Whether Intel acknowledges it or not, since the launch of AMD’s first-generation “Summit Ridge” Ryzen processors (spearheaded by the excellent Ryzen 7 1800X), the two companies have been in the midst of a Core War. With the launch of the $499 Intel Core i9-9900K, Intel now has a shot at claiming the mainstream-CPU productivity crown from AMD. Although raw performance isn’t the only factor to use to determine which components to buy, it is an important one. And without a doubt, on that front, Intel scores a definitive win with the eight-core Core i9-9900K. That said, while this epic mainstream processor delivers killer scores on benchmark testing, and is a top-flight pick for performance-minded PC gamers and content creators, you’ll need to factor in a decent thermal topper for this chip—and some extra budget for both that and the chip itself. It’s the priciest desktop CPU of recent memory on a mainstream, rather than specialized enthusiast, platform. But oh, is it ever a good one.
The Whole i9 Yards
If you’ve purchased an Intel Core processor (or a pre-built laptop or desktop) at any point in the last decade, you know how Intel segments its products into Core i3, Core i5, and Core i7 families. With the “Nehalem” microarchitecture, which launched in 2008, Intel began offering Hyper-Threading on its Core i7 chips, but not on the Core i5 ones. With Hyper-Threading, by letting each core handle up to two threads on these higher-tier processors, users realized approximately 30 percent better performance per core, depending on the application. For this reason, the Core i7 processors were a good deal more powerful than the Core i5s, especially when running highly threaded software. But with the 9th Generation Intel Core Processors, Hyper-Threading is no longer a standard feature on the Core i7s.
Intel already has a tier of processors above the Core i7s, which it introduced in May 2017, with the Skylake-X HEDT (High-End Desktop) processor family, the leading chip of which is the formidable, multi-kilobuck Core i9-7980XE Extreme Edition. This first wave of Core i9s was designed specifically for well-heeled extreme enthusiasts and pro content creators. The core counts with the Core X-Series Core i9 processors start at 10 and go up to 18, with Hyper-Threading a standard feature, but all of these chips require a pricey Intel X299-based motherboard, and the Core i9-line chip prices themselves start at a cool $1,000 with the Core i9-7900X. Apart from access to these supercar-speed processors, support for four-channel main system memory and lots of extra PCI Express lanes (to accommodate multiple video cards or PCI Express-based SSDs) are the reasons that buyers might opt for the X299 platform.
In contrast to the Core i9 chips of Intel’s Core X family, the Intel Core i9-9900K is the first i9 processor to become available on Intel’s mainstream desktop platform, and it’ll work in significantly more affordable Z370 motherboards, though a BIOS update will be required. This chip also offers two more cores than the former mainstream flagship chip, the six-core Core i7-8700K. When you look at the rest of the Core i9 processors in Intel’s stack, it’s easy to see the Core i9-9900K as a bit of a middle ground—extreme silicon, to be sure, but designed to appease enthusiasts with no need for Core X’s spare PCI Express lanes and quad-channel memory.
On the other hand, the $499 list price on the Intel Core i9-9900K (retail prices at the time I wrote this started at $530 and went up even higher) is a harder pill to swallow if you were expecting this chip to be a straight-up replacement for the Core i7-8700K. Instead, Intel is launching the $385 Core i7-9700K as the top-end Core i7 processor for the 9th Generation, and although it has eight cores, Hyper-Threading is not a supported feature. The core and thread jump from the Core i7-7700K to the Core i7-8700K is measurable in almost everything you do with your PC, but Intel doesn’t seem keen to tread the same path this time. Are there instances where the Core i7-8700K, with its six cores and 12 threads, will outperform an eight-core/eight-thread Core i7-9700K? Perhaps, though we suspect these will be the exception rather than the rule. I’ll save that discussion for that processor’s review, assuming I can get my mitts on one.
And what about AMD? At eight cores and 16 threads in their flagships, AMD and Intel may have achieved core parity on their respective front lines now, but AMD is once again wielding price as its greatest weapon. Before we talk about how Intel competes, however, let’s take a close look at what you get with the Core i9-9900K.
The Guts of 9th Gen Core
Although the Core i9-9900K looks just like all the rest of the LGA 1151 processors Intel has manufactured, there’s something new (well, new since the days of “Ivy Bridge”) under the hood. Instead of using silicon-based paste between the processor’s die surface and the glued-on heat spreader, Intel is bringing back bonded metal, or solder. This Soldered Thermal Interface Material (STIM) is considerably more capable when it comes to pulling heat away from the die when the CPU is under load. With a decent CPU cooler attached, STIM can help keep your processor running cooler.
Like the present of an unlocked multiplier, this new feature is a wink from Intel directly to overclockers, who’ve created a whole cottage industry around delidding Intel processors to replace the paste TIM with something more thermally conductive. It’s also worth noting that all of AMD’s AM4-based processors, with the exception of its “Raven Ridge” chips, rely on bonded metal between the die surface and the heat spreader.
The Core i9-9900K is built on Intel’s umpteenth revision of the 14nm process (dubbed here “14nm++”). However, the chip giant has managed to fit two more cores with similar clock speeds into the same package. I’ll take a closer look at the Core i9-9900K’s real-world power situation later on.
The Intel Core i9-9900K is a 95-watt TDP processor that features eight cores and 16 threads, and it’s built on an LGA 1151 package. This processor belongs to the Intel “Coffee Lake-S” family, and it has a 3.6GHz base clock and a maximum Turbo Boost frequency of 5GHz. Like the Intel Core i7-8086K Limited Edition I recently reviewed, this 5GHz Boost clock applies only when a single core is active. In my tests, this processor boosted up to 4.7GHz when all cores were active.
Other features include 16MB of Intel Smart Cache that’s available to all eight cores, a dual-channel memory controller, and Intel UHD Graphics 630 integrated graphics with a 350MHz minimum and 1.2GHz maximum GPU frequency. With the exception of the Smart Cache, the rest of these features are the same on the Core i7-8700K. The memory controller is rated to support up to DDR4-2666 memory, and Intel Extreme Memory Profile support means that Z370 and Z390 motherboards can support memory speeds beyond 4,000MHz.
The processor has 16 PCI Express lanes available for discrete graphics cards, and the integrated UHD Graphics 630 processor (the same graphics engine as on the Core i7-8700K) supports overclocking with an unlocked multiplier. When you slot this chip into a Z370 or Z390 motherboard, you’ll also get unlocked base clock and memory ratios, support for per-core overclocking, and adjustable voltages.
You may not be terribly inclined to use the UHD Graphics 630 for gaming (I was unable to test it because my system’s MSI MEG Z390 ACE motherboard does not feature a graphics output) but it’s not just some useless vestigial lump of silicon. With Intel Quick Sync Video Technology, this portion of the chip can rapidly convert HEVC 10-bit (H.265) video files and encode/decode premium 4K Ultra HD content, for instance from Netflix. This chip also supports the AVX2 instruction set, Intel Optane Memory, and Intel Turbo Boost Technology 2.0.
Intel is also launching a slightly-tweaked Z390 chipset to go with the 9th Generation Core processors, though if you already have a Z370 motherboard, there’s not a lot to get excited about. The upticks from Z370? Z390 features an integrated USB 3.1 Gen 2 controller for up to 10Gbps ports, and integrated Intel Wireless-AC with support for Gigabit Wi-Fi speed.
That said, not all Z390 motherboards will necessarily ship with said ports and Wi-Fi capabilities, but at the high end, Z390 motherboards equipped with an Intel Wireless-AC 9560 adapter will support theoretical data rates up to 1,733Mbps. Also, know that, as mentioned in passing earlier, not all Z390 boards will necessarily have video outputs that will let you use Intel’s integrated graphics. Shop with care if that matters to you. (See our preview of Z390 motherboards from Asrock and MSI.)
Testing Stock Performance
On paper, the Intel Core i9-9900K looks to be a powerful processor, but to see just how it compares to the other muscle-car chips currently on the market, I ran a passel of tests to determine how it handles a host of workloads at its default settings.
For my test setup, I installed the Intel Core i9-9900K into the MSI MEG Z390 ACE ATX motherboard mentioned earlier, and populated two of the DIMM slots with 16GB of dual-channel G.Skill Sniper X DDR4-3400 memory. For the Windows 10 boot drive, I relied on the 240GB Crucial BX300 6Gbps SATA SSD. I installed the components into an Alpine White EVGA DG-77 case and used the Fractal Design Celsius S36 closed-loop liquid cooler to flush heat away from the processor’s STIM-fused integrated heat spreader.
Keep in mind, like many of Intel’s enthusiast-centric processors, the Intel Core i9-9900K does not include a stock CPU cooler in the box, so you’ll need to have one on hand or buy one. For our game testing, I supplemented the Core i9-9900K with an Nvidia GeForce GTX 1080, operating at the Founders Edition clocks.
To compare this processor’s scores with that of other chips currently on the market, I included in the charts below scores for several chips mentioned earlier: the six-core/12-thread Intel Core i7-8700K and Core i7-8086K Limited Edition, and the 10-core/20-thread Intel Core i9-7900X, as well as the eight-core/16-thread Intel Core i7-7820X. The first two are on the same platform as the Core i9-9900K and will work with Z370 motherboards, while the two Core X-Series chips (the ones ending in “X”) rely on X299.
For the AMD side of the aisle, the competitors are the mainstream-flagship eight-core/16-thread AMD Ryzen 7 2700X, the step-down six-core/12-thread Ryzen 5 2600X, and for kicks and context, the much costlier 16-core/32-thread Ryzen Threadripper 2950X.
Maxon’s 64-bit Cinebench R15 is a CPU-centric test that lets us gauge both the single-core and multicore performance of the various processors I tested. The resulting scores are test-specific numbers that represent the processor’s performance while rendering a complex CPU-intensive image. This is considered a synthetic benchmark.
In the Cinebench R15 multi-threaded subtest, more cores tend to return higher scores. Despite this, the Core i9-9900K scores more in line with the much pricier 10-core Core i9-7900X. The eight-core Ryzen 7 2700X and Core i7-7820X are neck-and-neck, and the trio of six-core processors brings up the rear.
The Cinebench R15 single-threaded test is oblivious to more than one core, so megahertz matters most here. As you’d expect, the two 5GHz Intel processors take the top spots, but the Core i9-9900K has a slight edge.
iTunes 10.6 Conversion Test
The iTunes 10.6 Encoding Test is tragically single-threaded, which means that more cores simply don’t make a dent on these workloads. This test is designed to illustrate the performance you might expect when running legacy software that doesn’t scale well across more than one core.
The iTunes encoding test is a bit like Cinebench R15’s single-threaded test, where sheer one-core clock speed plays a big role. The Core i9-9900K once again comes out on top, but the rest of the Intel chips aren’t far behind. All three of AMD’s processors finish the encoding task 16 seconds or more behind the slowest Intel processor, the Core i9-7900X.
Handbrake is a classic (and popular) workstation application that is used to convert videos between formats. Typically, the more threads and cores a processor has, the better it will perform in this utility. I loaded up a 12-minute-long open-source 4K movie titled Tears Of Steel and used the software to convert it into a 1080p MPEG-4 video.
The AMD Ryzen Threadripper 2950X is a force to be reckoned with when it comes to Handbrake renders, but the Core i9-9900K comes in second place, beating even the about-a-grand Core i9-7900X. The AMD Ryzen 7 2700X clocks in almost a full minute behind the Core i9-9900K.
This benchmark is another one that’s generally considered synthetic; however, the highly threaded nature of the utility is getting to be more and more representative of the applications available today. (The benchmark tasks the processor with rendering a complex photo-realistic image using ray tracing.) I ran POV-Ray using both the multi-threaded “All CPUs” setting and the hamstrung “One CPU” setting.
POV-Ray’s single-threaded workload seems to favor Intel’s processors, and the Core i9-9900K has an impressive run once again to claim the top of the chart. The multi-threaded POV-Ray doesn’t deliver such one-sided results, however, with the Threadripper chip predictably taking the top spot by a huge margin. Just 11 seconds separates the AMD Ryzen 7 2700X’s fourth-place finish from the Core i9-9900K in third place.
Another of the real-world benchmarks I used is Blender, a popular open-source 3D rendering application that people far more creative and talented than I use to craft 3D visual effects, animations, and models. Our test file consists of a cartoonish flying-squirrel render that takes less than a minute to complete with most modern processors.
In Blender, the only processor that took more than 25 seconds to render our test image was the Ryzen 5 2600X. The rest of the processors managed it in between 22 and 18 seconds. The Core i9-9900K scored the fastest time, and the Ryzen 7 2700X was a mere 4 seconds behind it.
7-Zip File Compression
7-Zip is a widely used file-compression utility that features a built-in compression/decompression benchmark. It’s a real-world test that generally makes use of as many cores and threads as your processor has to offer.
A cursory glance at our results, and it’s immediately clear that 7-Zip likes multi-core processors. Although the 16- and 10-core processors capture first and second place, the Core i9-9900K is the fastest of the eight-core processors we tested. The six-core Intel chips (the Core i7-8700K and Core i7-8086K Limited Edition) left the AMD Ryzen 5 2600X in the dust.
The Ultimate Gaming CPU (With an Asterisk)
Due to time and hardware-availability constraints, I narrowed the game-benchmarking focus to just three processors: the Intel Core i7-8086K Limited Edition, the Intel Core i9-9900K, and the AMD Ryzen 7 2700X. The first has six cores, but it shares the Core i9-9900K’s 5GHz Boost clock, and the AMD processor has the same core count as the subject of this review. The component doing the real heavy lifting for the game benchmarks is the Nvidia GeForce GTX 1080 card.
To keep the platform comparisons as apples-to-apples as possible, I used the same graphics card, as well as the same memory kit, running at the same frequency (3,400MHz) and with the same timings. I ran in-game benchmarks from Far Cry Primal (at the High graphical preset) and Rise of the Tomb Raider (DX11, on the Very High preset) at three resolutions: 1080p, 1440p, and 4K. I also ran the benchmarks at both stock and overclocked settings, and occasionally saw a slight performance improvement for our efforts (but not always).
After I reviewed the Core i7-8086K Limited Edition, I was skeptical that Intel’s Core i9-9900K would be able to top its gaming performance, even with a Boost clock to match. But the numbers don’t lie. At the stock and overclocked settings, the i9-9900K scored 138fps and 140fps in Far Cry Primal and Rise of the Tomb Raider’s 1080p tests, respectively. The Core i7-8086K was right there, but the AMD platform only managed stock and overclocked frames-per-second (fps) scores of 106fps/109fps in Far Cry Primal and 132fps/133fps in Rise of the Tomb Raider. That’s a 26 percent deficit for the AMD platform in Far Cry Primal. In Rise of the Tomb Raider, the difference between the two platforms is less apparent; the AMD chip is just 6.6 percent behind Intel at 1080p. Although both platforms yielded frame rates above 100fps, this gulf is not inconsequential.
But what happens when we move the resolution to 1440p? As the CPU becomes less a bottleneck and the graphics card gets some room to run, the difference all but vanishes. Compared to the Ryzen 7 2700X, Far Cry Primal was 2 percent faster on the Core i9-9900K platform. In Rise of the Tomb Raider at 1440p, the difference at the stock settings was less than a single frame per second. At the 4K resolution, in both games, Intel’s and AMD’s platforms performed the same.
So to sum up: Yes, Intel’s Core i9-9900K is dominant in games, as long as your resolution and graphics card choices are such that the CPU becomes a bottleneck. (Generally, that will be an issue mainly at 1080p.) The testing I did was by no means exhaustive, and depending on the game, there will be wins for Intel and there will be wins for AMD. If you’re only gaming on your PC, then you can save hundreds of dollars by getting a six-core Ryzen 5 2600, a Core i5-8600K, or a Core i5-9600K and pouring that extra money into a beefier graphics card, or a monitor with a higher resolution or a high refresh rate. The extra frames Intel’s platform will get you don’t offer near the experience that increasing your resolution and/or refresh rate do. The games-only demographic will not find much value in the i9-9900K versus other high-end CPUs. Luckily for content creators, extreme multitaskers, and the rest of us power users, this processor speeds through virtually any task you toss its way.
To start, I loaded the BIOS menu of the MSI MEG Z390 ACE motherboard and raised the multiplier to 50. After a series of reboots and tests, I settled on a core voltage setting of 1.33V, which is still modest enough to maintain for the long term should I choose to make this clock speed a permanent setting.
Back in Windows 10, with 5GHz on all cores running stable and within reasonable temperature thresholds, I ran Cinebench to see the performance benefits. The Intel Core i9-9900K went from scores of 2,063 (multi-threaded) and 218 (single-threaded) to 2,188 and 222, respectively. My overclock shaved a second each off the Blender and iTunes encoding tests, 20 seconds off the Handbrake result, 4 seconds off the POV-Ray multi-threaded test result, and 3 seconds off the single-threaded POV-Ray test score. My results in the games didn’t show much improvement overall, but I achieved 4fps more in Rise of the Tomb Raider at 1080p, and 2fps more in the game’s 1440p result.
Power and Thermal Testing
When Intel says that the Core i9-9900K has a 95-watt TDP, that has virtually no bearing on the amount of power that this chip draws, even at stock settings. Back when I tested the six-core Core i7-8086K Limited Edition, another chip with a 5GHz Boost clock, the 95-watt TDP wasn’t far from the 103-watt total package power draw I saw in AIDA 64’s System Stability Test. When I overclocked the Core i7-8086K to run at 5GHz on all cores, power draw increased to 141 watts. At its stock settings, the Core i9-9900K running the AIDA 64 System Stability Test drew a whopping 165 watts. Keep in mind: This is out-of-the-box performance. You are likely to encounter loads similar to this in real-world computing scenarios.
Despite this amount of power, the Fractal Design Celsius S36 (the closed-loop liquid cooler I mentioned earlier, equipped with a 360mm radiator) was able to keep the processor running at between 65 and 75 degrees C. When I overclocked the Core i9-9900K to 5GHz on all cores, the power draw increased another 10 watts, but temperatures climbed into the mid-80s, with highs in the 90s. I have no problem running a system that occasionally spikes into the 90s for short bursts. If I’m doing something with this computer that takes hours to complete, however, I would back down to stock settings. If you keep your PC running heavy loads routinely, you may need a custom liquid-cooling system with more thermal capacity than a mere compact liquid cooler. If you plan to run an air cooler, I do not recommend overclocking the Intel Core i9-9900K, unless perhaps you equip it with a specialized high-end air monster such as one of the Socket 1151 models from Noctua.
When I first learned that Intel was using STIM between the die and the heat spreader of the Core i9-9900K, I thought that this was an example of Intel really listening to the enthusiast community. After working with the chip for a couple of weeks, however, I’ve concluded that anything less would have been a nonstarter.
9th Gen Core: Still Got the Goods
Even though it’s the product of an iterative refinement, the Intel Core i9-9900K is no less a wonder of modern silicon engineering. The fact that it does what it does, while having been built on the same manufacturing node as the previous Coffee Lake processors, boggles the mind. With Intel’s ongoing woes around the move to the 10nm process, few would say that the Santa Clara chipmaker is at the usual top of its leading-edge game, but this processor would not exist if some of the world’s best and brightest weren’t tweaking and iterating on an already solid platform.
If anybody was thinking that AMD’s second-gen Zen was going to be the blow that would put Intel in the underdog’s place at the mainstream, then that person just doesn’t know Intel very well. This processor isn’t for everyone, but for those who can afford both it and the serious cooling needed to overclock it, the Intel Core i9-9900K will not disappoint gamers, content creators, and extreme multitaskers who expect their CPU in a single-GPU system to do it all and do it all well. That said, Intel’s existing Hyper-Threading-capable 8th Generation Core chips and AMD’s top-end second-gen Ryzens continue to deliver superior value for those who can live with gear a notch or two below the leading edge and just one video card.