Snapdragon X Elite with 23W and 80W TDP: even the weaker version outperforms AMD, Intel and Apple

A new processor was introduced a week ago Qualcomm Snapdragon X Elite, which wants to break into the notebook market on the ARM architecture. He’s pretty good at it. It is manufactured using a 4nm process, contains 12 cores, with the Boost of all cores peaking at 3.8 GHz, one or two cores can even run up to 4.3 GHz. Now we finally find out what the consumption actually is. The original graphs were up to 50W, so it was safe to assume that the max consumption would be 50W or more. We now know that this variant has 80W consumption, and Qualcomm presented the results of tests where this processor was used in a 16.5″ notebook with a 4K display, a thickness of 16.8 mm, an 87Wh battery and LPDDR5x-8333 memories.

The second version of the processor (23W) was tested in a smaller 14.5″ body also with LPDDR5x-8333 memories, this time it was a notebook with a resolution of 2880×1800 pixels on an OLED display, the notebook was 15 mm thick and had a 58Wh battery. The boost of all cores is 3.4 GHz, Boost on one or two cores can then reach 4.0 GHz, but this relatively small reduction in clocks has significantly reduced power consumption to just 23 W.

Single-threaded performance in Cinebench is excellent, and even the 23W version outperforms both the 8-core AMD Ryzen 9 7940HS processor (at 80 W) and the 14-core Intel Core i7-13800H (at 115 W), as well as the Apple M2. But here we also see a minor snag. It is clear that the Apple M3 presented yesterday cannot be here, we would probably expect a slightly different order, but the issue will be more multi-threaded performance.

The problem is that the Apple M2 is an 8-core processor, while MacBook Pros also exist with M2 Pro processors with 10 or 12 CPU cores, which would certainly be a more relevant comparison to the 12-core Snapdragon. Basically, we could theoretically expect half the performance from Apple in these multi-thread tests (and if you look in the database, you will actually see results of around 780 points for the 12-core variant). However, we can see that the 80W Snapdragon X Elite outperforms both the 20-thread Intel and 16-thread AMD by a large margin. The economical 23W variant follows closely behind them with significantly lower consumption.

The Geekbench benchmark also shows a very high single-threaded performance of the new Snapdragons, even the one with a lower TDP outperforms AMD, Intel and Apple processors.

In the multithreaded variant, Intel is already sticking out its claws, but Qualcomm still holds first place with the 80W version. But Apple’s last place is due to the fact that a variant with a much smaller number of cores than it should be is used here (intentionally?). If we look at the performances of the 12-core M2 Pro, they range from 14.2 to 14.8 thousand points. So it would be enough for second place. In the same way, the most powerful Intel was not chosen, which is the 24-core (32-thread) Core i9-13980HX, on the other hand, the number of cores would really differ a lot (but it would be interesting if Qualcomm surpassed even this strongest Intel).

In the Aztec Ruins graphics benchmark, Qualcomm still holds the top spot, but even the basic Apple M2 with a 10-core GPU is making a lot of progress on the more economical version (here, let’s remind you that the 12-core version of the M2 Pro has 19 GPU cores).

Next, we have 3DMark Wildlife Extreme. Although Qualcomm holds the first position, here again it would be appropriate to say that if it had chosen the more adequate 12-core M2 Pro with a 19-core GPU for comparison, then it would not have been the first by far. This processor achieves about 65 fps in this test, so the new Snapdragon would run into the ground.

Thanks to a special chip for processing artificial intelligence algorithms in the Snapdragon processor, we see an order of magnitude higher performance in Qualcomm processors. The Apple M2 processor was missing this time (the benchmark is not for macOS). It is interesting that the Ryzen 9 7940HS had such low performance, although it supports Ryzen AI (it has a unit for calculating AI algorithms). Intel will only get this at the turn of the year with Meteor Lake processors.