Intel’s new desktop CPU units differ radically from the conventional Intel CPUs. Today marks the launch of Arrow Lake, the latest architecture for Intel’s desktop processors, officially sold as the Core Ultra 200S series. According to Intel’s claims, this batch of chips is faster and more energy-efficient than the latest-generation Raptor Lake CPUs, while still outperforming the competition.
However, Arrow Lake is not just remarkable for what it does, but also for what it is. Since the first set of x86 desktop processors were shipped but not manufactured by Intel, TSMC’s involvement in production is a new distinct aspect. Intel has also redesigned its chip design, with a complete overhaul coming with some hiccups, as we discovered during our tests.
To dive deeper into the productivity standards, check out Gordon’s review of the Core Ultra 9 285K in the video below.
Here, let’s delve into the top ten things you need to know about Intel’s Core Ultra 200S CPUs, after an intensive test of the flagship Core Ultra 9 285K processor.
New Names for a New Era
Last year, Intel revamped its labeling system for its core processors, dropping the “i” prefix before the numbers 3, 5, 7, and 9, and eliminating the reference to the generation. It also compressed the term “Ultra” to signify CPUs containing Arc graphics and a Neural Processing Unit (NPU) for AI-based tasks, along with resetting the numbering system to begin with the hundreds.
This change was first introduced in laptops powered by the Meteor Lake CPUs at the end of 2023. The Arrow Lake release marks the first time we see the new names on desktops, a significant shift after over a decade of the old naming conventions. Intel also chose to start numbering Arrow Lake in the 200 range, despite it being the first for desktops – likely to align the laptop and desktop architectures from the same year. (Lunar Lake, which just started shipping laptops in September 2024, also categorizes its processors in the 200 range.)
These are the five CPUs in the new Core Ultra 200S series, which include two different types without integrated graphics, referred to as “F” (this term remains the same as before):
- Intel Core Ultra 9 285K: 24 cores (8 P cores, 16 E cores, up to 5.7 GHz); 4 GPU cores, 13 TOPS NPU, $589
- Intel Core Ultra 9 265K: 20 cores (8 P cores, 12 E cores, up to 5.5 GHz); 4 GPU cores, 13 TOPS NPU, $394
- Intel Core Ultra 9 265K F: 20 cores (8 P cores, 12 E cores, up to 5.5 GHz); 0 GPU cores, 13 TOPS NPU, $379
- Intel Core Ultra 9 245K: 14 cores (6 P cores, 8 E cores, up to 5.2 GHz); 4 GPU cores, 13 TOPS NPU, $309
- Intel Core Ultra 9 245K F: 14 cores (6 P cores, 8 E cores, up to 5.2 GHz); 0 GPU cores, 13 TOPS NPU, $294
Hyperthreading is Dead (Again)
Intel introduced Hyperthreading technology to allow a CPU core to run two threads simultaneously, instead of the default single thread per physical core. However, in recent years, the company has considered phasing it out. For Arrow Lake, the feature has once again been bypassed.
As Intel recently explained, it abandoned Hyperthreading as part of efforts to improve energy efficiency and space in Arrow Lake, i.e., the amount of power the chip uses and can fit within a package of each processor.
Therefore, the core counts you see for the Core Ultra 200S CPUs are what you get – without any extra threads. However, the number of performance cores (known as P-cores) and efficiency cores (known as E-cores) packed in Arrow Lake are not few. The top Core Ultra 9 285K chip still contains 24 cores overall, with 8 P cores and 16 E cores.
Performance Gains for Content Creators
Arrow Lake’s performance will be a sensitive topic among chip enthusiasts, who have grown accustomed to significant performance gains over the past two years. However, this was possible due to high power limits, as modern flagship processors primarily operated by raising clock speeds as a default setup. In contrast, the key feature of the Core Ultra 200S series is energy efficiency, which Intel emphasizes as a selling point.
With this change, Arrow Lake provides improved performance for creators, but the gains are not spectacular enough to force most owners of the Core i9-14900K to upgrade yet.
Across our rendering and encoding benchmarks, the Core Ultra 9 285K consistently outperforms the latest-generation 14900K, with improvements ranging from 2 to 21 percent depending on the task. It also mostly outperformed the Ryzen 9 9950X. The only exception: it lagged slightly behind in DaVinci Resolve compared to older CPUs.
But Gaming is Less of a Win
Gamers love to see new CPUs crushing the current competition, and for those in this crowd, the Core Ultra 9 285K may seem disappointing. It’s a chip that falls behind the 14900K and 9950X in some games, equals them in others, and outperforms them in some games.
In our small set of gaming benchmarks, the 285K generally keeps up with the 14900K, but often trails by a two-digit margin behind the Ryzen 9 9950X from AMD. In Cyberpunk 2077, the 9950X maintained an 18 percent lead over the 285K, with similar performance in Formula 1 2023 and Rainbow Six Siege. The 285K triumphed in 3DMark’s Steel Nomad test, an artificial gaming benchmark simulating a dense game at 4K resolution.
Even when paired with faster and more performance-oriented CU-DIMM memory (instead of DDR5), gaming performance does not see a significant improvement – in Formula 1 2023, we saw an increase to 410 frames per second, or about 7 percent. This narrows the gap between the 285K and the 9950X, but the latter still leads.
The 285K chip is by no means a bad chip for gaming, but it won’t blow anyone’s socks off. So, while Intel was open about Arrow Lake’s gaming performance, motherboard compatibility may play a role.
Power Efficiency is Improved, But…
The reason behind the smaller performance gains of Arrow Lake lies in Intel’s claim that its new chips match the performance of the 14th generation Raptor Lake-R CPUs at half the power, but you may not see such a significant improvement.
Compared to the 14900K, our tests showed a 17 percent decrease in power consumption during the Handbrake AV1 benchmark (a difference of 65 watts). Similarly, during the Cinebench 2024 single-core benchmark, power usage decreased by about 16 percent (a difference of 22 watts). In idle mode, the power draw of the 285K increased slightly, rising by 3 percent (about 2.5 watts). These three benchmarks cover the general range of typical computer behavior.
The 285K performs better compared to top AMD performance, with a 4 percent decrease in power consumption in Handbrake, an almost 25 percent decrease in Cinebench 2024, and about a 28 percent decrease in idle power consumption compared to the 9950X.
Overall, Intel’s improvements give it an edge if you’re concerned about heat or power bills. But the story doesn’t end there. It turns out that when delving into power consumption, these new chips reveal an unexpected quirk.
Windows Optimizations May Be a Work in Progress
Desktop users may not pay much attention to power settings in the Windows operating system, as they are less relevant when connected to the wall all the time. At least, that’s usually the case.
However, being an odd bunch, we examined the power consumption of the 285K on various Windows 11 power plans. Strangely enough, 285K’s performance slowed significantly on the balanced and power saver plans.
In the Cinebench 2024 single-core test, the speed of the 285K noticeably slowed down while completing its task, extending to 30 minutes on the balanced plan. In the high-performance plan, the 285K remained closer to a 10-minute run (more common for high-end desktop chips). Highlighting this contrast, both the 14900K and 9950X finished at their usual times – meaning we could start benchmarks for these chips later than the 285K (15 minutes of 285K running) and still see them complete the task faster.
Additionally, the resulting comparison showed a 55 percent performance decrease compared to the 14900K and 9950X on the same balanced power plan. The difference was greater on the power saver plan, where it decreased by 67 percent.
At the same time, on the high-performance settings, the 285K actually outperformed its predecessors by about 5 percent and the 9950X by about 2 percent. This result is more puzzling, as our power consumption measurements showed minimal energy savings across the various power plans.
Does this mean Intel’s new chip could falter through your power settings? Right now, it’s likely yes, especially since the balanced power plan is the default Windows plan. Time will tell if this behavior can be fixed through better Windows optimization.
Intel Revamped Its Processor Design
The increased complexities with designing new chips are not extremely surprising, as Arrow Lake differs significantly from the 14th generation Raptor Lake-R.
So far, Intel has released homogenous processors for desktops; everything concerning the CPU was in a single package. But starting from Arrow Lake, Intel has transitioned to designing smaller chips, a step taken by the competing AMD for years.
Intel labels Arrow Lake as the “detailed” processor, where separate chips (“tiles”) for different functions are linked together as a single package. If you were to remove the cover of an Arrow Lake CPU, you would find compute squares, graphics processing unit, SOC, I/O, as well as “filler” and “substrate” blocks for stability. Intel’s Foveros technology links them all together. Each of these small chips is manufactured through different processes – you can read more about it in the initial overview of Arrow Lake.
For chip enthusiasts, this shift in approach represents a significant change for Intel, which previously advocated for a single package as the best performance. However, for most home users, what matters most is how well Intel managed to coordinate the package design.
You’ll Need a New Motherboard
Arrow Lake requires a new socket – LGA 1851 – meaning it will not be compatible with current LGA 1700 motherboards.
The first chipset series to be announced is the Z890, which supports up to 192GB of DDR5-6400 memory (up to 48GB per DIMM). You can use the more common SO-DIMM modules, but the latest CU-DIMM memory modules also work. Intel states that if you enable XMP for faster memory speeds, DDR5-8000 is the optimal choice.
For connectivity, the Intel 800 chipset offers up to 24 PCIe 4.0 lanes, along with 10 USB 3.2 ports, 14 USB 2.0 ports, and 8 SATA 3.0 ports. With the Arrow Lake CPU, you’ll get up to 48 PCIe lanes, 20 of which are PCIe 5.0.
