Intel's latest flagship processor delivers mixed results in benchmark testing, analysis suggests.

The Core Ultra 9 285K has arrived at $589, maintaining the same price point as its predecessor while promising substantial performance improvements. Tests reveal the processor struggles to deliver consistent gains across all workloads despite its new chiplet architecture manufactured on TSMC's N3B process node.

Independent testing shows gaming performance declining compared to the previous generation Core i9-14900K. Some titles record up to 43% lower framerates with the new chip. Although the processor supports DDR5 memory with bandwidth reaching 121-123 gigabytes per second, increased memory latency of approximately 80 nanoseconds appears to hamper gaming performance significantly.

Intel claims the Core Ultra 9 285K delivers up to 150W reduction in system power during gaming and 20% improvement in threaded performance. However, the removal of Hyper-Threading has reduced the total thread count from 32 to 24 (8P+16E), despite Intel maintaining this architectural decision enhances multi-threaded workloads.

The processor operates at a maximum boost clock of 5.7GHz—300MHz lower than its predecessor's 6GHz—though Intel reports a 9% increase in instructions per clock. These contradictions between marketing claims and real-world benchmark results raise questions about the chip's value as an upgrade option for most users.

Intel introduces first chiplet design with Core Ultra 9

Image Source: Chips and Cheese

Arrow Lake marks Intel's first chiplet-based design for desktop processors, moving away from the monolithic approach used in previous generations.

The new architecture, codenamed Arrow Lake, powers the Core Ultra Series 2 processors released in October 2024. Unlike earlier chips, these processors are primarily manufactured by TSMC, with the compute die using TSMC's N3B 3nm process node. This design change delivers up to 40% lower package power consumption compared to 14th Gen processors and introduces Intel's new Lion Cove P-cores with 9% higher instructions per clock. The Skymont E-cores show 32% higher integer IPC.

Intel's Core Ultra 9 features a disaggregated design with separate tiles for different functions. The compute tile houses the CPU cores, the GPU tile contains the graphics engine, and the SoC/I/O tiles handle memory controllers and connectivity. These components connect via Intel's Foveros 3D packaging technology, creating high-density microscopic connections between tiles. The approach reduces the package size by 33% and improves manufacturing yields as each tile can be produced independently.

For the first time since 2002, Intel has removed Hyper-Threading from its performance cores. The Core Ultra 9 285K features 24 cores and 24 threads (8P+16E), fewer than the 32 threads of its predecessor. This change reduced transistor count and power consumption, but could affect performance in memory-intensive workloads where simultaneous multi-threading typically provides 5-25% improvement.

Intel completely redesigned the core arrangement, placing E-core clusters between P-cores rather than separating them. Four P-cores sit on the borders with four more in the middle, while four E-core clusters (each containing four cores) are positioned between them. The E-cores now connect directly to the shared 36MB L3 cache, while P-cores receive larger 3MB L2 caches (increased from 2MB), and each E-core cluster shares 4MB of L2 cache. This configuration reduces thermal hotspots and improves inter-core communication latency.

Gaming performance slumps in Intel's latest flagship

"While it put up impressive productivity and multi-core results, its gaming performance sank below Intel's previous-gen Raptor Lake Refresh at launch." — Michael Crider, Technology journalist at Tom's Hardware

Tests show Intel's newest processor struggling with games despite architectural improvements, analysis reveals.

The Core Ultra 9 285K demonstrates unexpected weaknesses in gaming performance where previous Intel chips excelled. Benchmark results indicate consistent performance regression compared to the previous generation flagship.

Gaming tests reveal the Ultra 9 falling behind the 14900K across popular titles. The processor trails its predecessor by 8% in Shadow of the Tomb Raider, 9% in Cyberpunk 2077, and a substantial 20% in Forza Motorsport. In Final Fantasy XIV: Dawntrail, the 14900K establishes a 15% lead over the 285K. More concerning, Baldur's Gate 3 benchmarks showed up to 43% performance regression compared to previous-generation chips when using identical memory kits.

Increased memory latency appears to be the primary cause of this gaming performance decline. Intel's architectural decision to separate the memory controller and PHY into the I/O tile adds approximately 15-20 nanoseconds of memory latency. This creates a significant bottleneck for gaming workloads that depend heavily on rapid memory access.

Intel's new CUDIMM memory technology offers partial improvement but fails to fully address the issue. With standard DDR5-7200 memory, the 14900K remains 6% faster than the 285K. Even when equipped with high-end CUDIMM DDR5-8200 memory, the previous-generation 14900K still outperforms the 285K by 4%. Certain games like Baldur's Gate 3 show marked improvements with CUDIMM technology, primarily by mitigating the latency disadvantage.

System configurations further complicate performance comparisons. Testing reveals significant performance variations depending on motherboard BIOS settings. ASUS, ASRock, MSI, and Gigabyte boards showed inconsistent default configurations. Operating system versions also impact results, with Windows 24H2 generally performing better with the Core Ultra 9, though Cyberpunk 2077 ran notably better on Windows 11 build 2152.

Intel's Application Optimization (APO) software was designed to enhance performance by optimizing thread allocation. Initially supporting just two games, Intel has expanded compatibility to 26 titles. APO can boost framerates by up to 31% in specific situations, primarily by directing gaming workloads to P-cores while background tasks run on E-cores. However, current testing shows APO delivers minimal improvements without activating Windows "high performance" mode.

"This boost is likely primarily due to optimizations dealing with Intel's recent big-core/small-core design philosophy, using a mixture of performance and efficiency cores to balance out workloads." — Michael Crider, Technology journalist at Tom's Hardware

The Core Ultra 9 285K delivers significant performance gains in creative workloads while running cooler and more efficiently than its predecessor, testing shows.

Creative professionals see substantial improvements with the new chip. In Cinebench 2024, the 285K scores 17% higher than the 14900K in multithreaded tests and outperforms the Ryzen 9 9950X by 11%. Blender benchmarks reveal the Ultra 9 rendering 15% faster in the Monster workload, 19% faster in Classroom, and 5% faster in Junkshop compared to its predecessor.

Adobe applications particularly benefit from the new architecture. The Core Ultra 9 delivers 6% better performance in Adobe Premiere versus the 14900K and scores 16,231 points compared to the 14900K's 15,179. Photo editing sees up to 21% improvement over 12th-gen chips.

Perhaps most impressive is how the Core Ultra 9 285K achieves these gains while improving efficiency. Intel claimed a 40% reduction in package power consumption, and testing confirms substantial improvements. The chip peaks at 250W versus the 14900K's 253W, yet typically runs at 240W while the 14900K often reaches 270-280W.

Thermal performance sees dramatic improvements. Using identical cooling, the Ultra 9 peaks at 86°C compared to the 14900K's concerning 100°C. This allows for quieter, more efficient operation under heavy workloads.

For the first time, Intel brings dedicated AI hardware to desktop processors. The Core Ultra 9 includes a 13 TOPS NPU (Neural Processing Unit) alongside AI-capable CPU cores and integrated graphics.

In practical terms, the NPU accelerates certain AI workloads significantly. The Procyon AI Vision benchmark shows the NPU scoring 373 points versus 144 points when using only CPU cores. This represents an improvement, yet falls short compared to discrete GPUs.

Currently, the NPU's primary benefit is handling background AI tasks efficiently while preserving CPU and GPU resources for primary workloads. Real-world applications remain limited, and the 13 TOPS capability doesn't meet Microsoft's 40+ TOPS requirement for Copilot+ features.

Core Ultra 9 costs £140 more than 14900K but runs cooler

The Core Ultra 9 285K requires careful consideration before purchase, as the price-performance equation varies significantly depending on specific needs, analysis suggests.

At £589, the Ultra 9 285K costs substantially more than the 14900K now available for £433-£445, creating a £140-£150 price gap. For gaming workloads, the previous generation outperforms its successor by 5-6% across a 45-game test suite. In certain titles, the performance difference widens significantly—the 14900K delivers 15fps more in Assassin's Creed Valhalla and 17fps more at QHD resolution in some games.

The Ultra 9 excels in productivity tasks, offering 4.5% better performance with 11% faster Blender rendering. Perhaps most impressive, it consumes 36.4% less power while running 11.9% cooler under load.

Against AMD's lineup, the Core Ultra 9 faces formidable competition. The 9800X3D delivers 24% better gaming performance on average, while the 9950X3D achieves nearly 35% higher average frame rates and 30% better 1% lows. In productivity benchmarks, the Ultra 9 285K trails the Ryzen 9 9950X by up to 23% in 7-zip compression yet maintains competitive multi-threaded performance overall.

Adopting the Core Ultra 9 necessitates investing in the new LGA1851 platform. This requires a Z890 motherboard, as the chip won't work with existing Z790 boards. Although Intel hasn't officially committed to LGA1851's longevity, indicators suggest it might support 2025 and 2026 releases.

Memory costs present another consideration—CUDIMM technology delivers optimal performance but at a premium price of £300+ for DDR5-8000 kits. A more balanced option is DDR5-7200 CL34 memory at approximately £120.

The Core Ultra 9 285K makes most sense for productivity-focused users prioritizing efficiency and thermal performance. It excels in content creation tasks, outperforming competitors in Cinebench, Corona, and Blender benchmarks. Its significantly improved power efficiency makes it ideal for workstation builds running high-load applications for extended periods.

Pure gaming enthusiasts should look elsewhere—either at the more affordable 14900K or AMD's 3D V-Cache offerings. Even Intel's own Core i7-14700K at £350-£375 presents better gaming value while being £280 cheaper than the Ultra 9.

Intel Core Ultra 9 falls behind predecessor in gaming tests

Intel's latest flagship processor delivers mixed results in benchmark testing, analysis suggests.

The Core Ultra 9 285K arrives at $589, maintaining the same price point as its predecessor while promising substantial performance improvements. Tests reveal the processor struggles to deliver consistent gains across all workloads despite its new chiplet architecture manufactured on TSMC's N3B process node.

Independent testing shows gaming performance declining compared to the previous generation Core i9-14900K. Some titles record up to 43% lower framerates with the new chip. Although the processor supports DDR5 memory with bandwidth reaching 121-123 gigabytes per second, increased memory latency of approximately 80 nanoseconds appears to hamper gaming performance significantly.

Intel claims the Core Ultra 9 285K delivers up to 150W reduction in system power during gaming and 20% improvement in threaded performance. However, the removal of Hyper-Threading has reduced the total thread count from 32 to 24 (8P+16E), despite Intel maintaining this architectural decision enhances multi-threaded workloads.

The processor operates at a maximum boost clock of 5.7GHz—300MHz lower than its predecessor's 6GHz—though Intel reports a 9% increase in instructions per clock. These contradictions between marketing claims and real-world benchmark results raise questions about the chip's value as an upgrade option for most users.

Core Ultra 9 architecture marks significant shift for Intel

Arrow Lake represents Intel's first chiplet-based design for desktop processors, analysis shows. The Core Ultra 9 285K abandons the monolithic approach used in previous generations.

The processor serves as Intel's flagship in the Core Ultra Series 2 lineup released in October 2024. Unlike previous chips, it's primarily manufactured by TSMC, with the compute die using TSMC's N3B 3nm process. This shift delivers up to 40% lower package power consumption compared to 14th Gen processors and introduces Intel's new Lion Cove P-cores with 9% higher IPC alongside Skymont E-cores with 32% higher integer IPC.

The Core Ultra 9 uses a disaggregated design with separate tiles for different functions. The compute tile houses CPU cores, the GPU tile contains graphics, and the SoC/I/O tiles handle memory controllers and connectivity. These components connect via Intel's Foveros 3D packaging technology. This approach reduces package size by 33% and improves manufacturing yields as each tile can be produced independently.

For the first time since 2002, Intel has removed Hyper-Threading from its performance cores. Consequently, the Core Ultra 9 285K features 24 cores and 24 threads (8P+16E), rather than the 32 threads of its predecessor. This decision reduced transistor count and power consumption but potentially impacts performance in memory-intensive workloads where SMT typically provides 5-25% improvement.

Intel completely rethought the core arrangement, placing E-core clusters between P-cores rather than segregating them. Four P-cores sit on the borders with another four in the middle, with four E-core clusters (containing four cores each) sandwiched between them. The E-cores now connect directly to the shared 36MB L3 cache, while P-cores gain larger 3MB L2 caches (up from 2MB), and each E-core cluster shares 4MB of L2 cache. This layout reduces thermal hotspots and improves inter-core communication latency.

Gaming performance shows unexpected regression

"While it put up impressive productivity and multi-core results, its gaming performance sank below Intel's previous-gen Raptor Lake Refresh at launch," says Michael Crider, technology journalist at Tom's Hardware.

Despite architectural innovations, the Core Ultra 9 285K struggles in gaming performance. Recent benchmarks reveal regression compared to Intel's previous generation flagship.

Gaming tests show the Core Ultra 9 underperforming its predecessor across multiple titles. The Ultra 9 trails the 14900K by 8% in Shadow of the Tomb Raider, 9% in Cyberpunk 2077, and 20% in Forza Motorsport. In Final Fantasy XIV: Dawntrail, the 14900K establishes a 15% lead over the 285K. Baldur's Gate 3 benchmarks revealed up to 43% performance regression compared to prior-generation chips when using the same memory kit.

The primary culprit behind this gaming performance slump appears to be increased memory latency. Intel's decision to split the memory controller and PHY into the I/O tile adds approximately 15-20 nanoseconds of memory latency. For gaming workloads that depend heavily on quick memory access, this architectural trade-off creates a significant bottleneck.

Intel's new CUDIMM technology offers some relief but doesn't completely solve the issue. With both chips using standard DDR5-7200, the 14900K remains 6% faster than the 285K. Even with high-end CUDIMM DDR5-8200 memory, the previous-gen 14900K still outperforms the 285K by 4%. Some games like Baldur's Gate 3 show dramatic improvements with CUDIMM technology, primarily by overcoming the latency hurdle.

Further complicating matters, performance varies significantly depending on BIOS settings and Windows versions. ASUS, ASRock, MSI, and Gigabyte boards showed inconsistent default BIOS settings. Windows 24H2 generally performed better with the Core Ultra 9, though certain games like Cyberpunk 2077 ran significantly better on Windows 11 build 2152.

Intel's Application Optimization (APO) was designed to improve performance by optimizing thread allocation. Initially supporting only Metro Exodus and Rainbow Six Siege, Intel has expanded compatibility to 26 titles. APO can boost framerates by up to 31% in specific scenarios, essentially ensuring P-cores focus on gaming while E-cores handle background tasks. Nevertheless, in its current state, APO provides minimal improvements without setting Windows to "high performance" mode.

Productivity and thermal efficiency show significant gains

"This boost is likely primarily due to optimizations dealing with Intel's recent big-core/small-core design philosophy, using a mixture of performance and efficiency cores to balance out workloads," says Michael Crider, technology journalist at Tom's Hardware.

While gaming performance falters, the Core Ultra 9 285K excels in productivity tasks, showing the architectural trade-offs Intel made with Arrow Lake.

Creative professionals will appreciate the Core Ultra 9's substantial gains over previous generations. In Cinebench 2024, the 285K scores 17% higher than the 14900K in multithreaded tests and outperforms the Ryzen 9 9950X by 11%. Similarly, Blender benchmarks show the Ultra 9 rendering 15% faster in the Monster workload, 19% faster in Classroom, and 5% faster in Junkshop compared to its predecessor.

Adobe apps particularly benefit from this architecture. The Core Ultra 9 delivers 6% better performance in Adobe Premiere versus the 14900K and scores 16,231 points compared to the 14900K's 15,179. Photo editing sees up to 21% improvement over 12th-gen chips.

Perhaps most impressive is how the Core Ultra 9 285K achieves these gains while improving efficiency. Intel claimed a 40% reduction in package power consumption, and testing confirms substantial improvements. The chip peaks at 250W versus the 14900K's 253W, yet typically runs at 240W while the 14900K often reaches 270-280W.

Thermal performance sees dramatic improvements. Using identical cooling, the Ultra 9 peaks at 86°C compared to the 14900K's concerning 100°C. This allows for quieter, more efficient operation under heavy workloads.

For the first time, Intel brings dedicated AI hardware to desktop processors. The Core Ultra 9 includes a 13 TOPS NPU (Neural Processing Unit) alongside AI-capable CPU cores and integrated graphics.

In practical terms, the NPU accelerates certain AI workloads significantly. The Procyon AI Vision benchmark shows the NPU scoring 373 points versus 144 points when using only CPU cores. This represents an improvement, yet falls short compared to discrete GPUs.

Currently, the NPU's primary benefit is handling background AI tasks efficiently while preserving CPU and GPU resources for primary workloads. Real-world applications remain limited, and the 13 TOPS capability doesn't meet Microsoft's 40+ TOPS requirement for Copilot+ features.

Upgrade value depends on specific workflow needs

Deciding whether to invest in the Core Ultra 9 285K requires careful evaluation of its value proposition compared to alternatives. The price-performance equation differs markedly depending on specific needs.

The Core Ultra 9 285K costs approximately $589, whereas the 14900K has dropped to around $433-$445, creating a substantial $140-$150 price gap. For gaming workloads, the previous generation outperforms its successor by 5-6% across a 45-game test suite. In specific titles, the performance gap widens significantly—the 14900K delivers 15fps more in Assassin's Creed Valhalla and 17fps more at QHD resolution in some games.

Conversely, the Ultra 9 shines in productivity tasks, offering approximately 4.5% better performance with 11% faster Blender rendering. Most impressively, it consumes 36.4% less power while running 11.9% cooler under load.

Against the formidable AMD lineup, the Core Ultra 9 faces tough competition. The 9800X3D delivers 24% better gaming performance on average, whereas the 9950X3D achieves nearly 35% higher average frame rates and 30% better 1% lows. In productivity benchmarks, the Ultra 9 285K trails the Ryzen 9 9950X by up to 23% in 7-zip compression yet maintains competitive multi-threaded performance overall.

Adopting the Core Ultra 9 285K necessitates investing in the new LGA1851 platform. This requires a Z890 motherboard, as the chip won't work with existing Z790 boards. Although Intel hasn't officially committed to LGA1851's longevity, indicators suggest it might support 2025 and 2026 releases.

Memory costs present another consideration—CUDIMM technology delivers optimal performance but at a premium price of $300+ for DDR5-8000 kits. A more balanced option is DDR5-7200 CL34 memory at approximately $120.

The Core Ultra 9 285K makes most sense for productivity-focused users prioritizing efficiency and thermal performance. It excels in content creation tasks, outperforming competitors in Cinebench, Corona, and Blender benchmarks. Its significantly improved power efficiency makes it ideal for workstation builds running high-load applications for extended periods.

Notwithstanding these strengths, pure gaming enthusiasts should look elsewhere—either at the more affordable 14900K or AMD's 3D V-Cache offerings. Even Intel's own Core i7-14700K at $350-$375 presents better gaming value while being $280 cheaper than the Ultra 9.

Core Ultra 9 shows promise despite first-generation limitations

The Core Ultra 9 285K represents both a technological shift and strategic pivot for Intel, analysis suggests. The transition to a chiplet-based architecture brings substantial improvements in productivity performance and power efficiency, though gaming results disappoint compared to the 14900K.

Content creators, engineers, and professionals handling CPU-intensive tasks will appreciate the significant efficiency gains and cooler operation. Dedicated gamers might find better value in either the previous generation 14900K or AMD's gaming-focused X3D alternatives.

Platform costs require careful consideration. The mandatory upgrade to an LGA1851 motherboard coupled with potentially expensive CUDIMM memory creates a steeper entry price than the raw $589 CPU cost suggests. Nevertheless, the platform potentially offers future upgrade paths if Intel maintains socket compatibility with upcoming generations.

The Core Ultra 9 285K serves as a glimpse into Intel's chiplet-based future rather than an unequivocal recommendation for all users. This first attempt at desktop chiplet architecture shows promise in certain areas while revealing opportunities for improvement in others—particularly gaming latency.

Processors increasingly serve specialized niches rather than excelling universally. The ideal choice depends less on benchmark rankings and more on aligning with specific workflow requirements and budget constraints.

FAQs

Q1. Is the Intel Core Ultra 9 285K good for gaming? While the Core Ultra 9 285K performs well in 4K gaming, it may not offer significant advantages over previous generations for 1080p or 1440p gaming. Its strengths lie more in productivity tasks and power efficiency.

Q2. How does the Core Ultra 9 285K compare to previous Intel generations? The Core Ultra 9 285K offers improved power efficiency and cooler operation compared to previous generations. It excels in multi-threaded workloads but may have slightly lower single-threaded performance in some scenarios.

Q3. What are the main benefits of the Core Ultra 9 285K? Key benefits include significantly reduced power consumption, cooler operation, strong multi-core performance for productivity tasks, and support for high-speed DDR5 memory, including CUDIMM technology.

Q4. Does the Core Ultra 9 285K require special cooling? While the Core Ultra 9 285K runs cooler than previous high-end Intel processors, it still benefits from good cooling solutions. However, extreme cooling measures like delidding are generally not necessary for stock operation.

Q5. Is upgrading to the Core Ultra 9 285K worth it? The value of upgrading depends on your specific needs. It's most beneficial for users focused on productivity tasks, those seeking improved power efficiency, or building new systems. Gamers with recent high-end CPUs may see less benefit.