New direction for AI PCs, and Intel Firstpost

Pros:
– Massive gains in efficiency
– Comes with a lot of headroom if you want to overclock
– Manages heat well
– Strong multithread performance, despite no hyperthreading
– First desktop CPU to have an NPU
– Solid performance gains in creative workloads
– New chiplet-based architecture that has a lot of potential
– Comes with much-improved iGPU

Cons:
– New socket means investing in new motherboard & RAM
– Needs top-of the line, high-speed RAM to get the most out of it
– No major improvement in gaming performance. Dip in some titles
– NPU is under-utilised in Windows

Rating: 4.25/5

With the launch of their new line of CPUs, Intel is venturing into something new here. Not only are they going for a new architecture, they are also going for a whole new design philosophy. Moreover, they have done away with hyperthreading, the core technology that made their CPUs unbelievably fast especially in multithreaded workloads. Among their new line of CPUs, we first test the Core Ultra 9 285K a processor unlike any that I’ve seen before.

It marks a bold shift for Intel, reimagining everything from its design to its features with a clear focus on creating efficient desktop CPUs. Compared to what we’ve had for years, this chip accomplishes some pretty amazing feats that feel almost like a leap into the future, especially when you consider the Core i9-14900K, Intel’s last flagship-level desktop CPU.

The Core Ultra 9 285K is impressive in its own way. Intel has not only ramped up efficiency but also made strides in performance, and it looks like they’ve addressed previous stability concerns

Intel has revamped also its Core series to align with the naming conventions used for its mobile CPUs. What makes things really interesting though, is like those mobile chips, the Core Ultra 9 285K includes a dedicated NPU capable of 13 TOPS, a first for a desktop CPU.

The real question is whether that dedicated NPU is truly necessary for a gaming rig equipped with a graphics card, especially given Windows’ behaviour. Nevertheless, the Core Ultra 9 285K stands out as a worthy upgrade from the Core i9-14900K, particularly excelling in creative tasks and benchmarks. However, when it comes to gaming performance, that’s where things start to get intriguing.

Intel Core Ultra 9 285K Review: Specifications and features
Intel’s new Arrow Lake architecture is shaking things up with some cool changes, starting with its design. Like AMD, Intel has moved to a chiplet-based approach, but they’ve made it a bit more efficient and complex. At the heart of it all, you’ve got a foundational package, then layered on top are different tiles: a base tile, compute tile, I/O tile, SoC tile, GPU tile, and a filler tile. Each tile uses different manufacturing tech, with the compute tile being made on TSMC’s 3-nanometer N3B process, the graphics tile on TSMC’s 5nm N5P process, and the SoC and I/O tiles on TSMC’s 5nm N6 process.

After dealing with power and heat issues in the last couple of Raptor Lake generations, these upgrades were super important for Arrow Lake. Intel claims they’ve improved performance per watt by reducing package power consumption by 40% compared to the 14th Gen Raptor Lake Refresh processors, thanks to this new design.

The Core Ultra 9 285K features 8 Performance Cores and 16 Efficiency Cores, totaling 24 CPU cores. Unlike previous models, Intel has disabled Hyper-Threading, so all these cores are single-threaded. In the past, Intel’s desktop chips had P-Cores with Hyper-Threading, giving two threads per core, but thanks to better communication between cores, multi-core performance hasn’t taken a hit.

Intel has also slightly lowered the clock speeds. The Core Ultra 9 285K hits a max Turbo Boost of 5.7GHz, down from 6GHz in the Core i9-14900K. This means single-core performance won’t see a huge jump this time, but 5.7GHz is still lightning fast. The total cache has also increased, with the L3 cache staying at 36MB while the L2 cache bumps up from 32MB to 40MB.

A neat new addition is a 13 TOPS NPU (Neural Processing Unit) built into the Core Ultra 9 285K. This is the same NPU used in Intel’s 2023 Meteor Lake laptop processors, so it’s not the super-fast Lunar Lake NPU, but it still boosts AI performance compared to processors without one.

Intel also wanted to tackle power consumption and heat after a couple of hot generations. Power use is pretty much the same, with the Core Ultra 9 285K peaking at 250W, slightly lower than the 253W of the Core i9-14900K. However, temps have improved. Using the same cooler, the Core Ultra 9 285K peaked at 86°C — still warm, but much better than the 100°C of the Core i9-14900K.

On the graphics side, Intel has upgraded its integrated graphics hardware for Arrow Lake, using the Xe LPG architecture. This is closely related to what you find in Intel’s Arc graphics cards but reworked for integrated use. The shader count has doubled; where Raptor Lake’s Intel UHD Graphics 770 had 32 execution units (EUs) and 256 shaders, Arrow Lake processors now have 64 EUs and 512 shaders. These shaders can hit speeds up to 2GHz and come with dedicated AI and ray-tracing hardware.

Finally, the host processor and integrated graphics benefit from an improved memory controller that supports DDR5 RAM clocked at 6,400MT/s. With these upgrades and faster clock speeds, Arrow Lake processors should really outperform their predecessors when it comes to graphics performance.

Intel Core Ultra 9 285K Review: Compatibility & new chipset
Intel’s new Arrow Lake CPUs use the LGA 1851 socket, which means they aren’t compatible with any previous-gen motherboards. At launch, the new socket only supports one chipset: the Z890.

The Z890 chipset offers 20 PCIe 5.0 lanes (with x16 for expansion cards and x4 for storage) plus an extra 4 PCIe 4.0 lanes for storage. A cool upgrade is that the PCIe lanes for expansion cards can now be split among three devices (x8, x4, x4), compared to just two with the LGA 1700.

Like AMD’s AM5 socket, the Arrow Lake architecture is all about DDR5 SDRAM, dropping support for the older DDR4. So, if you’re upgrading to the Core Ultra 9 285K, you’ll need to pick up some new RAM. The maximum RAM capacity remains at 192GB, and while these chips don’t support DDR4, they can handle faster DDR5 memory up to 6,400MT/s right now. The chipset can theoretically support modules up to 9,200MT/s, which is a nice step up from the previous generation’s 7,200MT/s.

As for coolers, the new socket is compatible with existing ones and uses the same mounting hardware. However, since the hotspot on the die has shifted slightly toward the top, some cooler manufacturers might recommend a new mounting kit. If you’re using a good liquid or AIO cooler with a well-designed water block, you should be just fine—unless you’re going all out with overclocking.

The new Intel Z890 chipset is designed for the upcoming Intel Core Ultra processors and comes packed with impressive features to boost performance and connectivity for high-end desktop systems.

One standout is native Thunderbolt 4 support, ensuring lightning-fast data transfer and versatile connectivity for your peripherals. Plus, it includes support for Wi-Fi 6E, with motherboard manufacturers also offering Wi-Fi 7 options for even faster wireless internet and improved network performance.

Storage options have also gotten an upgrade, with PCIe 5.0 support for M.2 slots to facilitate faster SSD performance. The chipset boasts a total of 60 HSIO lanes (High-Speed I/O), providing plenty of expansion options for graphics cards and storage devices. You can also take advantage of Intel’s Rapid Storage Technology, which speeds up data access and supports RAID for data redundancy.

Intel has integrated a Neural Processing Unit (NPU) within select Core Ultra processors, significantly enhancing AI and machine learning tasks. The Z890 chipset supports various overclocking features, allowing users to squeeze out even more performance from their CPUs and memory.

Overall, the Z890 chipset is a solid choice for anyone looking to build a powerful gaming rig or workstation that can handle intense tasks, making it a worthy addition to Intel’s lineup of chipsets.

Intel Core Ultra 9 285K Review: Our test bench
For our testing setup, we used the Intel Core Ultra 9 285K paired with the Gigabyte Z890 AORUS Elite WiFi7 Gaming Motherboard, kindly provided by Gigabyte. To keep things cool, we opted for the AORUS Waterforce X II 360 AIO cooler, ensuring we squeezed every bit of performance from the chip. For memory, we went with a 96GB setup using two 48GB G.Skill Trident Z5 sticks rated at DDR5 6400MT/s.

For Graphics, we had the Gigabyte GeForce RTX 4080 Super, which boasts 16GB of GDDR6X memory. Everything was powered by an AORUS Elite 1000W Platinum modular power supply, featuring an 80 Plus Platinum rating.

Like other K-series Intel Arrow Lake CPUs, the 285K is ready for overclocking right out of the box, but we decided to keep things simple and tested it at factory settings. We did enable XMP from the BIOS to ensure the RAM performed at its peak.

Intel is banking on delivering the best performance at the highest thermal efficiency with its Core Ultra 200 series. Therefore, Intel’s new Adaptive Performance Optimization (APO) technology, which is designed to enhance power management and performance tuning across various workloads, becomes important. We therefore also had APO enabled. APO usually comes enabled straight out of the box, as it enables the processors to intelligently adjust power consumption, and optimise performance for both gaming and productivity tasks.

As for our choice of the motherboard, we chose the Z890 AORUS Elite WiFi7 for several reasons. Its AORUS AI SNATCH and AI SNATCH Engine features allow for one-click AI-driven overclocking, optimizing performance easily. And with a robust 16+1+2 phase VRM setup, stability is guaranteed.

On top of that, the motherboard is super user-friendly. Features like WiFi EZ-Plug simplify antenna connections, and the EZ-Debug Zone with diagnostic LEDs makes troubleshooting easier. There’s also a PCIe EZ-Latch Plus button for hassle-free GPU removal, and all four M.2 slots are tool-free, making the build process smooth and straightforward.

Intel Core Ultra 9 285K Review: Performance
With the Core 200 Desktop CPUs, Intel is focusing everything on efficiency and performance per watt — they’re aiming to deliver more performance while using less power and generating less heat compared to the previous 14th Gen chips and for the most part, they’ve hit the mark,

When it comes to gaming, things get a bit tricky, but more on that later.

A key feature of the Core Ultra 9 285K is its 13 TOPS NPU, but chances are you’ll pair this CPU with a powerful graphics card like the Nvidia GeForce RTX 4080 or a 4090. In those cases, the NPU might not seem all that useful.

Having said that, it does give a nice boost to AI performance compared to just using the CPU.

For example, in the Procyon AI Vision benchmark, the Core Ultra 9 285K scored 144 points when only the CPU was tested. But when the NPU comes into play, that score jumps to 373 points.

While the NPU can speed up AI tasks, it doesn’t hold a candle to what a dedicated graphics card can do. Intel claims that the NPU is there to handle background AI tasks while the GPU is busy, but honestly, there aren’t many tasks right now that can really take advantage of it.

When it comes to synthetic CPU benchmarks, the Core Ultra 9 285K holds its own. In Geekbench the Core Ultra 9 outperforms the 14900K in single-core performance only to lose out in Cinebench R23’s single-core benchmark. Meanwhile, in Cinebench R23’s multi-core performance, the Core Ultra 9 285K comes in faster than the 14900K. We see somewhat of a similar story in Cinebench 2024. However, in Geekbench 6, the Ultra 9 285K is outscored on the multi-core front.

Surprisingly, Pugetbench does not sit well with Ultra 9 285K as it was handily beaten by the 14900K in both, Photoshop and Premiere. Maybe the next Adobe update will start utilising everything that the Ultra 9 285K has to offer.

Even with some less impressive results, the Core Ultra 9 285K still shines in multi-core performance.

However, when it comes to gaming, the Core Ultra 9 285K takes a slight hit compared to the 14900K. The performance isn’t terrible, but it’s not topping the charts like some of the previous chips. Intel had stated that with the new CPUs, they are not looking to make massive jumps in gaming performance. However, seeing a dip, as marginal as it may be, can be a bit flustering for some.

Don’t get me wrong, the numbers put up by the Core Ultra 9 285K aren’t disappointing or bad in any way. Intel is delivering exactly what they had said. But, why do we not see that big a jump in gaming performance with Arrow Lake? There are a few factors at play, and it’s not just one thing causing the drop in performance.

For starters, the Core Ultra 9 285K has a lower maximum P-core clock speed (5.7 GHz vs. 6.0 GHz), and the ring bus connecting the cores to the shared L3 RAM is slower too (3.8 GHz versus up to 5.0 GHz). This lower clock speed likely explains some of the gaming performance dips, but it also helps with the improved energy efficiency we’re seeing in Arrow Lake.

Plus, the P-cores don’t have hyperthreading anymore; the upgraded E-cores are now taking over the tasks that the old P-core hyperthreading used to manage.

There are a lot of changes happening under the hood with Arrow Lake, and figuring out exactly why there’s a performance gap will take some digging.

To be fair to Intel, they’ve been upfront about their focus on energy efficiency with these Arrow Lake Desktop CPUs. The big question is: have they delivered on that promise?

Intel Core Ultra 9 285K Review: Power consumption & thermals
Short answer is, yes. During our tests, the Intel Core Ultra 9 285K typically peaked around 240 watts, while the i9-14900K could hit between 270 and 280 watts at times.

When it comes to temperatures, the 14900K could reach as high as 95 degrees Celsius under peak workloads, usually settling around the 90 to 92-degree mark during sustained use. In contrast, the Intel Core Ultra 9 285K managed to stay cooler, rarely exceeding 86 degrees and often settling between 82 and 84 degrees during our testing.

Given that the Intel Core Ultra 9 285K is built on a completely new design and approach to chip manufacturing, the efficiency gains for a first-generation product are quite impressive. That said, there’s still plenty of room for improvement, especially regarding the power draw.

Intel Core Ultra 9 285K Review: Verdict
There is no doubt about the fact that the Core Ultra 9 285K is a mixed bag — we gain some noticeable performance in certain areas, and take a small hit in others, especially when we compare it to the outgoing set of CPUs, both from AMD and Intel.

What I am really impressed by is the significant efficiency gains that Intel has made and the clever way Intel has managed to pack in 24 threads, that too when giving up on hyperthreading.

In my mind, this can be Intel’s Ryzen moment — when AMD moved away from their FX series processors to a whole new architecture and went on to mount an impressive resurgence.

But for that to happen Intel has to build on the foundation that they have laid with the Core Ultra 9 285K. It may not sell in the volumes that Intel wants it to, or in fact even needs it to, but it sure is a start of something fresh, something new.

But as much as I am rooting for Intel, the fact of the matter is, upgrading to this generation of Intel CPUs, only makes sense if you have an Intel 11th or 12th Gen system or perhaps even older, and for some reason would not want to go for something from AMD. If you’re a gamer, the Ryzen 7 7800X3D is still king, thanks to its 3D V-Cache

For everyone else, the only way the new cost of the platform and the new CPUs make sense is that you run your PC throughout the day, and really need to cut down on your power bills.