小芯片技术可一次性同步助力项目设计突破
By Majeed Ahmad, EDN (June 13, 2023)
What’s the state of chiplet technology today? As the cost advantages of silicon process scaling driven by Moore’s law start to dwindle, will the chiplet approach replace system-on-chip (SoC) designs with multi-die heterogeneous implementations? Are small steps toward implementing chiplet technology sufficient for this landmark semiconductor industry undertaking?
There is no simple answer to these questions yet. But one thing is clear: multi-die architectures are becoming increasingly critical in handling the needs of compute-intensive applications in data centers, cloud computing, and generative artificial intelligence (AI), technologies that require large amounts of memory and fast inter-chip communications.
Then there are automotive and gaming applications that mandate much more reliable and cost-effective solutions than what the current advanced packaging solutions can offer. So, where do the high-performance and highly scalable multi-die architectures for compute-intensive applications actually stand?
To read the full article, click here
Related Semiconductor IP
- NPU IP Core for Mobile
- NPU IP Core for Edge
- Specialized Video Processing NPU IP
- HYPERBUS™ Memory Controller
- AV1 Video Encoder IP
Related News
- 基于Silterra高压制程工艺验证的创飞芯反熔丝一次性可编程解决方案
- 德累斯顿工业大学、曼彻斯特大学和格芯成功流片AI云系统突破性SpiNNaker2,不仅将实时AI延迟降低至毫秒级而且实现云规模级高能效
- Sondrel开发独特建模流程软件,ASIC建模时间缩短至数日
- Fraunhofer IPMS 在 TSN/A 技术研讨会上展示其可扩展TSN 多端口交换机