In addition to Weibo, there is also WeChat
Please pay attention
WeChat public account
Shulou
2025-01-24 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >
Share
Shulou(Shulou.com)11/24 Report--
CTOnews.com, December 19 (Xinhua)-- Micron and AMD have set up a joint server lab in Austin to reduce server memory verification time and jointly conduct workload testing during product verification and release, according to Micron.
At present, Meguiar's DDR5 memory and fourth-generation AMD EPYCTM processors for data centers have been shipped, and officials have conducted some common high-performance computing (HPC) workload benchmark tests.
For a long time, supercomputers bear the workload of high-performance computing. Such large-scale data-intensive workloads need to run TB-level data volumes to perform millions of parallel operations to solve human world problems such as weather and climate prediction, seismic modeling, and chemical, physical and biological analysis.
With the development of computer architecture, such workloads are often hosted in very large "scale-out" high-performance server clusters. These server clusters need to combine the most powerful computing power, architecture, memory, and storage infrastructure to meet the scalability, low latency, and high performance needs of critical workloads. However, with the continuous growth of server CPU performance and throughput, DDR4 cannot provide sufficient memory bandwidth to meet the growing high-performance kernel.
To alleviate this bottleneck, Meguiar DDR5 memory is combined with fourth-generation AMD EPYC processors with Zen 4 server architecture to enable server CPU to better match memory products to meet the performance and efficiency needs of data-intensive workloads. CTOnews.com was informed that Micron conducted an industry-wide high-performance computing workload benchmark test on the latest AMD Zen 496-core CPU and Meguiar DDR5, all of which showed a tripling of performance.
STREAM1 is a common benchmark tool for measuring the memory bandwidth of high-performance computers and capturing the peak memory bandwidth of high-performance computing systems.
The software stack used by this workload
● Alma 9 Linux kernel 5.14
Released on November 29, 2008 by ● STREAM.f,2021
Test Settin
● DDR4 system with the third generation 64-core 3.7GHz AMD EPYC processor; DDR4 3200 MHz system 2's RDIMM memory slot is full, a total of 64GB
● DDR5 system with fourth generation 96-core 3.7GHz AMD EPYC processor; RDIMM memory slot of DDR5 4800 MHz system 3 is full, total 64GB
Test result
● DDR5 system memory bandwidth per slot doubled to 378GB / s
● this result means that customers can run larger artificial intelligence / machine learning (AI / ML) projects or take advantage of the increased memory bandwidth of DDR5 for more high-performance computing.
The high-performance computing workload code used in this test is for weather and climate. The WRF model performs well in some traditional high-performance computing architectures such as high-performance floating-point processing, high memory bandwidth and low-latency networks, and is tested in the continental United States (CONUS) with a horizontal resolution of 2.5km.
The software stack used by this workload
● Alma 9 Linux kernel 5.14
● WRF 2.3.5 & 4.3.3
● Open MPI v4.1.1
Test Settin
● DDR4 system with the third generation 64-core 3.7GHz AMD EPYC processor; DDR4 3200 MHz system 2's RDIMM memory slot is full, a total of 64GB
● DDR5 system with fourth generation 96-core 3.7GHz AMD EPYC processor; RDIMM memory slot of DDR5 4800 MHz system 3 is full, total 64GB
Test result
● Meguiar DDR5 with the fourth generation AMD EPYC processor can achieve 2.8533 time steps / second of 1.3567 time steps / second VS DDR4 system.
Faster ● means that you can use larger databases or run more models to make weather forecasts, thereby improving the accuracy of forecasts.
OpenFOAM is an open source high-performance computing workload of Computational fluid Dynamics (CFD), widely used in many industries, helping to shorten development time and reduce costs. From consumer product design to aerospace design, OpenFOAM can simulate physical interactions in different applications, including motorcycle windshield turbulence.
In this simulation, OpenFOAM can calculate the steady air flow around the motorcycle and the rider. OpenFOAM can do load balancing calculation according to the number of processes specified by users, so that the grid can be divided into multiple parts and assigned to different processes to solve. After the solution is completed, the grid is reassembled into a single domain.
The software stack used by this workload
● OpenFOAM CFD software (version 8), in which the motorcycle grid size is: 600x240x240
● Alma 9 Linux kernel 5.14
● Open MPI v4.1.1
Test Settin
● DDR4 system with the third generation 64-core 3.7GHz AMD EPYC processor; DDR4 3200 MHz system 2's RDIMM memory slot is full, a total of 64GB
● DDR5 system with fourth generation 96-core 3.7GHz AMD EPYC processor; RDIMM memory slot of DDR5 4800 MHz system 3 is full, total 64GB
Test result
The test results show that Meguiar's DDR5 product portfolio improves the performance of OpenFOAM by 2.4 times. OpenFOAM is one of the top five high-performance computing software platforms with a large open source community. The software is widely used in universities and R & D centers, and can make use of high-bandwidth memory and high-performance CPU with dense kernel to achieve a high degree of parallel operation.
CP2K is an open source quantum chemistry tool for many applications, including solid-state biological system simulation. CP2K can provide a general framework for different modeling methods. The object of this test is the density functional theory (DFT) of water (H2O). The simulation box contains 6144 atoms (2048 water molecules).
The software stack used by this workload
● H2O-DFT-LS.NREP4 and H2O-DFT-LS
● Alma 9 Linux kernel 5.14
Test Settin
● DDR4 system with the third generation 64-core 3.7GHz AMD EPYC processor; DDR4 3200 MHz system 2's RDIMM memory slot is full, a total of 64GB
● DDR5 system with fourth generation 96-core 3.7GHz AMD EPYC processor; RDIMM memory slot of DDR5 4800 MHz system 3 is full, total 64GB
Test result
The test results show that Meguiar's DDR5 product portfolio can improve the molecular dynamics performance by 2.1 times. As the number of cores and memory bandwidth increases, so does the performance of such workloads.
Summary
At present, only a small number of high-performance computing workloads are tested, so the above are only preliminary results. Combining high-performance, high-bandwidth memory with the latest server processors, such as fourth-generation AMD EPYC processors, creates new possibilities for high-performance computing customers.
1 STREAM Benchmark-- with 2.5 billion vectors configured in the STREAM benchmark runs on a single AMD CPU system
2 AMD DDR4 system is a 64-core AMD EPYC 7763 processor, DDR4-3200 MHz RDIMM memory slot is full, a total of 64GB
3 AMD DDR5 system is a 96-core AMD EPYC 9654 processor, DDR5-4800 MHz RDIMM memory slot is full, a total of 64GB
(4) the running time of WRF with a horizontal resolution of 12.5km CONUS is 929s on DDR4 system and 287s on DDR5 system (including the input / output time of memory). In this test, WRF is configured as 2.5km CONUS, and the test result is 1.3567 time steps per second, compared with 2.8533 time steps per second for DDR4.
5 three variants were run for OpenFOAM:
The running time of 5a:1004040 runtimes,DDR4 system is 1144 seconds, and that of DDR5 system is 478 seconds.
The running time of 5b:1084646 runtimes,DDR4 system is 1633 seconds, and that of DDR5 system is 698 seconds.
The running time of 5c:1305252 runtimes,DDR4 system is 2522 seconds and that of DDR5 system is 1091 seconds.
6 the running time of molecular dynamics workload is 2519 seconds on DDR4 system and 1242 seconds on DDR5 system.
Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.
Views: 301
*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.
Continue with the installation of the previous hadoop.First, install zookooper1. Decompress zookoope
"Every 5-10 years, there's a rare product, a really special, very unusual product that's the most un
© 2024 shulou.com SLNews company. All rights reserved.