Qualcomm Snapdragon X Elite Analysis - More efficient than AMD & Intel, but Apple stays ahead

The smallest chip is the Snapdragon X Plus with 10 CPU cores, while the Snapdragon X Elite SoCs have 12 CPU cores (Oryon CPU) with two clusters (performance cluster with 8 cores & efficiency cluster with 4 cores). The cores are based on the ARM v8.7 micro architecture and do not support Hyperthreading. The naming scheme of the new Snapdragon X Elite is pretty cryptic and many of the launch models are equipped with the entry-chip X1E-78-100, which does not support dual-core turbo, so all cores can reach the specified 3.4 GHz. The faster models support dual-core turbo frequencies in the range between 4.0-4.3 GHz and the two top models have a higher frequency of 3.8 GHz on all cores. However, our two review units already show that the processor does not really say much about the dual-core performance. Like on AMD and Intel chips, the results heavily depend on the power limits set by the manufacturer. Simply said, the Elite SoCs can operate in a TDP range between 15-80 Watts, which will obviously have a big effect on the performance.

The official designation of the integrated Adreno GPU is Adreno X1-85 and there are currently two versions depending on the processor. The GPUs of the two faster Snapdragon X Elite SoCs manage 4.6 TFLOPS according to Qualcomm, while the slower unit manages 3.8 TFLOPS (also applies for the Snapdragon X Plus). The GPU supports DirectX 12 and features 6 shader processors clocked at 1.5 GHz in the top model and 1.2 GHz in the slower version. there is no dedicated GPU driver like we are used to from AMD and Intel. Instead, there are additional settings (like Super Resolution) available in the GPU settings of Windows.

The NPU of the Snapdragon X SoCs is called Qualcomm Hexagon NPU and manages 45 TOPS. This means the NPU is slightly faster than the required 40 TOPS specified by Microsoft for the Copilot+ features.

We want to compare the different processors and integrated graphics adapters as fairly as possible. In addition to the pure performance in synthetic benchmarks, we also measure the power consumption to get efficiency results. We perform the power measurements with an external screen in order to eliminate impacts caused by different display sizes and technologies. We still measure the overall system consumption and do not solely rely on indicated values for the CPU and GPU. 

We usually use Cinebench R23 for our CPU measurements since the benchmark also runs natively on Apple's M-SoCs and we have plenty of data for this benchmark. However, Cinebench R23 has to be emulated on ARM Windows and the additional translation layer would affect the results. Therefore, we switched to Cinebench 2024, but this means we only have limited data to compare. We will also check the efficiency of the Snapdragon SoC with different power limits and compare the results with modern AMD/Intel chips (also with different power levels) as well as the current Apple M3 as well as the older M2 Pro. We still use the game The Witcher 3 for GPU efficiency results, since it is among the recommended games in Qualcomm's official review guide and we also have plenty of comparison data for this scenario.

We start with the single-core performance and the two new Qualcomm chips perform very well. The X1E-80-100 with the turbo frequency of 4.0 GHz is around 15 % faster compared to the entry-level X1E-78-100 without turbo, both in Cinebench 2024 as well as Geekbench. The X1E-78-100 is sitting right between the two current chips from AMD and Intel and is even slightly ahead in Cinebench. The X1E-80-100 on the other hand is faster than both x86 chips and competes with the old Apple M2 Pro. Apple's M3 and the new M4 are out of reach, though. Apple's latest M4 chip has a massive advantage of 30 % in Geekbench.

Let's talk about single-core efficiency next. At the bottom of our comparison chart are the two CPUs from AMD as well as Intel, while the new Snapdragon chips manage about twice the number of points per Watt. It is also interesting to see that the single-core efficiency of the two Snapdragon X Elite SoCs is basically identical. Apple's wo CPUs are out of reach, though. The old M2 Pro manages 25 % more points per Watt while the M3 almost manages twice the points per Watt. Unfortunately, we have no efficiency data for Apple's M4 yet since it is only available in the new iPad Pro and we are not able to perform accurate power measurements with an external screen. 

Update June 22nd: The single-core performance of the X1P-64-100 is on par with the X1E-78-100, which is hardly surprising considering the same core clock of 3.4 GHz. However, it seems that the Snapdragon X Plus is slightly more efficient here (at least our test unit) with better results compared to the two Snapdragon X Elite units.

The Qualcomm SoCs also perform well in multi-core scenarios and Cinebench 2024 shows that they scale nicely. At 35 Watts, the Intel Core Ultra 7 155H (even at 50 Watts) is beaten and only the Ryzen 7 8845HS at 54 Watts is a bit faster. Apple's M2 Pro SoC is beaten when the Qualcomm chip can use 45 Watts or more. The Geekbench performance on the other hand is pretty much identical with different power settings and comparable to Apple's M3 Pro with 14 CPU cores. 

Qualcomm and some of the laptop manufacturers advertise the performance advantage over the Apple M3, and we can confirm that claim in multi-core tests. However, the M3 is usually passively cooled in most devices and the real competitors for the Snapdragon SoCs are the M3 Pro chips. We will add the data for the M3 Pro chips as quickly as possible.

Our efficiency numbers clearly indicate that the new Snapdragon CPUs are generally more efficient than the two chips from AMD and Intel in multi-core scenarios. The most efficient power range of the Snapdragon X Elite chips seems to be 20-30 Watts, where they even slightly beat the Ryzen 7 8845HS at 20 Watts. Apple's M2 Pro is also beaten in this range. However, if you increase the power limits, the Snapdragon X Elite loses efficiency quickly, especially at v45 Watts or more. This is not a good sign for the faster Snapdragon variants. Apple's M3 is much more efficient and based on efficiency comparisons we did with Cinebench R23, we know that the M3 Pro chips are much more efficient than the M2 Pro models. It is therefore safe to say that the M3 Pro CPUs will also be much more efficient than the Snapdragon models.

Update June 22nd: The Snapdragon X Plus in the Surface Pro Copilot+ is between 8-11 % slower compared to the Snapdragon X Elite in the Surface Pro OLED Copilot+. It is also 31 % faster than the Apple M3 in Cinebench 2024, but the advantage shrinks to about 10 % in Geekbench. The efficiency numbers of the Snapdragon X Plus is a bit worse compared to the X Elite SKU, but the difference is not huge. Depending in the TDP limits, these values can obviously be different for other devices.

We use the native benchmarks 3DMark Wild Life Extreme Unlimited, Geekbench, as well as GFXBench for our GPU comparison. The benchmarks show big differences and the OpenCL performance, for example, is worse compared to the immediate rivals. the new Adreno GPU sits between the Radeon 780M and the Intel Arc Graphics with 8 Xe cores in 3DMark, but Apple's M3 chips (and obviously M4 & M2 Pro) are faster. 

Qualcomm's Adreno GPU is slightly faster than the 8-core GPU of the Apple M3 in the two High-Tier tests of GFXBench, but all the other M3 GPUs as well as the M2 Pro are once again faster.

Our GPU efficiency results are similar to the CPU results, because the Adreno GPU (which should consume around 12-15 Watts based on our power measurements) is once again more efficient than the competition from AMD and Intel, but is clearly lagging behind the Apple GPUs.

We also checked the power consumption of the laptops while idling and during playback of a 4K YouTube video (full screen). We once again used an external 4K display to avoid impacts from different panel sizes and technologies. The test video was our review of the Asus Vivobook S 15 OLED and we used the default browsers.

The idle measurements show that in addition to the SoC itself, the results also depend on the optimization by the manufacturers. The Snapdragon X Elite in the Surface Pro OLED Copilot+ is slightly more efficient compared to the Asus Vivobook S 15 OLED, for example. The RedmiBook Pro 14 with the Intel Core Ultra 7 155H is sitting between the two Qualcomm chips, while the AMD Ryzen 7 8845HS in the Schenker Via 14 Pro is lagging behind. We had no access to the respective U-series chips for this test, but we will add the data as soon as possible. Apple's M3 chip is still the most efficient SoC.

This is also the case for video playback, but the two Qualcomm SoCs are not far behind. Both the Intel as well as AMD system consume significantly more power in this scenario.