NVDA needs more chip capacity to feed the AI monster.
And NVDA needs a domestic (US) supplier of chips to comply with and benefit from the Chips Act.
Intel to the rescue?
And power saving chips preferred. (My take)
Any suggestions?
Intel needs a big customer for its new manufacturing business. A recent comment from Nvidia's CEO offers the beleaguered chip giant some hope.
www.barrons.com
Intel Demos Meteor Lake's AI Acceleration for PCs, Details VPU Unit
For Computex 2023, Intel announced new details about its new AI-focused VPU silicon that will debut in the company's new Meteor Lake chips. The company also outlined its efforts to enable the AI ecosystem for its upcoming Meteor Lake chips. Intel plans to release the Meteor Lake processors, it's first to use a blended chiplet-based design that leverages both Intel and TSMC tech in one package, by the end of the year. The chips will land in laptops first, focusing on power efficiency and performance in local AI workloads, but different versions of the design will also come to desktop PCs.
....The focus here is the VPU unit, but don't let the first image, which is Intel's simplified illustration it shared for today's announcement, mislead you -- the entire tile is not dedicated to the VPU. Instead, it is an SoC tile with various other functions, like I/O, VPU, GNA cores, memory controllers, and other functions. This tile is fabbed on TSMC's N6 process but bears the Intel SoC architecture and VPU cores. The VPU unit doesn't consume this entire die area, which is good -- that would mean Intel was employing nearly 30% of its die area to what will be a not-frequently used unit, at least at first. However, as we'll cover below, it will take some time before developers enable the application ecosystem needed to make full use of the VPU cores.....
The VPU is designed for sustained AI workloads, but Meteor Lake also includes a CPU, GPU, and GNA engine that can run various AI workloads. Intel's Intel says the VPU is primarily for background tasks, while the GPU steps in for heavier parallelized work. Meanwhile, the CPU addresses light low-latency inference work. Some AI workloads can also run on both the VPU and GPU simultaneously, and Intel has enabled mechanisms that allow developers to target the different compute layers based on the needs of the application at hand. This will ultimately result in higher performance at lower power -- a key goal of using the AI acceleration VPU.
Intel's chips currently use the GNA block for low-power AI inference for audio and video processing functions, and the GNA unit will remain on Meteor Lake. However, Intel says it is already running some of the GNA-focused code on the VPU and achieving better results, with a heavy implication that Intel will transition to the VPU entirely with future chips and remove the GNA engine.
Intel also disclosed that Meteor Lake has a coherent fabric that enables a unified memory subsystem, meaning it can easily share data among the compute elements. This is a key functionality that is similar in concept to other contenders in the CPU AI space, like Apple with its M-series and AMD's Ryzen 7040 chips.
...
The industry will face similar challenges in bringing AI acceleration to modern operating systems and applications. However, having the capability to run AI workloads locally isn't worth much if developers won't support the features due to difficult proprietary implementations. The key to making it easy to support local AI workloads is the DirectML DirectX 12 acceleration libraries for machine learning, an approach championed by Microsoft and AMD. Intel's VPU supports DIrectML, but also ONNX and OpenVINO, which Intel says offers better performance on its silicon. However, ONNX and OpenVINO will require more targeted development work from software devs to extract the utmost performance.
Many of today's more intense AI workloads, such as large language models like ChatGPT and the like, require intense computational horsepower that will continue to run in data centers. However, Intel contends that it presents latency and privacy concerns, not to mention adding cost to the equation. Some AI applications, like audio, video, and image processing, will be able to be addressed locally on the PC, which Intel says will improve latency, privacy, and cost.
AI makes a splash with Meteor Lake
www.tomshardware.com