While the GPU conversation may well have seemed like a one-way street in favour of NVIDIA, 2018’s CPU market was the most competitive it’s been in years. That competitiveness extended from top to bottom, encompassing the gamut of budget desktop offerings through to high-end server chips.
We ended 2017 with AMD and Intel duking it out in an exceptionally tight battle. Ryzen’s release had forced Intel into a corner, pushing them towards bringing six cores into their mainstream platform as the Coffee Lake-based Core i7-6700K. Threadripper too triggered a more aggressive Intel approach, as the Core X HEDT platform would now offer up to 18 cores (36 threads) in a single package known as the Core i7-7980XE. For the first time in a long while it felt like Intel were playing catch-up, and it seemed like AMD had more in the tank. So it proved in 2018.
The first major release of 2018 was AMD’s. Desktop APUs, long the mainstay of AMD’s processor lineup in the troubled times between 2011 and 2016, hadn’t been updated to reflect either the new graphics or CPU architectures available to AMD. That would be rectified with the first Ryzen 2000-series CPUs, incorporating Zen CPU architecture and Radeon Vega Graphics in one package on the AM4 socket. AMD 300-series motherboards would already support the new APU thanks to the single-socket mainstream desktop platform approach.
AMD were far from done. As Winter gave way to Spring, so the Ryzen 1000-series was refreshed into the 2000-series. Although Zen+ was primarily designed to exploit TSMC’s 12nm process – itself a derivative of the 16nm process on which first gen Ryzen was built – it would allow for a few tweaks here and there that culminated in a substantially superior package. AMD’s motherboard chipset range was also refreshed to the 400-series, but the platform remained broadly backwards- and forwards-compatible.
Perhaps the greatest innovation 2nd gen Ryzen brought was the introduction of a new multi-core turbo mode known as Precision Boost 2. Leveraging improved sensor data (thanks to AMD SenseMI technology), PB2 dynamically boosts operating frequency no matter the number of cores loaded. Previously, once more than two cores were under heavy load, the Precision Boost algorithm drastically limited operating frequencies across all cores.
Precision Boost 2 makes post-purchase overclocking all but obsolete for all but the most adept of enthusiasts, while at the same time increasing the value of high quality cooling. Suddenly, even the rankest of amateurs could exploit even more that the underlying silicon had to offer, beyond the raw specifications on the box.
Not everything in 2018 went AMD’s way. A mini-controversy brewed in March as cybersecurity researchers CTS Labs accused AMD of having “complete disregard of fundamental security principles” due to “the basic nature of some of [..] vulnerabilities [outlined by CTS Labs.]”. They labelled these vulnerabilities ‘RyzenFall’ and ‘Masterkey’, it what appeared to be a rather brazen attack on AMD specifically.
AMD’s response was measured, while cybersecurity experts seemingly queued up to debunk the claims made by CTS Labs. In particular, it was quickly noted that many of the vulnerabilities actually exploited 3rd party hardware rather than AMD’s own, and that many required root and physical access to the hardware to exploit. In essence, the requirements necessary to take advantage of the vulnerabilities would mean that the intruder could already do far more damage through more direct means.
Outrage soon petered out, and it’s unlikely to have done any long-term damage.
While the GPU market remained dormant CPUs took centre-stage at Computex in 2018. Intel, seeking to capture the headlines, reeled off a range of accomplishments and announced a new Core X HEDT lineup during their Keynote presentation. Perhaps the highlight of the conference however was a demonstration of their upcoming Core X flagship, a 28-core monster chip operating at 5GHz that would cut through multi-threaded benchmarks like a hot knife through butter.
All was not as it appeared.
Later on that week AMD presented their own upcoming HEDT platform, a refresh of 2017’s Threadripper. Soon to be available in 32-core configurations, up from 16, and eclipsing all of Intel’s HEDT lineup (current and announced), 2nd Generation Threadripper brought plenty of the good stuff from their EPYC server lineup. To really put the boot in, AMD re-ran the same benchmarks previously blitzed by Intel’s 28-core 5GHz demo, putting the results in the shade without the need for exotic cooling and an extreme overclock.
Summer and Autumn months drove home the quickly-changing nature of the CPU market in 2018. AMD led the way with 2nd Generation Threadripper in August, boasting both the 16-core and 32-core models at extremely competitive prices for an HEDT chip. A gauntlet was thrown down in the form of the 32-core, 64-thread Threadripper 2990WX for an MSRP of only $1799 (less than $60 per core), one which Intel would struggle to respond to.
Nonetheless, Intel finally had their own many-core answer to the vexing problem of 2nd Gen. Ryzen in the mainstream. October saw the launch of the 9th Generation CPUs, based on Coffee Lake and debuting the Core i9 branding in the mainstream consumer platform. So marked the entrance of an Intel 8-core/16-thread option on the mainstream platform, the Core i9-9900K, as well as 6c/6t now being something of a baseline thanks to the i5-9600K. Hitting 5GHz Turbo, the flagship swiftly became the CPU to beat in gaming. At an MSRP of $488 however, and low availability pushing street prices higher, AMD was still able to match on a price/performance basis.
Intel's Core i9-9900K has genuine flagship performance, check out our review
As the leaves left the trees Intel launched their refreshed Core X lineup. Now entirely based on a refreshed Skylake X architecture, it dispensed with the irrelevant quad-core entry level Kaby Lake-X models in favour of starting at 8-cores with Hyperthreading. Clock speeds were adjusted upwards across the range (considerably in the case of the i9-9980XE), and L3 cache quantities were re-jigged in a couple of segments, but little meaningfully changed compared to their 7000-series counterparts otherwise.
More recently, Intel’s 28-core behemoth exhibited at Computex appeared on websites as the Xeon W 3175X, priced at $3999. That rather stark price differential compared to the 32-core Threadripper all but consigns it to a very different market, even if it is marketed as a HEDT chip.
AMD’s aggressive Threadripper pricing, even at the top-end where they appear to offer features (i.e. 32 cores) that Intel cannot, means that the CPU market is as exciting at the end of the year as it was at the start, if not more so.
There’s plenty more to come in 2019. Zen 2 will make an appearance in the radically re-designed 2nd generation of EPYC server CPUs code-named Rome, and TSMC’s 7nm manufacturing process will be used for the first time on these processors. Transitioning to 7nm may also allow AMD to compete in pure clock speed with Intel, making them more attractive still to gamers as and when Zen2 arrives on the mainstream and HEDT platforms. For their part Intel still have work to do on a die shrink of their own, but don’t count out their talented engineers (or deep pockets) prematurely.