Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
Although AMD loses out compared to the Core Ultra 7 265K and Core i7-14700K in productivity workloads, its simple arrangement of eight cores within a single CCD leads to much lower power draw. As you can see from our results in demanding, heavily-threaded workloads like Cinebench, Handbrake, and Blender, peak power consumption is around 170W, with many workloads running around 130W to 140W even while the chip is pushed. The Core i7-14700K, meanwhile, has some of the highest peak power consumption out of our test pool, only rivaled by other Raptor Lake (and Refresh) offerings. The Core Ultra 7 265K shows off Intel’s power improvements with Arrow Lake, but it still isn’t enough to match most Zen 5 chips.
The Ryzen 7 9850X3D and Ryzen 7 9800X3D carry the same 120W TDP, but there are significant deviations in real-world power consumption across both gaming and productivity apps. The updated chip consumed 20% more power in the Junkshop scene from Blender, and it consumed 22% more power at idle. In the latter case, we’re looking at the difference between 18W and 22W, so the power jump is less significant than it sounds. However, there’s still a jump.
AMD has clearly shown how efficient its Zen 5 chips are in demanding workloads, with even the Ryzen 9 9950X3D coming in below Intel’s slate of Core i7 offerings from the past few generations when pushed. On the flip side, AMD has much higher active idle power draw. For this test, we run a browser with three tabs open, one of which is playing a YouTube video. Intel shows better results in this scenario, and once again, the Ryzen 7 9850X3D consumes more power than its predecessor — a relatively benign 10% jump in this particular test.
For raw efficiency numbers, there’s a clear split between Arrow Lake and AMD, and all previous Intel chips in our test pool. Arrow Lake has put AMD back into the efficiency conversation, but it’s still a game that AMD is mostly winning.
Our scatterplots show this more clearly. In Handbrake, you can almost literally see the line between the two batches of CPUs described above. The Core Ultra 7 265K really stands out with its balance of overall performance and power consumption, while the Core Ultra 9 285K almost matches the Ryzen 9 9950X3D.
Blender shows clearer scaling between power usage and performance, but with a distinct difference between how that scaling looks between AMD and Intel. In both of these scatterplots, we’re looking at cumulative energy usage across the benchmark pass rather than an average. That gives us useful data for how much energy a task actually uses (measured in Kilojoules), which we can then plot against performance.
Test Setup
We run nearly identical test benches to reduce as many variables as possible when benchmarking. All three test benches used for this review use identical hardware, short of the motherboard and, of course, the CPU. We used XMP/EXPO for all testing (memory speeds below). In addition, all tests were run with VBS disabled and Resizeable BAR turned on.
There are several important software notes to keep in mind, particularly with this test pool. Zen 5 originally launched with laggardly gaming performance, so we tested with the latest AGESA. In addition, we used the 105W power mode for the Ryzen 7 9700X, as that is an officially-warrantied operating mode for the Ryzen 7 9700X, and it improves performance. We also applied microcode 0x12B to Alder, Raptor, and Raptor Lake Refresh chips.
Our test images are frozen for periods of time to keep testing consistent. For this series of tests, we used Windows 11 24H2. The latest 25H2 update doesn’t show notable performance differences — short of a handful of spotty Nvidia drivers — so we kept the version the same. We did, however, update to the latest Nvidia driver, not only to support Doom: The Dark Ages, but also to bring a more accurate picture of performance on a modern platform.
Intel LGA 1851 (Arrow Lake) | Row 0 - Cell 1 |
Motherboard | ASRock Z890 Taichi |
RAM | 2x16GB G.Skill Trident Z Neo RGB DDR5-7200 |
Intel LGA 1700 (Raptor Lake, Alder Lake) | Row 3 - Cell 1 |
Motherboard | MSI MPG Z790 Carbon Wi-Fi |
RAM | 2x16GB G.Skill Trident Z Neo RGB DDR5-7200 |
AMD AM5 (Zen 5, Zen 4) | Row 6 - Cell 1 |
Motherboard | MSI NPG X870E Carbon Wi-Fi |
RAM | 2x16GB G.Skill Trident Z Neo RGB DDR5-6000 |
All Systems | Row 9 - Cell 1 |
Gaming CPU | Nvidia GeForce RTX 5090 Founder’s Edition |
Application GPU | Nvidia GeForce RTX 2080 Ti Founder’s Edition |
Cooler | Corsair iCue Link H150i RGB |
Storage | 2TB Sabrent Rocket 4 Plus |
PSU | MSI MPG A1000GS |
Other | Arctic MX-4 TIM, Windows 11 Pro, Alamengda open test bench |
MORE: CPU Benchmark Hierarchy
MORE: AMD vs. Intel
