AMD 2nd-Gen 32-core Ryzen Threadripper 2990WX arrives this week

Double the cores, double the fun! AMD's highest-end processor jumps from 16 to 32 cores in the new Ryzen Threadripper 2990WX, based on AMD Zen+ architecture, slated to ship on Aug. 10. At the end of August, the current 16-core flagship, the 1950X, gets its 2nd-generation upgrade to the 2950X but a demotion to middle-of-the-line of AMD's high-end processors.

For the majority of people who don't build their own systems, this means that PCs from custom builders like Origin PC and Falcon Northwest are imminent, and likely closely followed by the bigger gaming OEM lines such as Alienware, Republic of Gamers and MSI. They'll be expensive, though, based on the boxed-CPU prices -- the 2990WX is $1,800 and its 24-core sibling the 2970WX $1,300. They're still cheaper than Intel's 18-core i9-7980XE, though.

The Zen+ architecture is based on a smaller 12nm process than the previous Zen 14nm process. But perhaps more important from a performance standpoint, the updated CPUs incorporate AMD's Precision Boost 2; the first version took an all-or-nothing approach to its decisions about when to boost clock speed, but the newer version makes more intelligent decisions about how many cores to boost and when to boost them for better, more consistent and more efficient operation. 

That means even though the 2950X has the same number of cores as before, the increased clock speeds enabled by the Zen+ architecture plus the the improved Precision Boost should deliver noticeably better performance.

AMD claims its team successfully overclocked the 2990WX to a clock speed of 5.1GHz and a score of 7,618 on the Cinebench R15 multi-core (with liquid nitrogen cooling at a frigid -385° F), significantly more than any other shipping consumer processor; Intel's best recorded OC score for its Core i9-7980XE is 5,828. We'd like to see how the Threadripper fares in a system with a higher-end graphics card than AMD's Radeon RX 460.

And these competitions don't tell you how well the system can sustain the boosted clock speeds on all cores for real work; a consistent 4.8GHz on all cores, for example, is probably more useful than random bursts to 5.1GHz followed by a typical 4GHz for intensive video encoding  or rendering. Or how well it performs in comparison when not overclocked.