[et_pb_section fb_built=”1″ admin_label=”section” _builder_version=”4.16″ custom_padding=”0px||0px||false|false” global_colors_info=”{}”][et_pb_row admin_label=”row” _builder_version=”4.16″ custom_padding=”||0px||false|false” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.16″ custom_padding=”|||” global_colors_info=”{}” custom_padding__hover=”|||”][et_pb_text admin_label=”Breadcrumb” _builder_version=”4.16″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n
Scientific Computing and Data<\/a>\u00a0\/\u00a0High Performance Computing<\/a> \/ Documentation<\/a> \/\u00a0gpgpu<\/p>\n [\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ admin_label=”section” _builder_version=”4.16″ custom_padding=”0px||0px||false|false” global_colors_info=”{}”][et_pb_row _builder_version=”4.16″ _module_preset=”default” custom_margin=”|auto|-20px|auto|false|false” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.16″ _module_preset=”default” global_colors_info=”{}”][et_pb_text admin_label=”Title” _builder_version=”4.16″ _module_preset=”default” header_font=”|700|||||||” header_text_color=”#221f72″ custom_margin=”||0px||false|false” custom_padding=”||0px||false|false” global_colors_info=”{}”]<\/p>\n [\/et_pb_text][\/et_pb_column][\/et_pb_row][et_pb_row admin_label=”row” _builder_version=”4.16″ background_size=”initial” background_position=”top_left” background_repeat=”repeat” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.16″ custom_padding=”|||” global_colors_info=”{}” custom_padding__hover=”|||”][et_pb_text admin_label=”Overview” _builder_version=”4.27.4″ text_line_height=”1.5em” hover_enabled=”0″ global_colors_info=”{}” sticky_enabled=”0″]In modern computing, GPUs (Graphics Processing Units) have evolved far beyond their initial role as graphics rendering accelerators like 3D rendering and image processing. This evolution has led to the emergence of GPGPU (General-Purpose computing on Graphics Processing Units), a paradigm shift where GPUs are utilized for a broader range of computational tasks traditionally handled by CPUs.<\/p>\n The NVIDIA CUDA toolkit includes everything needed to develop GPU-accelerated applications, including GPU-accelerated libraries, a compiler, development tools, and the CUDA runtime.\u00a0NVCC (NVIDIA CUDA Compiler), the compiler driver, acts as a high-level interface, simplifying the CUDA compilation process for developers. NVCC separates the CUDA code into two parts: host code (for the CPU) and device code (for the GPU), and sends the host code to a standard C++ host compiler (like GCC or Clang) and\u00a0the device code to the CUDA backend compiler. NVCC handles the creation of fat binaries, which bundle the compiled device code for different GPU architectures within the host executable. NVCC manages the linking of the compiled host code, the device code, and necessary CUDA runtime libraries.<\/p>\n Minerva has a total of 81 GPU nodes. Please check below for more information.<\/p>\n [\/et_pb_text][et_pb_text admin_label=”Requesting GPU Resources” _builder_version=”4.27.4″ text_line_height=”1.5em” global_colors_info=”{}”]<\/p>\n The GPU nodes must be accessed by way of the LSF job queue. There are nodes on the To request GPU resources, use the Full details of the By default, GPU resource requests are exclusive_process<\/i>.<\/strong><\/p>\n The available GPU models and corresponding resource specifications are:[\/et_pb_text][et_pb_code admin_label=”Table” _builder_version=”4.27.4″ _module_preset=”default” position_origin_a=”center_center” position_origin_f=”center_center” text_orientation=”center” width=”50%” max_width=”50%” custom_margin=”20px||30px|150px|false|false” custom_padding=”|||200px|false|false” custom_css_main_element=”.centered-content {|| display: flex;|| justify-content: center; \/* Center horizontally *\/|| align-items: center; \/* Center vertically *\/|| height: 100%; \/* Ensure full height is available *\/|| text-align: center; \/* Optional: Center the text within the content *\/||}||||.centered-table {|| margin-left: auto;|| margin-right: auto;|| text-align: center;||}|| ||.centered-table td, .centered-table th {|| text-align: center;|| vertical-align: middle;||}” global_colors_info=”{}”]<\/p>\nHarnessing GPUs for Scientific Computing: Concepts and Usage Etiquette<\/h1>\n
\n
Requesting GPU Resources<\/h2>\n
interactive<\/code>, gpu<\/code> and gpuexpress<\/code> queues.<\/p>\n-gpu<\/code> option of the bsub command. At a minimum, the number of GPU cards requested per node must be specified using the num=<\/code> sub-option. The requested GPU model should be specified by using an LSF resource request via the -R<\/code> option.<\/p>\nbsub -gpu num=2 -R a100<\/code><\/pre>\n-gpu<\/code> option can be obtained\u202fhere<\/a>.<\/strong><\/p>\n