G-Dep Advance Co., Ltd. Announcement of start of handling of NVIDIA(R)H200 NVL 141GB PCIe

[G-Dep Advance Co., Ltd.] Announcement of start of handling of NVIDIA(R)H200 NVL 141GB PCIe ​
G-DEP Advance Co., Ltd. Press release: November 12, 2024 Announcement of start of handling of
NVIDIA(R)H200 NVL 141GB PCIe
https://prcdn.freetls.fastly.net/release_image/90150/23/90150-23-c1138acf832211ed8c30c1b57c29b25b-1030×400.jpg

GDEP Advance, Inc. (Headquarters: Chuo-ku, Tokyo, Representative Director and CEO Masamichi Iino, TSE Standard Stock Code: 5885), an AI and visualization solutions company, announced on November 12, 2024 ( Tuesday), the GPU card “NVIDIA(R)H200 NVL 141GB” equipped with Hopper architecture from NVIDIA, Inc. We have started accepting orders for PCIe. NVIDIA(R) H200 NVL 141GB PCIe is equipped with GPU architecture “Hopper”. It is a GPU optimized for AI generation with a large capacity of 141GB and wide bandwidth performance of 4.8TB/s by adopting HBM3e memory. The inference speed of the large-scale language model “Llama2 70B” with 70 billion parameters is 1.9 times faster compared to the previous model, NVIDIA H100, and the inference speed of the large-scale language model “GPT3-175B” with 175 billion parameters is 1.6 times faster. For HPC applications, CPU (Xeon(R) Platinum 8480 2CPU) Achieves 110 times faster speed than . This has made it possible to perform advanced computational processing such as the development of large-scale language models and complex simulations in HPC in a short time. It also includes a five-year subscription to NVIDIA AI Enterprise, which is essential for developing and supporting applications built with NVIDIA’s extensive framework libraries, providing a smoother AI development environment. NVIDIA(R) H200 NVL 141GB PCIe achieves a vast memory space of 282GB by connecting two H200 NVLs with NVLink(TM), and each GPU can communicate with a wide bandwidth of 900GB/s, so It has high-end specs that are ideal for learning large-scale AI models such as generative AI. NVIDIA(R) H200 is more than just a high-performance GPU; it can quickly accelerate AI projects and increase competitiveness in fields that aim for strategic use, such as for companies, startups, and researchers. By constantly providing the latest AI development environment, G-Dep Advance will accelerate AI development in all industries and contribute to the creation of a prosperous society. About NVIDIA H200 NVL 141GB PCIe Product name: NVIDIA H200 NVL 141GB PCIe Sales price: 1 piece ¥6,000,000 (excluding tax and shipping) 2 sheets NVLink set ¥12,299,571 (excluding tax and shipping) Product page URL:
https://www.gdep.co.jp/products/list/v/67296ae729fd7/ ​ *NVIDIA H200 NVL 141GB PCIe does not guarantee operation on servers other than those certified by NVIDIA. Please contact our sales department for compatible models. NVIDIA H200 NVL 141GB PCIE Specifications Product name NVIDIA H200 NVL 141GB PCIe Architecture Hopper Process rules 4nm(TSMC) Double precision performance | FP64/Tensor Core FP64 34 TFLOPS/67 TFLOPS Single precision performance | FP32/Tensor Core TF32 67 TFLOPS/495 TFLOPS Bfloat16 BF16 134 TFLOPS/989 TFLOPS
Half-precision performance FP16/Tensor Core FP16 134 TFLOPS/989 TFLOPS Tensor Core INT8 1,979 TOPS Integer performance 31 TOPS GPU memory 141GB HBM3e Memory bandwidth 4.8TB/s Interconnect PCI-Express 5.0 x16 : 128GB/s 4th generation NVLink: 900GB/s (When using NVLink Bridge with 2GPU) Form factor PCIe Dual Slot Multi-instance GPU (MIG) Up to 7GPU Maximum thermal design power 600W Cooling method Passive NVLINK bridge 900GB/s (4-way) NVIDIA AI Enterprise 5 years Guarantee 3 years What’s included Warranty *Auxiliary power cable is not included ■ About G-Dep Advance Co., Ltd. G-Dep Advance Co., Ltd. (TSE Standard, stock code: 5885) is an “Advance with With the mission of “You”, we provide work and support services, from the provision and construction of operational tools such as GPGPU and other accelerators, high-end workstations, broadband networks and high-speed storage, as well as operational tools such as libraries, compilers, and job schedulers. We are an AI and visualization solution provider that comprehensively provides the means to advance research in all forms, including on-premises, rental, and cloud. We are active as an elite partner in NVIDIA’s NPN (NVIDIA Partner Network). URL: https://www.gdep.co.jp/ *Company names, product names, etc. listed are registered trademarks or trademarks of each company. *The information listed is as of the time of release. Specifications and prices may change or sales may be discontinued or postponed without notice. ■Inquiries regarding products and services Gdep Advance Co., Ltd. Solution Department E-mail: sales@gdep.co.jp