memory-based-computing with a natural merge of processor and memory, which might break the bandwidth-bottleneck of the conventional von Neumann architecture. A range of NVM

نویسندگان

  • Yang Shang
  • Wei Fei
چکیده

Hybrid integration of CMOS and nonvolatile memory (NVM) devices has become the technology foundation for emerging nonvolatile memory based computing. The primary challenge to validate a hybrid memory system with both CMOS and nonvolatile devices is to develop a SPICE-like simulator that can simulate the dynamic behavior accurately and efficiently. Since memristor, spin-transfer-toque magnetic-tunnelingjunction (STT-MTJ) and phase-change-memory (PCM) devices are the most promising candidates of next generation of nonvolatile memory devices, it is under great interest in including these new devices in the standard CMOS design flow. The previous approaches either ignore dynamic effect without consideration of internal states for dynamic behavior, or need complex equivalent circuits to represent those devices. This paper proposes a new modified nodal analysis for nonvolatile memory devices with identified internal state variables for dynamic behavior. As such, compact SPICE-like implementation can be derived for all three new nonvolatile memory devices in the design of large-scale memory circuits. As demonstrated by a number of experiment examples on hybrid memory circuits with both CMOS and nonvolatile memory devices, our newly developed SPICE-like simulator can capture dynamic behaviors of memristor, STT-MTJ and PCM devices, and can also reduce CPU runtime by 20 ~ 69 times when compared to the previous equivalent circuit based approaches.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

A High Performance Parallel IP Lookup Technique Using Distributed Memory Organization and ISCB-Tree Data Structure

The IP Lookup Process is a key bottleneck in routing due to the increase in routing table size, increasing traıc and migration to IPv6 addresses. The IP address lookup involves computation of the Longest Prefix Matching (LPM), which existing solutions such as BSD Radix Tries, scale poorly when traıc in the router increases or when employed for IPv6 address lookups. In this paper, we describe a ...

متن کامل

A High Performance Parallel IP Lookup Technique Using Distributed Memory Organization and ISCB-Tree Data Structure

The IP Lookup Process is a key bottleneck in routing due to the increase in routing table size, increasing traıc and migration to IPv6 addresses. The IP address lookup involves computation of the Longest Prefix Matching (LPM), which existing solutions such as BSD Radix Tries, scale poorly when traıc in the router increases or when employed for IPv6 address lookups. In this paper, we describe a ...

متن کامل

Towards Memory Oriented Scalable Computer Architecture and High Efficiency Petaflops Computing

The separation of processor logic and main memory is an artifact of the disparities of the original technologies from which each was fabricated more than fifty years ago as captured by the “von Neumann architecture”. Appropriately, this separation is designated as “the von Neumann bottleneck”. In recent years, the underlying technology constraint for the isolation of main memory from processing...

متن کامل

Three Highly Parallel Computer Architectures and Their Suitability for Three Representative Artificial Intelligence Problems

Virtually all current Artificial Intelligence (AI) applications are designed to run on sequential (von Neumann) computer architectures. As a result, current systems do not scale up. As knowledge is added to these systems, a point is reached where their performance quickly degrades. The performance of a von Neumann machine is limited by the bandwidth between memory and processor (the von Neumann...

متن کامل

Toward on-chip acceleration of the backpropagation algorithm using nonvolatile memory

By performing computation at the location of data, non-Von Neumann (VN) computing should provide power and speed benefits over conventional (e.g., VN-based) approaches to data-centric workloads such as deep learning. For the on-chip training of largescale deep neural networks using nonvolatile memory (NVM) based synapses, success will require performance levels (e.g., deep neural network classi...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2011