Page 2 - Hot Chips News Today

Nvidia's Fermi architecture was ambitious and innovative, offering advances in GPU compute along with features like high tessellation performance. However Terascale 2's more traditional approach delivered better power efficiency. Fermi's GTX 500 series refresh narrowed the efficiency gap, but then AMD hit back with the GCN architecture on 28 nm. In developing their first 28…

A mutual love of Labradors and hot dogs inspired a Yakima couple to open Lab Dogz, a restaurant featuring gourmet hot dogs, last month near the Fred Meyer in northwest

Multicore CPUs have to give user programs a way to synchronize between different cores, and ensure a coherent view of memory. A write from one core has to be made visible to others, even though each core has private caches. Cache coherency protocols like MESI, MESIF, or MOESI help make sure this happens. They're typically…

CPU-Z is a hardware information tool from a company called CPUID, not to be confused with the CPUID instruction. Besides showing basic CPU, motherboard, and memory information, CPU-Z features a built-in benchmark. While the benchmark isn't its primary function, it has made its way into some reviews, as well as AMD's slides. Its free, accessible…

NorthPole is an advanced AI chip from IBM Research that integrates processing and memory on a single chip, improving energy efficiency and processing speed for AItasks.

Spanning the full range from huge chips for AI to virtual cores in Minecraft.

A YouTuber called Moore's Law is Dead recently leaked a couple AMD slides about Zen 5. I typically find leaks uninteresting as they are impossible to verify and often don't correspond to reality. One example is leakers expecting RDNA 3 to one-up Nvidia's Ada architecture. AMD is fighting two larger competitors on two fronts and…

Cell phones have to handle telecommunications and audiovisual processing while maximizing battery life. Digital Signal Processors, or DSPs, use specialized hardware to offload these tasks from the CPU which lowers power usage. From an old Qualcomm presentation on Snapdragon 800, showing Hexagon's evolution. Note the large blocks of L2/TCM SRAM Qualcomm's Hexagon has a long…

NVIDIA "Huang's Law" is the primary catalyst for driving chip performance and efficiency to over 1000x in less than a decade.

Huangs Law for GPU performance has delivered 1000x improvements in the last decade and is chiefly driven by ingenuity, says Nvidias Chief Scientist.