When it comes to memory for compute engines, FPGAs – or rather what we have started calling hybrid FPGAs because they have all kinds of hard coded logic as well as the FPGA programmable logic on a single package – have the broadest selection of memory types of any kind of device out there. …
Intel recently announced that High-Bandwidth Memory (HBM) will be available on select “Sapphire Rapids” Xeon SP processors and will provide the CPU backbone for the “Aurora” exascale supercomputer to be sited at Argonne National Laboratory. …
If the HPC and AI markets need anything right now, it is not more compute but rather more memory capacity at a very high bandwidth. …
We have heard much about the concept of dark silicon but there is a separate, related companion to this idea. …
Increasing parallelism is the only way to get more work out of a system. …
Whether being built for capacity or capability, the conventional wisdom about memory provisioning on the world’s fastest systems is changing quickly. …
A new crop of applications is driving the market along some unexpected routes, in some cases bypassing the processor as the landmark for performance and efficiency. …
As Moore’s Law spirals downward, ultra-high bandwidth memory matched with custom accelerators for specialized workloads might be the only saving grace for the pace of innovation we are accustomed to. …
We have heard about a great number of new architectures and approaches to scalable and efficient deep learning processing that sit outside of the standard CPU, GPU, and FPGA box and while each is different, many are leveraging a common element at all-important memory layer. …
All Content Copyright The Next Platform