AMD’s New Instinct MI100 Accelerator to Boost HPC Workloads


Advanced Micro Devices


AMD

recently

unveiled

AMD Instinct MI100 accelerator chips, which are built on its CDNA architecture and feature 120 compute units. The chips are based on the 7 nanometer (nm) technology.

The chips will be capable of speeding up high performance computing (HPC) tasks in scientific research amid emergence of exascale computing when used in combination with AMD EPYC processors and the ROCm 4.0 open software platform. The processors are backed by upgraded computers from Dell, Supermicro, Gigabyte and Hewlett Packard.

Robust features to drive adoption

Per AMD, MI100 accelerator is the first x86 server GPU to exceed the 10 teraflops (FP64) performance limit. The GPU also provides peak performance up to 11.5 TFLOPS (FP64) for HPC.

MI100 accelerator boasts 32 GB high band width HBM2 memory with clock rate of 1.2Ghz. The chips are based on PCIe Gen 4.0 technology and has peak transport data bandwidth (theoretical) of 64GB/s from CPU to GPU. It also has memory bandwidth of 1.23 TB/s to facilitate management of huge data sets.

The impressive features are likely to result in boosting adoption of these new GPU processors in the market amid increasing demand for HPC applications.

Per

Grand View Research Report

, the global market for HPC is forecast to witness a CAGR of 6.5% between 2020 and 2027 driven by migration to cloud and emergence of exascale computing. Exascale refers to computing power capable of a quintillion (1018) calculations per second.

It will enable AMD to fortify its competitive positioning against

Intel


INTC

and

NVIDIA


NVDA

and boost investors’ confidence in the company’s stock.

Notably, shares of AMD have rallied 82.5% in the year-to-date period against the

industry

’s return of 36.3%.

AMD’s Azure Partnership

AMD also announced that

Microsoft’

s

MSFT

Azure platform will use the next generation EPYC processors (dubbed Milan) for its HB series virtual machines (VM) for HPC workloads.

At present, Microsoft is deploying EPYC (second generation) chips for its HBv2 virtual machines (VMs) that are used for HPC tasks.

AMD stated that it expects to begin shipping third generation of EPYC chips in the first quarter of 2021. However, AMD will start shipping EPYC (third generation) processors to few customers, (names were not provided), based on OEM availability, in the fourth quarter of 2020.

AMD

continues

to benefit from strong momentum witnessed in the adoption of EPYC processors, especially second generation, across virtualized and HCI architecture.

Further, the momentum is likely to be driven by rapid growth in the cloud computing market. Per a

Verified Market Research Report

, cloud computing market is projected to witness a CAGR of 13.9% between 2020 and 2027.

The EPYC processors facilitate

efficient running

of virtual desktop workloads and help to sustain a steady performance with comparatively lesser number of servers.

However, increasing expenses on product development amid stiff competition in the processors’ market is likely to lead to pricing pressure and limit margin expansion in the near term.

Currently, AMD currently carries a Zacks Rank #3 (Hold).

You can see


the complete list of today’s Zacks #1 Rank (Strong Buy) stocks here.

Legal Marijuana: An Investor’s Dream

Imagine getting in early on a young industry primed to skyrocket from $17.7 billion in 2019 to an expected $73.6 billion by 2027.

Although marijuana stocks did better as the pandemic took hold than the market as a whole, they’ve been pushed down. This is exactly the right time to get in on selected strong companies at a fraction of their value before COVID struck. Zacks’ Special Report, Marijuana Moneymakers, reveals 10 exciting tickers for urgent consideration.



Download Marijuana Moneymakers FREE >>

Want the latest recommendations from Zacks Investment Research? Today, you can download 7 Best Stocks for the Next 30 Days.

Click to get this free report


To read this article on Zacks.com click here.


Zacks Investment Research