FPGA | 現場可程式化邏輯閘陣列
What is it?
It may surprise you to know that many of the integrated circuits (or "microchips") we install in our PCs and servers were built from the beginning to serve a very specific purpose. Such chips are known as ASIC, or "application-specific integrated circuit". A prime example of an ASIC is the early iteration of the GPU, a processing unit designed specifically to render graphics—though the advent of GPGPUs has certainly blurred that line. The flip side of the ASIC is the FPGA, or "field-programmable gate array". This is an integrated circuit that can be reprogrammed by the customer after manufacturing to perform different functions. The FPGA is able to accomplish this due to the integration of reprogrammable logic gates in its architecture, as well as careful "floorplanning" that enables flexible resource allocation.
To use an everyday example to illustrate our point, you may think of ASIC chips as pieces of furniture you have in your home: a chair is designed to serve as a chair, a table is designed to serve as a table, and so on. An FPGA chip is like a piece of furniture you can reassemble to serve as a chair or table or bed or closet—it is just that versatile. The only drawback is that FPGA chips usually command a heftier price than ASIC chips, just as you might imagine a piece of multifunctional furniture to be.
To use an everyday example to illustrate our point, you may think of ASIC chips as pieces of furniture you have in your home: a chair is designed to serve as a chair, a table is designed to serve as a table, and so on. An FPGA chip is like a piece of furniture you can reassemble to serve as a chair or table or bed or closet—it is just that versatile. The only drawback is that FPGA chips usually command a heftier price than ASIC chips, just as you might imagine a piece of multifunctional furniture to be.
Why do you need it?
FPGA is mainly positioned as a source of adaptable and accelerated computing that will enable future breakthroughs in data center technology. Accelerators based on FPGA in your server set-up can complement other processors through heterogeneous computing or parallel computing, enabling you to complete tasks faster. Since FPGA is reprogrammable, you will also be able to repurpose its functions to meet the constantly changing needs of the modern data center or server room; this is important if you want to develop groundbreaking prototypes or incorporate ready-made computing solutions that will shorten your time to market (TTM).
How is GIGABYTE helpful?
GIGABYTE Technology has long been a pioneer of FPGA applications. Back in the mid-aughts, GIGABYTE introduced the i-RAM, which was solid state storage that used Xilinx FPGA. In recent years, GIGABYTE has worked closely with FPGA market leaders such as Intel and Xilinx (which has been acquired by AMD) to incorporate FPGA-based accelerators in its server solutions. For example, GIGABYTE's H-Series High Density Servers, G-Series GPU Servers, and R-Series Rack Servers all offer the acceleration card density and supportability to accommodate FPGA accelerators. This makes them the ideal solutions for deep learning, AI inferencing, high performance computing (HPC), and more.
WE RECOMMEND
RELATED ARTICLES
Tech Guide
如何挑選您的AI伺服器?(上)CPU和GPU
生成式人工智慧和其他AI工具盛行的當下,挑選合適的AI伺服器成為各產業的首要任務。技嘉科技最新發表《科技指南》,帶領讀者認識AI伺服器的八個關鍵零組件,本篇從最重要的元件開始,即中央處理器(CPU)和圖形處理器(GPU)。挑選適當的運算晶片,打造量身訂做的人工智慧超算平台,可以讓工作事半功倍,為使用者開創全新的巔峰。
Tech Guide
CPU vs. GPU:淺談伺服器的兩大運算力
順應科技趨勢,業界持續追求運算力更強大的伺服器,其中的關鍵元素除了大家熟知的中央處理器CPU之外,圖形處理器GPU近年來也受到重視;但您是否知道,什麼是GPU?它與CPU的差別何在?又該如何運用兩種不同類別的處理器,發揮最強大的運算力?技嘉科技是伺服器與尖端科技解決方案的知名品牌,發表本篇《科技指南》,目的是比較CPU與GPU的原理與功能,進一步協助您挑選合適的技嘉伺服器產品,將「GPU運算力」導入您的伺服器機房與資料中心。