Quick Specs
Our Verdict
Arrow Lake's flagship is a genuinely inventive tiled processor that finally makes Intel run cool and power-efficient, with strong multi-thread performance and a much-improved iGPU. It doesn't beat the 14900K outright, trails AMD on efficiency and lags the X3D chips in games, but it does its work far more comfortably — a promising fresh start.
Our editorial process
Research method, author and affiliate-independence details
Our editorial process
Research method, author and affiliate-independence details
Prepared by our editorial team using verified source material, product research, and a British-English editorial rewrite before publication.
- We review the working bundle for product facts, comparisons, and buyer-relevant tradeoffs before publishing.
- Non-English source material is translated into British English and rewritten into our house style without carrying over publication branding.
- Affiliate links and price references are handled separately from editorial judgements and never determine the verdict.
Read our full methodology: how we review
Affiliate links never determine our verdicts. Commercial relationships are disclosed separately from the editorial assessment, and we aim to keep buyer guidance clear, specific, and evidence-based.
Pros & Cons
Pros
- Strong performance, especially multi-threaded
- Runs far cooler and draws much less power than the 13th/14th gen
- Much-improved integrated Xe graphics
- Good efficiency-per-pound thanks to keen launch pricing
Cons
- Core i9 and Ryzen X3D chips stay ahead in games
- HyperThreading is gone
- Windows 11 24H2 scheduler bugs marred the launch
- NPU is a token 13 TOPS (below Copilot+ needs)
- Still less efficient and hotter than AMD's Ryzen 9000
Full Specifications
Key Features
Strong performance, especially multi-threaded
Runs far cooler and draws much less power than the 13th/14th gen
Much-improved integrated Xe graphics
Good efficiency-per-pound thanks to keen launch pricing
Intel has been through a rough patch, and with Arrow Lake it finally seems to be rethinking its desktop processors. This is the first generation to borrow so heavily from Intel's laptop approach, betting on energy efficiency rather than ever-higher clock speeds. So how good is the flagship Core Ultra 9 285K? This review is based on that published test, not on a hands-on trial of our own.
A Small Revolution: The Arrow Lake Architecture
Where Alder Lake was a big change in its day, Arrow Lake goes further: for the first time on the desktop, the 285K is a true system-on-a-chip, built from separate "tiles" bonded together by Intel's Foveros 3D packaging. Tellingly, every one of the four essential tiles — compute, GPU, SoC and I/O — is manufactured by TSMC, on nodes down to 3 nm; Intel handles the design and assembly rather than the fabrication itself.
The compute tile pairs new Lion Cove performance cores with Skymont efficiency cores, and the headline change is the disappearance of HyperThreading — each P-core now handles a single thread. Intel claims a 9% IPC gain over the 14th generation, with far bigger strides on the efficiency cores (up to 32% single-thread and 72% multi-thread over the old Gracemont), finer clock control in 16.67 MHz steps, down from the old 100 MHz, and up to 36MB of combined cache. Those efficiency gains lean on a wider branch predictor, more dispatch ports and a doubled L2 cache on the Skymont side. An improved Thread Director acts as conductor, allocating work to the right cores at the right moment, better classifying tasks and shifting them from the efficiency to the performance cores as demand rises.
iGPU, NPU and Platform
The 285K is not an APU, but its integrated graphics have grown up: first-generation Xe cores (not the Xe2 destined for the Battlemage graphics cards), with four Xe cores, full DirectX 12 support and roughly double the graphics performance of the 14th generation — not enough for demanding games, but fine for lighter titles and everyday display. The NPU is more of a token effort at 13 TOPS, well below the 40 TOPS that Copilot+ requires. The SoC tile also brings two Thunderbolt 4 ports, PCIe Gen 5, a generous spread of USB ports running at up to 5, 10 and 20 Gbps, and support for the new CUDIMM memory standard. The 285K itself packs 24 cores — eight performance plus sixteen efficiency — reaches up to 5.7 GHz, sits in the new LGA1851 socket, and needs a new chipset: the Z890 leads the way, with more affordable boards to follow. At launch it headed a three-strong K-series line-up alongside the Core Ultra 7 265K and Core Ultra 5 245K.
Performance
In applications the 285K is a strong multi-threaded performer: it clearly improves on the Core i9-14900K and, in Cinebench multi-thread, pulls comfortably ahead, though single-thread and rendering benchmarks such as Blender and a 50 min HandBrake encode still hand the win to AMD's Ryzen 9 9950X, usually by a modest margin. It dominates the memory tests. Gaming is more mixed. In Cyberpunk 2077 the Arrow Lake chips trail the field — beaten by the previous Raptor Lake generation, by Ryzen 9000, and comprehensively by AMD's gaming champion, the Ryzen 7 7800X3D — while Total War: Warhammer 3 and Shadow of the Tomb Raider are far more competitive, the 285K sitting ahead of Ryzen 9000 even if the older Core i9s and the X3D stay in front. Testing was, in fairness, complicated throughout by scheduler problems in the Windows 11 24H2 update, which Intel and Microsoft will need to iron out.
Heat, Power and Efficiency
The most striking progress is in temperatures and power. Intel used to be a byword for running hot; the 285K is not. It runs markedly cooler than the 13th- and 14th-generation chips, and its power draw is transformed — 244W under a Cinebench multi-thread load, against 338W for the Core i9-14900K in the same test. That is only half the story, though: despite TSMC's 3 nm process, the 285K still runs a touch hotter and less efficiently than AMD's Ryzen 9000 parts, which retain the efficiency-per-watt crown. Where Intel does score well is efficiency-per-pound, thanks to fairly aggressive launch pricing.
Verdict
New architectures rarely arrive without teething trouble, and the 285K's launch — tangled up with the messy Windows 11 24H2 scheduler — was no exception. On one hand this is a genuinely inventive chip that shows real intent from an Intel going through a difficult spell; on the other, Arrow Lake does not seize the lead from the competition. Optimists will point to the big efficiency gain over Raptor Lake Refresh (though much of that is TSMC's 3nm doing); pessimists will note that, even so, the chip stays hotter and less efficient than AMD's, and lags the X3D parts in games. What is not in doubt is the comfort of the thing: the Core Ultra 9 285K may not beat the Core i9-14900K outright, but it does its work far more calmly — no throttling worries, cooler, and much less power-hungry. A promising, if imperfect, fresh start.
Ready to Purchase?
Check current prices and availability
Affiliate Disclosure: As an Amazon Associate I earn from qualifying purchases. Truthful Reviews may receive a commission when you buy through eligible Amazon UK or Amazon US links, at no extra cost to you. Affiliate partnerships never determine our verdicts or recommendations.


