’s begin.

Testing Methodology

We ran the Geekbench 6 CPU test on the Snapdragon X Elite (X1E-78-100), both on power and battery. And there wasalmost no difference in CPU performance, while plugged in or on battery. Keep in mind that all of the tests were done on Best Performance mode and Full-speed mode turned on in the MyASUS app.

If we talk about Snapdragon X Elite’s thermals, well, the CPU temperature at idle was 41 degrees Celsius before the Geekbench test, and during the test, it peaked at 75 degrees Celsius. The surface temperature around the keyboard was 29.7 degrees Celsius at idle, rising to 35.5 degrees Celsius during the intense workload.During Geekbench 6 CPU TestSnapdragon X Elite(Plugged-in)Idle CPU Temperature41 degrees CelsiusIdle Surface Temperature29.7 degrees CelsiusPeak CPU Temperature75 degrees CelsiusPeak Surface Temperature35.5 degrees CelsiusPeak Power Consumption47.6 WSustained Clock SpeedAround 3400 MHz (3.4GHz)

The Snapdragon X Elite’s peak power draw during the Geekbench CPU test was47.6 watts. As for the peak frequency, the Snapdragon X Elite sustained clock speeds of3400 MHz(3.4 GHz) for an extended period. It means that the Oryon cores didn’t throttle down the frequency and sustained the CPU performance for much longer.

While the CPU power consumption is notably high, especially compared to Apple’s M-series chips, the credit goes to Asus for keeping thermals in check. The company integrates its IceCool thermal technology featuring two fans and two heat pipes for better heat dissipation.

Overall, the CPU performance of the Snapdragon X Elite (X1E-78-100) is top-notch, both on battery and power. However, to squeeze the best performance out of the chipset, you will have to change the performance profile both in Windows Settings and MyASUS app. Note that this will result in higher battery drain and louder fan noise, but the fan noise is still much quieter than any x86 processor.

The frequency ranged between 2.5GHz to 3.4GHz, and occasionally, it even crossed the rated peak frequency of 3.4GHz, operating at 3.6GHz. While we didn’t notice performance degradation in full power mode, it’s worth noting that Snapdragon X Elite’s power consumption under maximum load is significantly higher than expected.

As this is the first PC chipset from Qualcomm featuring the new Oryon cores, we anticipate the next iteration of Snapdragon X Elite prioritizes efficiency, aiming to match the efficiency figures of Apple M-series chipsets.

To test the integrated Adreno X1 GPU on the X Elite, we ran the Geekbench GPU test, based on OpenCL and Vulkan APIs. On the OpenCL API, the Adreno X1 GPU achieved adecent score of 20,543 pointswhile plugged in and 20,417 points on battery. On Vulkan API, the GPU performed notably better. It scored 23,635 points on power and 24,227 on battery.Geekbench 6 GPUSnapdragon X Elite(Plugged-in)Snapdragon X Elite(on Battery)OpenCL score20,54320,417Vulkan score23,63524,227

The Adreno X1 GPU on this X Elite SKU (X1E-78-100) runs at 1.25GHz. Admittedly, the GPU on the Snapdragon X Elite is underwhelming at best. Evenan 8-core Apple M2 GPU scores north of 24,000points in the Geekbench OpenCL test.

That’s not to say that the Adreno X1 GPU is not usable or performant enough for light gaming, but compared to its competitors, there is definitely a performance gap. There is also much work to be done on the software and driver side to improve the gaming performance.

The Adreno X1 GPU on the Snapdragon X Elite does support Microsoft’s DirectX 12 API which means you can play games like Cyberpunk 2077 on low settings. So we ran the 3DMark Wild Life Extreme test on the DX12 API and it did perform well. It scored 37.82 FPS and 6,316 points while plugged in.

To test the AI capability of the Snapdragon X Elite chipset, we ran the Geekbench ML test. While using the ONNX DirectML API (using the GPU), it scored 2,350 points, and with the ONNX CPU API, it got the inference score of 2,885 points — all while plugged in. The NPU support for X Elite has not been added to Geekbench ML yet.

If we compare the CPU score, Apple’s M3 chipset scores above 4,000 in the same test, with its own Core ML CPU API. With the Apple M3 GPU for AI inference, it achieves around 7,800 points. I would say that for AI inference using the CPU or GPU, Snapdragon X Elite does not perform well, at least when compared to Apple’s M3 chipsets.

After the Geekbench ML benchmark, let’s check out how well the X Elite performs in practical AI applications. We installed the beta version of LM Studio for Snapdragon X Elite and loaded the Phi 3 Mini 3B model. Keep in mind that LM Studio only supports CPU inferencing on the Snapdragon X Elite right now.

In our test, Snapdragon X Elite’s CPU generated response at 24 tokens per second which is amazing for on-device inferencing. In another test, it generated a response at 21 tokens per second. Keep in mind, this speed is based on the 3 billion model. If we run a larger model, the token generation speed will get slower.

Nevertheless, to run 3B and 7B models, even on the CPU, Snapdragon X Elite does a splendid job. With support for GPU and NPU offloading in the future, the experience might get even better.

Finally, we move to the browser benchmark test. We ran Speedometer 3.0 and JetStream 2.0 tests on the native ARM64 version of Google Chrome and Microsoft Edge. Surprisingly, Google Chrome performed better than Edge in the Speedometer test.

The Snapdragon X Elite, in this case, the X1E-78-100 SKU, is undoubtedly apowerful processor, especially when compared to x86 chipsets. However, its high power draw is disheartening as the low power consumption is a key advantage of ARM processors.

While the new Oryon cores are performant, they draw more power than Apple’s M-series CPU cores. Laptops powered by the Snapdragon X Elite will surely offer better battery life than Intel and AMD-powered laptops. However, it’s worth noting that these devices still require active cooling, similar to their x86 counterparts.

The Asus Vivobook S 15 is a compelling product, but from a chipset perspective, we expected X Elite from Qualcomm to match Apple’s performance-per-watt figures more closely.

Notably, Qualcomm has to significantly improve its Adreno GPU for PCs. Currently, it seems like a scaled-up version of the mobile Adreno GPU. Streamlined driver support and gaming compatibility will be key to enticing long-time Windows users to jump on the ARM platform. Finally, framework support for the NPU is crucial for enabling efficient AI processing on the device.

Passionate about Windows, ChromeOS, Android, security and privacy issues. Have a penchant to solve everyday computing problems.