Intel has introduced Cascade Lake superior efficiency, a model new class of server CPU primarily based on its current 14nm course of. The first Xeon processors below the Cascade Lake umbrella will launch within the first half of 2019 with one main promoting level: they’ll supply extra cores than ever due to a number of discrete ‘glued’ collectively dies – similar to AMD’s EPYC processors.
Back when AMD first launched EPYC, a supposed inside slide from Intel was leaked that claimed AMD’s server chips have been delivering inconsistent efficiency as a consequence of their “4 glued together desktop die”. It was somewhat stab at AMD’s Infinity Fabric, the interconnect mesh that permits for a number of discrete processing dies to speak and performance collectively.
That similar dig at AMD by no means made it into the press slides for Cascade Lake, and for apparent purpose. Intel’s present era Xeon processors – the aptly-named Xeon Scalable – depend on a single die stuffed to the brim with cores. The present high chip is the Xeon Platinum 8180, a 28-core, $10,000 processor utilising a single, fully-functional XCC die package deal.
But Intel could have given in to AMD’s scalable answer with Cascade Lake. Intel is providing as much as 48 cores per chip with the upcoming era, made attainable by a multi-chip module ostensibly not conceptually dissimilar to AMD’s EPYC chips. The superglue method will seemingly see Intel pairing two 24-core packages collectively utilising a excessive pace interconnect.
While Intel’s ‘glued together’ phrasing would possibly initially have sounded unnecessarily derogatory, the time period is usually used when particularly speaking about technical interconnects, and never at all times negatively.
And Intel already has fairly a little bit of expertise with interconnect expertise. The firm utilises mesh interconnects inside present Xeon traces to cut back latency between bodily cores, Ultrapath Interconnect expertise to meld two discrete server CPUs collectively, and its embedded multi-die interconnect bridge carried out again with Kaby Lake G. This expertise bridged the hole between AMD’s Radeon Vega GPU and HBM2 reminiscence on the hybrid chip.
AMD’s EPYC processors can be found with as much as 32 cores of processing grunt. These chips are reportedly in trend proper now as Intel struggles to satisfy its shoppers wants as a consequence of 14nm shortages.
Cascade Lake is to be constructed upon the 14nm node, as have been the previous few generations of Intel server processors. This is essentially as a consequence of a slowdown in growth of its 10nm course of – a severely delayed node not anticipated to reach till the tip of 2019. This has considerably stifled not simply product development over the previous few years, however, extra lately, provide, too.
HP Enterprise, one of many world’s preeminent server corporations, reportedly started recommending AMD EPYC chips in lieu of Intel’s Xeon processors. Intel’s third quarter financials point out the corporate has since stabilised, or a minimum of floated, the provision constraints, however AMD’s nonetheless wanting sturdy within the information centre market after years of Intel dominance.
Intel’s Cascade Lake processors might be out there someday subsequent 12 months, and the highest chip nearly definitely costing upwards of $10,000. For the time being Intel is introducing the entry-level Intel Xeon E-2100, a lowly hexacore half. Meanwhile, AMD is probably solely sooner or later away from saying the full details of its 7nm EPYC ‘Rome’ chips.