Technology

Gazzillion Misses™
Unlocking High-Bandwidth AI Performance

A breakthrough memory management technology that eliminates bottlenecks and maximizes throughput for AI and HPC workloads  

Unrivaled Memory Bandwidth

Optimized for Low-Latency, High-Bandwidth

Seamless Data Flow

Seamless Data Flow

Supports up to 128 simultaneous memory requests, ensuring continuous data flow

High-Performance Memory

Optimized for high-bandwidth, memory-intensive applications such as AI inference, key-value stores, and sparse data processing

Stall-Free Pipelines

Eliminates traditional stop-and-go latency stalls, keeping AI pipelines saturated

Multiprocessor Ready

Scalable, Efficient, and Flexible for AI

lorem ipsum

Atrevido Ready

Works with Atrevido core to maintain peak performance

From Edge to HPC

Scalable from edge AI to high-performance compute clusters

Ideal for

AI acceleration, machine learning, inference engines, high-speed data analytics, and memory-bound HPC workloads

What is Gazzillion Misses™ technology?

Built into our fully customisable 64-bit RISC-V family cores and meant for memory intensive applications, Gazzillion Misses™ helps to avoid latency issues when accessing off-chip memory

F=Fetch, D=Decode, R=Rename, I=Issue, A=ALU or Address Generation, M=Cache Miss. In the figure, after five cache misses, the processor stops and needs to wait until data returns from memory. In the example, showing a memory 50 clocks away, the processor will be idling doing nothing for almost 100 clocks

Figure 1, Traditional cores waste 100s of clock cycles waiting for data due to cache misses

F=Fetch, D=Decode, R=Rename, I=Issue, A=ALU or Address Generation, M=Cache Miss. In the figure, the processor can make up to 128 cache misses, thereby hiding all the 100 clocks of memory latency. Processing continues while the first (and following) cache misses are sent to the memory system

Figure 2, Semidynamics Gazzillion™ technology allows the processor to continue doing useful processing while previous misses are served

Figure 3, Atrevido stand out from all other cores with their ability to process up to 64 or 128 misses respectively

To learn more about how Gazzillion Misses™ can enhance your system's performance, read our detailed blog post