Intel Core i9-9900K Review: Welcome to an Intel-AMD 8-Core Slugfest

The CPU market has changed more in the past 22 months than it did from 2008-2017 combined. That's not to say that CPU architectures didn't change, or that we didn't see per-core pricing come down. But Intel had a remarkably successful run with the product positioning it introduced at the high-end with Nehalem in 2008 and refined with Sandy Bridge in 2011. Quad-cores with Hyper-Threading at the top, a midrange Core i5 sans HT in the middle, and a Core i3 dual-core with HT re-enabled to anchor the low end. From 2011-2017, that was Intel's desktop product line in a nutshell -- until AMD launched Ryzen. Fast forward to today, and we find ourselves in a completely different ballgame.

Intel's 9th Generation family completes the transformation of the product line that began with 8th Generation parts. Hyper-Threading has vanished from the stack, save for the Core i9-9900K. The difference between the Core i7 and Core i5 has similarly shrunk, at least in terms of thread count. Previously, chips like the Core i7-8700K(Opens in a new window) or Core i7-7700K supported twice the total threads of the Core i5 family, with the caveat that these were logical rather than physical processors and did not deliver anything like the scaling of a full core. We typically assume Hyper-Threading support adds ~20 percent performance. The Core i7-9700K no longer offers Hyper-Threading, but the core count has been bumped up to eight to compensate.

This launch is a critical chance for Intel to recover some of the prestige the company has lost over the past 10 months. While the company's earnings have been excellent, it's taken a hammering in the press for a variety of reasons, including (in no particular order): security problems like Spectre and Meltdown, the unexpected sudden resignation of the CEO, a significant delay stacked on top of an already-significant delay to its 10nm process deployment, and a host of downstream impacts from that, including a CPU shortage and a rumored delay to its EUV deployment timeline. As we've written, these downstream effects should be viewed as the logical downstream impact of the 10nm delay rather than a series of separate, unrelated problems, but they still add up to a rocky year.

The Core i9-9900K offers an opportunity to change that narrative -- but only if it can get past the Ryzen-sized competitor standing in its path.

What to Watch for

We're evaluating a large range of chips today, covering multiple product families. We've pulled in the Core i7-7700K to illustrate the performance gains from Intel's last quad-core CPU, included both the Core i7-8086K and Core i7-8700K to cover enthusiasts who may have picked up a previous generation top-end part (or are simply curious about the 8086K in general), and tossed the 10-core Core i9-7900X in for good measure to illustrate the potential performance benefits that come with that platform's quad-channel DDR4 support and additional two cores.

Our benchmarks on the AMD side are thinner -- AMD simply hasn't been building high-end parts for as long as Intel has -- but they'll answer the important questions. The Ryzen 7 2700X(Opens in a new window) is our primary point of comparison, but we've kept the 2950X in the mix as well. First, it allows us to showcase AMD's $900 CPU to match our inclusion of the $1,000 Core i9-7900X. Second, it illustrates what kind of value the Threadripper 2950X provides against a CPU from AMD's greatest rival.

We've also separated out some test results in places where we updated benchmarks or added tests.

We tested the Intel Core i9-9900K on an Asus Maximus XI Hero motherboard with 32GB of DDR4-3200 installed in all four DIMM slots, a GeForce GTX 1080 Ti running Nvidia 411.63 drivers and a 1TB Samsung 970 EVO M.2 SSD for storage. Windows 10 Build 1803 was used for the testbed with the latest patches and updates installed. Our test results are embedded in the slideshow below.

Gaming Benchmarks

There are two schools of thought on gaming benchmarks. One argument says to test CPUs with GPU quality levels that won't actually stress the card, allowing the reviewer to highlight architectural differences. The other says that CPUs should be benchmarked with detail levels that gamers themselves are likely to use, particularly when testing high-end hardware. I agree with both views, but I deploy them at different times: Lower detail levels make sense when the goal is to investigate low-level differences in a new CPU architecture, while normal detail levels make more sense for benchmarking a CPU family once it's established in-market. And while I'm sure there's someone out there who likes running a GeForce GTX 1080 Ti(Opens in a new window) at 1080p medium detail because they really hate frame rate drops, I can't say I've ever met them.

The honest truth is this: I run these benchmarks to give the review a sense of completeness, not because there's any question about what I'm going to find. Games are the least effective way to highlight differences between Intel and AMD CPUs at reasonable detail levels for a GTX 1080 Ti, and our tests show a minimal deviation between chips. The only real exception to this is Ashes of the Singularity's CPU test, which is specifically designed as a heavily CPU-bound scenario and should scale with chip clock and core count. There are few games that fit this performance profile, however, and gamers should not worry about whether a Ryzen 7 or older Intel CPU is capable of running the latest games. When the Core i7-8700K came out, the big discussion was around whether it represented a good gaming upgrade to 2011's Sandy Bridge, a six-year-old chip at the time. When it comes to gaming, older Intel chips are not a problem and neither is Ryzen.

Intel Retakes Performance Lead, but AMD Has a Death Grip on Performance Per Dollar

Intel's Core i9-9900K sets new records in both single and multi-threaded performance as far as Intel's mainstream desktop family is concerned. It's faster than AMD's Ryzen 7 2700X and that chip took our top recommendation from the year-old and still-stellar Core i7-8700K. If you play exclusively within the Intel ecosystem, there's a great price story as well. Up until now, buying an eight-core Intel CPU meant paying a substantially larger premium because it required customers to commit to either the Core X family or a Xeon processor. List price on the Core i7-7820X (8 cores, 3.6-4.3GHz) is $600 and high-clock Xeon processors aren't cheap either. In addition, Xeon and HEDT customers pay more on average for motherboards.

At ~$480, the Core i9-9900K knocks $120 off the list price of the 7820X, while Z370 motherboard compatibility could open a $30-$50 gap between the cheaper consumer boards and more expensive LGA 2066 options. The Core i9-9900K is a genuine improvement to Intel's product stack as far as available performance per dollar. It's also the best CPU on the market for those who demand uncompromised performance in single-threaded and multi-threaded workloads. It's significantly faster and cheaper than older cores like the Broadwell-E based Core i7-6900K were when they debuted -- the 6900K was a $1089 part in 2016, compared with an expected $480 for the Core i9-9900K today.

The CPU's only flaw is that Intel knows what it has and priced it accordingly. In doing so, it's making a calculated bet that AMD's ability to take market share in the desktop space is still largely limited to low-cost systems or the retail channel. Intel's listed CPU prices are generally understood to only apply to the retail channel -- OEMs don't pay full cost -- and Intel may be willing to sacrifice some channel market share so long as OEMs continue mostly shipping high-end Intel rigs.

To date, this strategy seems to have worked. Its Core X CPUs aren't competitive with AMD based on the retail list price -- the Core i9-7900X loses every test to the Threadripper 2950X, including benchmarks that Intel tends to win. Ordinarily, we'd predict that Intel would lower its per-core prices, but the company hasn't done so since it first introduced the Core X family in June 2017. This suggests that Intel isn't worried about the pricing disparity between itself and AMD when it comes to the Core X family, but we don't know if this will continue to be the case for the Core i9-9900K.

There was a time when making the price/performance argument for AMD felt a bit like trying to convince someone that a Peel P50(Opens in a new window) and a Mini Cooper S offered an equivalent driving experience. Those days are over. The Ryzen 7 2700X isn't just a fast CPU "for the money," it's an objectively fast CPU period. There's a difference between a product that's managed to wedge itself into a niche where it can make a narrow argument for its own utility and hardware that represents a damn good value in its own right. AMD has moved from the former category to the latter.

The irony of the Core i9-9900K is that absent Ryzen, it's incredibly unlikely this chip would exist at all. In 2016, Intel launched Broadwell-E and priced its 10-core CPUs at $1,723. The 8-core Core i7-6900K retailed for $1,089, a $90 increase over the previous Haswell-E generation. Intel had no intention of cutting its CPU prices or offering more cores except as its competitive response to Ryzen dictated. Compare the Core i7-7700K with the Core i9-9900K if you want to see just how seriously Intel takes Ryzen as a competitive threat, and then remember that Kaby Lake isn't even two years old. There's a marked disparity between the adjustments Intel made to its product family after Ryzen launched -- the company has completely overhauled its product segmentation at every price point -- and how much it's been willing to adjust its prices.

The Core i9-9900K does not generally justify its price premium when compared with AMD's Ryzen 7 2700X or even cheaper Intel CPUs. With the Ryzen 7 2700X priced at ~$300, buying the Core i9-9900K for an additional $180 constitutes a 1.6x price penalty for an average 1.15x performance gain across our entire suite of benchmarks. Keep in mind, however, that this varies fairly reliably by workload, with AMD competing more closely in rendering (9900K leads by 6 percent) then in 7zip compression or H.265 encoding (9900K transcodes to that standard in 75 percent the time it takes a Ryzen 7 2700X). A gap of that size is an advantage some individuals would be happy to pay for -- and that means there are going to be some specific areas, possibly including H.265 transcoding, where the Core i9-9900K is a very good deal.  On the other hand, if you're a gamer, the $180 you'd save by buying AMD or a cheaper Intel chip would make the difference between buying a GeForce 1060 6GB ($250) and a GTX 1070 Ti ($409) with $21 to spare. That swap would make a far greater impact on your gaming performance than stepping up to the 9900K.

The value proposition of the 9900K also ultimately depends on whether you view it as a potential upgrade to the quad-core or six-core CPU you were otherwise going to buy or as a potential alternative to the eight-core+ Core X chip you intended to purchase. Compared with Intel's pre-existing top-end ecosystem, it's a great step forward. Compared with what you can buy from AMD or what Intel offers to people who don't need all eight cores, not so much. But that's more an argument for careful consideration than it is an argument against the CPU, and we always recommend people consider their likely usage patterns and general needs before pulling the trigger on any hardware purchase.

Intel's walking away from this dust-up with the performance crown and AMD's keeping its own grip on excellent performance per dollar. Nobody actually loses here, except maybe Core i7-7700K customers. We do kind of feel sorry for you guys, even if it's still an objectively good CPU.