[ad_1]
AMD (AMD) is eager to take advantage of an unprecedented manufacturing process in the growing markets of processors and server GPUs.
At an event held in San Francisco on Tuesday morning, AMD officially unveiled the first GPUs to be manufactured using the advanced 7 nanometer (7 nm) manufacturing process of Taiwan Semiconductor (TSM). The company also introduced its first 7nm server processors, as well as shared information on the core architecture of the Zen 2 processor, which will power 7nm products in the PC and server markets.
In addition, AMD announced that Amazon.com (AMZN) had joined the ranks of cloud infrastructure service providers to offer cloud computing instances based on the AMD server's Epyc processors.
The announcements gave a boost to AMD shares, which sold sharply in October thanks to a technical correction and the light Q4 forecasts published in the third quarter report. AMD rose 3.9% Tuesday to $ 20.68. Although far from their September highs, AMD shares have again risen about 100% over the year.
AMD's first 7 nm GPUs, the Radeon Instinct MI60 and MI50, are based on Vega's GPU architecture, launched in 2017, and are designed for a workload for the AI server and the High performance computing (HPC). As such, they address Nvidia's Volta architecture server GPUs (NVDAs), which dominate the accelerator market for AI / AI training, and also represent an important part of the market. HPC accelerators. The MI60 is expected to be delivered by the end of the year and the MI50 by the first quarter of 2019.
The MI60, aided by its support for a healthy memory bandwidth of 1 terabyte per second (TB / s), would produce 14.8 teraflops per second (TFLOPS) with single-precision performance (FP32) and 7.4 TFLOPS with double-precision performance (FP64) performance. While the performance of the FP32 is only slightly improved compared to AMD's MI25 graphics processor, which uses an older 14-nm manufacturing process, the performance of the FP64 is much better. In addition, the chip size of the MI60 is lower than that of the MI25, resulting in power savings and manufacturing costs.
The official performances of the MI60 FP32 and FP64 are close to those of the Tesla V100 GPU, a flagship product of Nvidia, launched in 2017 and using an older 12nm process. It would not be surprising to see Nvidia, which relies both on performance and a strong ecosystem of developers to dominate the artificial intelligence training market (and premium pricing), launches the GPUs of server at 7 nm in 2019.
AMD promises that the 7-nm, named Rome-based server processors, scheduled for launch in 2019, will deliver an impressive performance gain of 2x per processor socket compared to first-generation Epyc server processors. AMD (baptized Naples), thanks to the use of a 7 nm process and Zen 2 cores. The Rome processors will have 64 cores (the first generation of Epyc exceeding 32 cores) and will support two times more memory bandwidth per channel than their predecessors. Like the first Epyc processors, they will present a modular architecture in which several processor chips are housed in the same package.
"I've been in the semiconductor business for a very long time, you do not get it twice," said CEO Lisa Su. She added that AMD was "deeply involved" in the development of its successor, Milan, which received the code name.
AMD claims that its use of 7nm doubles processor core density compared to older chips with its first-generation Zen processor cores (based on 12nm and 14nm processes), as well as reducing 50% energy consumption when the performance is unchanged. Apple's A12 Bionic SoC, which powers iPhone XS and XR, is also based on TSMC's 7 nm process.
The company has not yet released the details of its first 7-nm PC processors and processors. According to speculation, AMD will do so at the CES annual trade show in January, where Su will give a lecture.
At the same time, the AWS partnership follows similar agreements to provide chips for Microsoft's cloud platforms (MSFT), Baidu (BIDU) and Tencent (TCEHY). Amazon, which remains by far the leading player in the cloud computing infrastructure market, deploys three types of cloud computing devices with first-generation Epyc processors.
The AWS agreement "is testament to Epyc's capabilities," said Patrick Moorhead, Technical Analyst. He was impressed by the fact that AWS customers can get a 10% cost reduction by switching to Epyc from similar IT instances based on Intel INTC Xeon processors, without any modification to the software or of the script.
AMD's announcements come one day after Intel launches Xeon server processors based on the next generation platform, Cascade Lake. The Cascade Lake chips, which will be available from the first half of 2019, support up to 48 cores per processor, as well as memory modules incorporating Optane's next generation Intel memory.
Like Intel's older Skylake processors, Cascade Lake is relying on Intel's 14nm manufacturing process node, just like a platform that will succeed it, Cooper Lake, scheduled for later in 2019. After several delays, the first server processors based on Intel's 10 nm node, as competitive as the TSMC 7 nm node, are not expected until 2020.
Intel's shares fell 0.9% to 47.25 dollars on Tuesday. Nvidia shares closed down 0.3% to 211.06 dollars.
Amazon and Microsoft are stakes in Jim Cramer Action Alerts PLUS club member. Do you want to be alerted before Jim Cramer buys or sells AMZN or MSFT? Learn more now
Source link
