Intel Server Roadmap Pegs DDR5 & PCIe 5.0 for 2021
A recent server roadmap was leaked from a Huawei presentation (see below for irony), and it outlines some serious I/O advancements headed for the datacenter between now and 2022. The roadmap introduces new codenames Sapphire Rapids and Granite Rapids, set to succeed both Cooper Lake and Ice Lake in the server segment.
The Sapphire Rapids SP seems poised for a 1Q21 launch, and brings with it DDR5 memory support and PCIe 5.0 bandwidth. Sapphire Rapids will make use of the Eagle Stream platform, a 2P machine platform that will also be used for Granite Rapids. No word on core counts or frequencies for Sapphire Rapids yet. Sapphire Rapids will be succeeded by Granite Rapids in 2022, and could bring generation-over-generation improvements in regards to silicon, clockspeeds, and efficiency.
It appears both processor families will concur with Intel’s new 7nm nodes, so it could be Sapphire Rapids will be based on Intel’s first generation of 7nm, while Granite Rapids could leverage 7nm+.
Tech Companies Cut Ties with Huawei Amidst US-China Tensions
In what’s been an unprecedented move in both the tech sector and the Android landscape, Huawei has found itself on a commerce trade blacklist, compliments of the US Federal Government. Following Huawei’s addition to the entity list, several important companies have been forced to sever ties.
Among the most critical are Google and ARM, which will absolutely cripple Huawei’s ability to ship phones and other Android devices. Google announced that it will revoke Huawei’s Android license, thus limiting Huawei to using the Android Open Source Project (AOSP). Without access to Google’s proprietary services like apps and Play Store, this would make any fork of AOSP unrecognizable.
Huawei has silently been working on its own operating system, as the company anticipated losing Google. However, losing ARM is nothing short of devastating. Nearly every chip inside a smartphone or tablet is based on ARM designs, and without some foundation or IP on which to base chips, Huawei’s chip building ambitions are effectively dead.
Also on the growing list of companies turning away from Huawei is Intel, Qualcomm, Broadcom, and Xilinx, all pledging not to supply the company until further notice. Huawei has suggested it could replace its American suppliers with those found in China, but that’s a dubious claim to say the least.
There is a chance Huawei could potentially replicate some components, but it’s unlikely it could be done at the scale Huawei would need to do it at -- Huawei is the second largest phone maker in the world, behind Samsung. More to the point, Huawei needs access to American software, IP, and technology for chips. China simply can’t develop chips without American technology, at least not yet.
Stay tuned, as there will surely be much more to this story.
Huawei Ban Has US Tech Companies Caught in Crossfire
Due to Huawei’s trade blacklist, several American suppliers are expecting losses, with some being more exposed than others. As American companies have to cease any and all trade with the besieged Huawei, this is obviously catastrophic for Huawei’s business -- but the trade ban is one that cuts both ways.
Broadcom, Qualcomm, Micron, Intel, AMD, Microsoft and more are all set to see serious revenue impacts. Early estimates point to an $11 billion loss for American suppliers, collectively. AMD seems to be the most exposed to loss, as it doesn’t enjoy the absolute market share dominance that Intel, Nvidia, Broadcom, and Microsoft do.
AMD will lose about $39 million, which doesn’t sound like much compared against Intel’s $85 million; the difference is AMD’s sales to Huawei account for a much greater percentage of its revenue than that of Intel’s. Intel’s losses only account for about 1% of its total revenue.
Broadcom stands to lose the most in terms of absolute revenue at $508 million, but its exposure is far from the highest.
Intel Ends Samsung’s Semiconductor Reign After One Year
As IC Insights reports, Intel is once again the number one semiconductor vendor in the world, dethroning Samsung, who defeated Intel for the crown in 2017. This is in no small part thanks to the abrupt downturn of the DRAM market, offsetting Intel’s CPU shortage.
For Samsung, the writing has been on the wall for some time; the company went as far as warning investors about its 1Q19 revenue greatly missing the mark due to massive price declines in memory. Samsung’s lead was built on the massive memory business it was enjoying during the several quarters of booming memory prices, something GN looked into extensively.
So, Intel’s propulsion back to the top was almost expected -- even if it may be something of a default.
HPE Will Acquire SuperComputer Pioneer Cray
Hewlett Packard Enterprise and Cray have jointly confirmed that HPE will buy the supercomputer maker for $1.3 billion.
Antonio Neri, President and CEO of HPE, stated “Answers to some of society’s most pressing challenges are buried in massive amounts of data. Only by processing and analyzing this data will we be able to unlock the answers to critical challenges across medicine, climate change, space and more. Cray is a global technology leader in supercomputing and shares our deep commitment to innovation. By combining our world-class teams and technology, we will have the opportunity to drive the next generation of high performance computing and play an important part in advancing the way people live and work.”
Cray got its start in 1972, with the Cray-1 supercomputer. Since then, Cray is often associated with the most powerful supercomputers on the planet, including the recently announced Frontier, slated to be the world’s first exascale supercomputer. HPE will leverage Cray and its associated technologies to address the emerging HPC (High Performance Computing) and exascale market segments, where HPE intends to bolster its HPC-as-a-Service offerings.
HPE also intends to build an end-to-end HPC portfolio consisting of storage, software, compute, and interconnects. This will allow HPE to extend its footprint across a broader set of markets.
Adata & Micron Leapfrog Towards 6GHz DDR4
Micron and Adata seem content to keep trading the crown for world’s fastest kit of DDR4 memory, and all the bragging rights afforded therein. Just recently, Micron laid claim to the world record for DDR4 memory frequency with a HWBot validated overclock of 5,726 MHz, blowing past Adata’s previous record of 5,634 MHz.
Now, Adata is back to leapfrog Micron’s score, with an impressive overclock of 5,738MHz, also validated by HWBot. Adata achieved the overclock with a kit of its own Spectrix D60G RGB DDR4 memory.
Obviously, these records are set using LN2, of which by now many of you have likely seen GamersNexus using for our own overclocking ambitions. Aside from the bragging rights, memory makers love these overclocking competitions because it’s great self promotion. For many years, Samsung’s B-die memory were the ICs of choice for many an overclocker. However, with Samsung discontinuing the B-die chips this year, there’s room at the top for the next gold standard.
Micron and Adata are certainly making their case.
Editorial: Eric Hamilton
Host: Steve Burke
Video: Andrew Coleman, Keegan Gallick