08 / Industry · Analysis
Intel D270 extends support to high-performance computing
TechM reported that Intel's D270 supports low-precision AI inference and double-precision calculations used in science, engineering and finance.
Intel unveiled high-performance computing support for its D270 GPU, which uses the Xe 3P architecture, has 160GB of memory and contains 32 Xe cores, according to TechM, a Korean technology publication.
TechM reported that the expansion brings calculations requiring numerical accuracy within the scope of a product designed for AI inference. Its support spans low-precision inference and double-precision computing.
According to TechM, Intel also released initial performance test results alongside the high-performance computing features. The announcement covered both the hardware configuration and the calculation formats it supports.
Crescent Island's configuration
The D270 carries the codename Crescent Island and includes 256 XMX engines in its computing configuration, TechM reported.
According to TechM, Intel's own-brand card uses LPDDR5X memory. This is a low-power type of memory that holds data for processing.
Double precision broadens the scope
TechM reported that the D270 natively supports FP4 and FP8, low-precision calculation formats used for AI inference. Native support means the hardware can process those formats directly.
The D270 also natively supports FP64 double-precision calculations, according to TechM. That support extends its scope to calculations in science, engineering and finance where numerical accuracy matters.