When you use a laptop, you might assume that it will always perform the same, whether you use it while plugged in or unplugged. After all, these systems are designed to be used on the go. But for a long time, that hasn't been the case.
More recently, some silicon vendors have been boasting about closer parity in plugged and unplugged performance, offering you similar power whether you're working at your desk or anywhere else. To see the current state of this, we've taken samples of the latest laptop from each major CPU manufacturer: Intel, AMD, Qualcomm, and Apple, and run a few tests to see how they compare.
For the Windows systems, we used the balanced power profile, which was already the default on the majority of these laptops, while the MacBook Pro has an "automatic" power management profile out of the box.
Where we could, we went for top-end processors, including the Intel Core Ultra X9 388H, Qualcomm Snapdragon X2 Elite Extreme, and Apple's M5 Max. In AMD's case, we had two Gorgon Point processors, a Ryzen AI 9 465 and a Ryzen AI 7 445. To see some generational gains, we also included an Intel Core Ultra 7 258V "Lunar Lake" chip.
We tested three ways: in Cinebench 2026 single-core and multi-core workloads, as well as using Handbrake, an open source video conversion tool, to transcode a 4K video to 1080p.
Note that we're not comparing the scores of the systems to each other. We're comparing the systems to themselves, to see how performance changes on the battery.
With those tests run, we can calculate percentage differences to see percentage changes. Notably, a number of systems actually increased performance ever so slightly on battery. This occurred the most on multi-threaded Cinebench, where the Zenbook A16 with Snapdragon X2 Elite Extreme increased by 13.06%, while the ThinkPad X9 bumped up by 4.98%.
| Row 0 - Cell 0 | Asus Zenbook Duo (Intel Core Ultra X9 388H) |
Asus Zenbook A16 (Snapdragon X2 Elite Extreme) |
Apple MacBook Pro (14-inch, M5 Max) |
Acer Swift Go 16 AI (AMD Ryzen AI 7 445) |
Average |
+0.01% |
-0.41% |
+0.08% |
-50.06% |
Cinebench 2026 (Single-threaded) |
0% |
-0.96% |
+0.27% |
-56.25% |
Cinebench 2026 (Multi-threaded) |
-3.41% |
+13.06% |
+0.85% |
-42.28% |
Handbrake |
+3.43% |
-13.33% |
-0.89% |
-51.64% |
Qualcomm's chip was by far the most variable. While it dramatically increased in performance on battery for Cinebench nT, it dropped a similar amount for Handbrake.
But it was Apple and Intel that led the pack. While the Core Ultra X9 388H had the most consistent score on average, it did show 3% swings in multi-threaded Cinebench and on Handbrake. Apple's M5 Max results were the most consistent throughout, never swaying even a full percentage point on battery.
The least consistent system on battery power was the Acer Swift Go 16 AI, running on an AMD Ryzen AI 7 445. That system lost a ton of performance when unplugged from the wall, dropping as much as 56.25% performance off the wall. On average, benchmark scores were 50% lower on that system. On the HP OmniBook X Flip 14, Ryzen AI 9 465, it still dropped, but far less, losing an average of 5.45% performance on power based on these tests.
On the bright side, things are going in the right direction. Intel and Apple's most advanced systems give you very similar experiences on battery and power — at least in these tests. Snapdragon can, but sometimes it swings a bit. It's really AMD that needs to catch up in this regard. And we know that can be done. The Steam Deck, which was released in 2022, was specifically designed to have similar performance on and off the charger, and it uses a semi-custom AMD SoC. Its AMD-based opponents, however, still get better performance on a charge. You not only have to have your chip designed around battery and performance. You also need to make sure that the system around it is optimized for that.
The other side of this coin is efficiency. If you're able to perform heavy tasks on battery without losing performance, you're likely going to do it. That's especially the case if you do field work in creative or industrial fields. That means ensuring that your system can handle your workload long enough to get it done. But hey, there's always an external battery, right?
