Intel Core Ultra 9 285K review: impressive, not necessarily good

Intel's Core Ultra 9 285K is an undeniably impressive CPU, but it's still tough to recommend given some inconsistent performance.

Intel Core Ultra 9 285K review: impressive, not necessarily good

intel core ultra 9 285k review 1

Intel Core Ultra 9 285K

MSRP $589.00

“Intel is finding its footing with the Core Ultra 9 285K — for better and worse.”

Pros

Insane efficiency improvements Delivers multi-threaded performance gains without Hyper-Threading Runs very cool in games A lot of headroom for CPU and memory overclocking

Cons

Gaming performance struggles Inconsistent productivity performance High-end memory required for best experience

I’ve never seen a processor as radically different as the Core Ultra 9 285K. It’s a complete departure for Intel, from design to features, all focused on a new vision of highly efficient desktop CPUs. And it actually delivers on that vision. Even compared to the best processors, the Core Ultra 9 285K pulls off things that seem like magic compared to the CPUs we’ve seen for the past decade or more.

Calling the Core Ultra 9 285K anything but impressive is seriously selling it short. Intel has improved efficiency, it’s delivered performance improvements, and it seems to have gotten its instability struggles under control.

Get your weekly teardown of the tech behind PC gaming

But you’ve read the headline; you’ve seen the score. As impressive as the Core Ultra 9 285K is, it ends up being a tough CPU to actually recommend you go out and buy.

Intel Core Ultra 200S at a glance

The back of the Core Ultra 9 285K CPU.Jacob Roach / Digital Trends

At first glance, the lineup of Core Ultra 200S processors — formerly known as Arrow Lake — doesn’t look much different than 12th-, 13th-, or 14th-gen CPUs. Core counts are the same, with the same split between performance (P) and efficient (E) cores, and the processors have the same rated power draw. Clock speeds have shifted slightly down, but not by much. The specs really sell the story short here, however.

The Core Ultra 9 285K, and the entire range of chips, is a radical departure. First and foremost, Intel killed Hyper-Threading here. One core now equals just one thread, putting to rest a practice of placing two threads on each core that Intel (and AMD) has maintained for decades. That immediately puts the Core Ultra 9 285K at a thread disadvantage compared to the Ryzen 9 9950X, meaning it needs more powerful individual cores to make up the thread disparity in heavily threaded workloads.

And indeed, these individual cores are more powerful. The Core Ultra 9 285K still has P-cores and E-cores, but unlike the Core i9-12900K, which debuted Intel’s hybrid architecture, the P-cores aren’t the main drivers of performance — the E-cores are. That’s a strange shift. The E-cores drive the bedrock performance, while the P-cores step in for intense workloads that need a lot of speed on, say, one or two threads. This architectural change brings a lot of improvements to efficiency, as I’ll dig into with my benchmarks below.

Cores (P+E) / Threads Base clock (P/E) Boost clock (P/E) Total cache (L2+Smart cache) Power (Base/Max)
Core Ultra 9 285K 24 (8+16) / 24 3.7GHz / 3.2GHz 5.7GHz / 4.6GHz 60MB 125W/250W
Core Ultra 7 265K/KF 20 (8+12) / 20 3.9GHz / 3.3GHz 5.5GHz / 4.6GHz 50MB 125W/250W
Core Ultra 5 245K/KF 14 (6+8) / 14 4.2GHz / 3.6GHz 5.2GHz / 4.6GHz 38MB 125W/159W

The setup here is actually identical to Intel’s mobile Lunar Lake CPUs, just juiced up with more cores, higher clock speeds, more cache, and higher power limits. The E-cores use the Skymont design, which Intel says comes with a 2% improvement in IPC (instructions per clock) over the Raptor Cove P-core design featured in 13th-gen and 14th-gen chips. The P-cores use the Lion Cove design. Based on the crossover, you can think of Arrow Lake as using P-cores and P+-cores, or something along those lines. Skymont is the main performance driver, rather than some weaker additional cores that can step in for heavily threaded workloads.

Another big departure comes in the manufacturing process. This is the first desktop CPU that Intel didn’t manufacture itself. The Core Ultra 9 285K is built on TSMC’s N3B node, as Intel ditched its plans to build on the 20A node it announced years ago.

Finally, the Core Ultra 200S range comes with an on-board 13 TOPS NPU, though I suspect it won’t be super relevant in a desktop for at least a few years, particularly if you already have a discrete GPU.

Test configurations and a note on compatibility

The Asus ROG Maximus Hero Z890 motherboard sitting on a table.Jacob Roach / Digital Trends

As with all of my CPU reviews, I stuck with almost identical test benches across Arrow Lake, Zen 5 and Zen 4, and Intel’s previous-gen Raptor Lake processors, only swapping out the CPU and motherboard between the various platforms. I’m using an RTX 4080 and a kit of DDR5-6000 memory to isolate the performance of the processor as much as possible. However, Arrow Lake chips support much higher DDR5 speeds than Raptor Lake processors, which are now possible thanks to CUDIMMs. I’ve included a few tests below with faster memory to show the differences.

For the Arrow Lake bench, I used the high-end Asus ROG Maximus Z890 Hero motherboard, which is absolutely packed to the brim with features. I actually tested two different motherboards — the aforementioned Asus board and the MSI MEG Z890 Unify-X. I ran into several issues with getting the platform to boot properly across both motherboards, but the Asus board is what eventually got the test bench stable. I don’t suspect most people will have these problems, as it was some combination of a dying SSD and pre-release BIOS versions, but it’s important to note nonetheless.

  Intel LGA 1851 Intel LGA 1700 AMD AM5
GPU Nvidia RTX 4080 Founders Edition Nvidia RTX 4080 Founders Edition Nvidia RTX 4080 Founders Edition
RAM 32GB Gigabyte Aorus DDR5-6000 32GB Gigabyte Aorus DDR5-6000 32GB Gigabyte Aorus DDR5-6000
Motherboard Asus Z890 ROG Maximus Hero MSI Z790 Tomahawk Wi-Fi Gigabyte X670E Aorus Master
CPU cooler MSI CoreLiquid i360 MSI CoreLiquid i360 MSI CoreLiquid i360
Power supply Gigabyte Aorus P1200W Gigabyte Aorus P1200W Gigabyte Aorus P1200W
Storage Boot: Samsung 990 Pro 2TB / Tests: MSI M450 1TB Boot: Samsung 990 Pro 2TB / Tests: MSI M450 1TB Boot: Samsung 990 Pro 2TB / Tests: MSI M450 1TB

Regardless, the Core Ultra 9 285K introduces the new LGA 1851 socket. Physically, it’s the same size as the LGA 1700 socket that Intel used on 12th- to 14th-gen CPUs, but the Core Ultra 200S range isn’t backwards compatible. Thankfully, existing LGA 1700 coolers will fit on the new LGA 1851 socket without issue. The MSI MAG CoreLiquid i360 cooler I used came with an offset bracket specifically for LGA 1851.

As previously reported on, the hotspot location on Arrow Lake chips is higher up on the processor, so an offset cooler like the one I used targets the hottest area more directly. Any LGA 1700 cooler will work, but you’ll want to check with the brand of your cooler to see if an offset bracket is available to get the best performance.

Productivity performance

Performance for the Core Ultra 9 285K in Cinebench R24.Jacob Roach / Digital Trends

I said up top that the Core Ultra 9 285K was impressive, and I stand by that. Look no further than Cinebench R24 for proof. This is the gut check of CPU reviews — the first test you run to make sure everything is operating the way it should — and Intel wasn’t messing around. Arrow Lake is fast. 

In fact, it’s 15% faster than the Ryzen 9 9950X and 20% faster than the Core i9-14900K in multi-core performance. And remember, the Core Ultra 9 285K has access to 24 threads, not the 32 available to the other two chips.

Performance of the Core Ultra 9 285K in Geekbench 6.Jacob Roach / Digital Trends

Intel specifically said that the Core Ultra 9 285K shows big gains in a ray-traced renderer like Cinebench, and sure enough, Geekbench 6 flattens the differences between the CPUs a bit. Here, we’re actually seeing a slight reduction in single-core speed compared to the Ryzen 9 9950X and Core i9-14900K, but it’s not huge. And even with less impressive results overall, the Core Ultra 9 285K still lands at the top of multi-core performance — and even more so if you have a speedy kit of memory to pair the CPU with.

Performance of the Core Ultra 9 285K in Handbrake.Jacob Roach / Digital Trends

Cinebench and Geekbench produce the most impressive results for the Core Ultra 9 285K, but as you start digging into other benchmarks, the performance ranges from competitive to downright disappointing.

At the competitive end, we have transcoding in Handbrake, where the Core Ultra 9 285K was just a touch slower than the Ryzen 9 9950X in an x265 video transcode. It’s technically a loss, but I’m willing to call this identical performance considering the generational improvement on display.

Performance of the Core Ultra 9 285K in Blender.Jacob Roach / Digital Trends

The Ryzen 9 9950X is really the toughest competition for the Core Ultra 9 285K, too. As you can see in Blender, the generational improvements in rendering applications are certainly present — there’s a huge uplift over the Core i9-14900K. However, AMD’s new Ryzen 9 9950X still comes out on top, and in some of the rendering tests, by quite a significant margin.

Performance of the Core Ultra 9 285K in Y-Cruncher.Jacob Roach / Digital Trends

Another point against the Core Ultra 9 285K is the lack of AVX-512 support. This isn’t relevant to the vast majority of users, but if you have an application that can leverage AVX-512 instructions, the performance uplift with native support on AMD’s latest Zen 5 CPUs is quite dramatic. Y-Cruncher is a great example of that. Once again, Intel is showing a massive generational improvement, but it’s really no contest when you bring the 512-bit data path available to Zen 5 CPUs into the fold.

Short of the limited applications of AVX-512, this is really impressive performance. So, why is the processor so slow in Adobe Premiere Pro and Photoshop? Premiere Pro is bad enough, with the Core Ultra 9 285K falling short of every CPU I tested short of the 12-core Ryzen 9 9900X. Photoshop is somehow even worse, with the Core Ultra 9 285K showing the lowest result out of the test suite. It’s not good when the Core i9-13900K and Ryzen 9 7950X — two older components — are offering clearly better performance.

I hear you — Jake, something must be wrong with your testing. I assumed the same thing. I tried two different motherboards, different memory kits, overclocked and stock settings, and no power limits. I even tested two different Core Ultra 9 285K models, and I shared this data directly with Intel. This is all fresh data, too — all of the results you see in this review were collected within days of each other. And yet, test after test, variable change after variable change, the same results showed up time and again.

You would assume that the performance advantages across all of my other tests would translate into apps like Premiere Pro and Photoshop, but they just don’t. And that’s largely been my experience with the Core Ultra 9 285K. There are a lot of impressive things Intel is doing here, some of which has big performance implications. But when the rubber meets the road, the Core Ultra 9 285K isn’t strictly faster than its last-gen counterparts. There’s no place that’s more clear than gaming.

Gaming performance

I didn’t have high expectations for gaming performance on the Core Ultra 9 285K. Intel’s stated goal with the processor was to simply maintain gaming performance, not improve it. Based on 10 games I tested, the Core Ultra 9 285K is on-par to slightly slower than the last-gen Core i9-14900K, and it gets absolutely crushed by AMD’s Ryzen 7 7800X3D.

Let’s get some easy games out of the way first. Most games are limited by your graphics card, even at 1080p, and that limiting factor only becomes more pronounced as you climb to higher resolutions and graphics settings. At 1080p High settings, there is very little movement in Returnal, Assassin’s Creed Mirage, and Black Myth: Wukong. There are differences, with the Core i9-14900K beating the Core Ultra 9 285K in Mirage and Black Myth, but we’re talking about a few frames at most. This is identical performance, at least for the gen-on-gen comparison.

Performance of the Core Ultra 9 285K in Red Dead Redemption 2.Jacob Roach / Digital Trends

Black Myth: Wukong centers in on the trend throughout my gaming tests, though — it’s really futile to compare gaming performance with the Ryzen 7 7800X3D around. It’s just the fastest gaming CPU you can buy, and in most games, it’s not even close. Look at Red Dead Redemption 2 as an example. This is a game I would normally toss into the graphically-intensive bucket, and you can see a fairly flat line in performance across the CPUs I tested — except for the Ryzen 7 7800X3D.

Nevermind the fact that the Core Ultra 9 285K is beaten by its last-gen counter-part and the Ryzen 9 9950X. If you want good gaming performance, none of these CPUs really make sense.

Performance of the Core Ultra 9 285K in F1 22.Jacob Roach / Digital Trends

Things are a little more interesting elsewhere. The Ryzen 7 7800X3D still leads the pack, unsurprisingly, but the Core Ultra 9 285K showed some odd performance. In F1 2022, the processor was slower than both the Core i9-13900K and Core i9-14900K. However, when paired with fast DDR5 memory, it actually showed a slight lead.

Performance of the Core Ultra 9 285K in Final Fantasy 14.Jacob Roach / Digital Trends

Similarly, in Final Fantasy XIV Dawntrail, the Core Ultra 9 285K jumped from a slight loss to the Core i9-14900K to a slight lead with fast memory. In both F1 2022 and Final Fantasy XIV, AMD’s Ryzen 9 9950X is faster, and the Ryzen 7 7800X3D tops the charts, so although it’s an interesting exercise to look at memory scalability on the Core Ultra 9 285K, it doesn’t make much of a difference when it comes to making a buying decision.

Performance of the Core Ultra 9 285K in Hitman 3.Jacob Roach / Digital Trends

The only game where the Core Ultra 9 285K topped the charts was Hitman 3, though you could call it a win by technicality considering how close the top three chips are — and the fact that the Core Ultra 9 285K needed much faster memory to even beat its last-gen counterpart.

There are some gen-on-gen improvements, which you can see in Ashes of the Singularity and Tiny Tina’s Wonderlands. The AMD competition is really the problem here; by simply maintaining gaming performance, Intel has essentially conceded gaming performance wholesale to AMD this generation.

Performance of the Core Ultra 9 285K in Cyberpunk 2077.Jacob Roach / Digital Trends

To show all of my work, you can see results for Cyberpunk 2077 above. The trend is pretty clear here regardless, but it’s worth showing all of the games I tested — it took me a long time, you know?

Temperatures and power

Jacob Roach / Digital Trends

Intel may not have improved gaming performance on the Core Ultra 9 285K, but its goal was to improve efficiency. And true to its word, there are some big improvements in temperature and power draw. The chart above really tells the whole story here. While delivering similar performance, the Core Ultra 9 285K is often drawing 50% less power, and sometimes it’s even more efficient. I saw the numbers fly by on screen during benchmarking, but I’m still taken back seeing them put up next to each other on a graph.

But that pesky Ryzen 7 7800X3D. Impressive as Intel’s improvements in efficiency are here, the Ryzen 7 7800X3D is delivering much higher performance at an even lower power draw. It’s another example of the disconnect with the Core Ultra 9 285K. You can see that Intel pulled off a lot of impressive things, but it rarely shifts the buying decision.

Jacob Roach / Digital Trends

Although the Core Ultra 9 285K consumed more power, it ran cooler overall. It doesn’t seem like much here, but it’s important to remember that I was benchmarking with a 360mm all-in-one liquid cooler that’s more than capable of keeping these CPUs cool. Scale down to a small form factor PC, and the temperature advantages of the Core Ultra 9 285K could make a big difference.

Jacob Roach / Digital Trends

Gaming is really where the big efficiency improvements are. As you can see from the chart above, the Core Ultra 9 285K is more efficient in heavy workloads like Cinebench and Handbrake, though it’s still totally comfortable going above 200 watts when it needs to. What’s great here is that the Core Ultra 9 285K is providing a clear performance improvement while drawing lower power — unlike gaming, you’re not trading performance for better efficiency.

Jacob Roach / Digital Trends

The translation into temperatures isn’t as stark in productivity applications, but once again, those differences would become more pronounced the more thermally constrained the environment is.

What about instability?

The LGA 1851 socket on a motherboard.Jacob Roach / Digital Trends

It’s impossible to talk about the Core Ultra 9 285K outside of the context of Intel’s recent desktop struggles. The Core i9-14900K and Core i9-13900K faced dire instability problems, which Intel has only recently solved. Will we see the same issues with the Core Ultra 9 285K? Intel says no, but I don’t have a crystal ball.

Here’s what I can tell you — Intel is clearly on top of any instability concerns with the Core Ultra 9 285K. One glance in the BIOS of a Z890 motherboard, and it becomes abundantly clear that Intel doesn’t want to repeat past mistakes. Both motherboards I looked at were running with default power settings, and they were both clearly labeled to show that they were Intel’s default settings.

In addition, the company let me know that it’s opened some more strict lines of communication with its motherboard partners moving forward. That should mean more consistency across different motherboard models, and it should mean that instability problems, if they do show up, won’t be a widespread or long-lasting. But frankly, I say “should” because I don’t know. Intel has all of the pieces in place, and my guess is that the Core Ultra 9 285K won’t have instability problems. That’s something that’s impossible to test for a day-one review, though. Only time will tell.

Should you buy the Core Ultra 9 285K?

A hand holding the Core Ultra 9 28K.Jacob Roach / Digital Trends

Part of me is enamored with the Core Ultra 9 285K. I see the huge efficiency improvements; I see the voodoo that Intel has pulled off with 24 threads. More than anything, I see a company that’s willing to strip all of its core tenets, from in-house manufacturing to Hyper-Threading, to deliver a radically different CPU that’s not just more of the same. Say what you want about the Core Ultra 9 285K — there’s no denying that it’s impressive.

But impressive doesn’t always mean good. I hope Intel can continue building from the Core Ultra 9 285K to deliver highly efficient components, and based on my testing, it seems like even one generation down the road we’ll see an offering from Intel that’s seriously special. From where I’m standing today, however, I can’t recommend the Core Ultra 9 285K outright.

First, for gamers, there’s really not much for you here. Yes, the Core Ultra 9 285K is way more efficient than the Core i9-14900K, but it’s not any faster. More importantly, the Ryzen 7 7800X3D exists, and the Ryzen 7 9800X3D is on the way. Intel really gave up on the gaming fight before the Core Ultra 9 285K ever released, and if your focus is gaming, AMD’s 3D V-Cache CPUs are still the way to go.

On the productivity end, the Core Ultra 9 285K and Ryzen 9 9950X trade blows. However, I saw much more consistent performance out of the AMD chip. It’s possible Intel still has software improvements it can deliver down the road to gain the upper hand, and I’d love to see them. That could earn this chip the Recommended badge it just barely missed. But for now, the Ryzen 9 9950X makes more sense than the Core Ultra 9 285K.

I don’t want to discount the Core Ultra 9 285K completely. There’s certainly a use case here if you have a small form factor PC, for example, and want a balance of good productivity and gaming performance. The Core Ultra 9 285K slots into that system perfectly. But for everyone else, there are better options.