At the annual Taipei International Computer Show in Taiwan, Intel announced the pricing of a new generation of data center processors, an AI accelerator suite, and the architecture of AI PC chips. Intel claims that its new Xeon 6 processors deliver higher performance and power efficiency for high-density, scale-out workloads in the data center than previous generations of chips.
Intel said that the main highlight of the Gaudi 2 and Gaudi 3 AI accelerator kits announced at the Taipei event is that they are high-performance and affordable, and their cost is only one-third of that of other competitors. "The combination of Xeon processors and Gaudi AI accelerators in the system provides a powerful solution to make AI faster, cheaper, and more accessible," Intel noted in a statement. ”
On the PC side, Intel unveiled the infrastructure of Lunar Lake and said that the architecture could lead to the next generation of AI PCs, which consume 40% less power than previous generations of chips.
The Lunar Lake chip is a high-performance product carefully built by Intel for the AI PC market. For the first time, the chip was all manufactured by TSMC and was manufactured using TSMC's advanced N3B and N6 process nodes, resulting in significant advantages in process technology. Through TSMC's advanced process technology, Lunar Lake has achieved comprehensive improvements in key performance such as CPU, GPU, and NPU, bringing users a smoother and more efficient AI experience.
In terms of performance, the Lunar Lake chip performs well. It is designed with the Hybrid core architecture, with 4 Lion Cove P-core performance cores and 4 Skymont E-core efficiency cores to form a powerful 8-core hybrid design. This makes it easy for Lunar Lake to multitask, delivering excellent performance for work, entertainment, and gaming. At the same time, Lunar Lake's GPU performance has also been greatly improved, comparable to the Xe2 core graphics architecture, bringing users more realistic graphics rendering and smooth gaming experience.
Figure: Intel's next-generation new mobile processor (Source: Intel)
What's more worth mentioning is that the Lunar Lake chip has made a huge leap forward in AI performance. It is equipped with four NPUs (Neural Network Processors), which provides up to 3 times the AI performance of the previous generation. This enables Lunar Lake to complete AI tasks such as image recognition, speech recognition, and natural language processing more quickly and accurately, bringing users a more intelligent user experience.
In addition to the performance improvement, the Lunar Lake chip has also made innovations in memory integration. It uses the advanced Foveros packaging process, which integrates 16/32GB of LPDDR5X RAM directly into the package. This design not only reduces the load on data transmission and reduces energy consumption by about 40%, but also makes the operation of the whole system more stable and efficient. Users can enjoy a smooth experience without additional memory, and the SoC (system-on-chip) architecture similar to smartphones also makes Lunar Lake more competitive on mobile devices.
When it comes to packaging and connectivity, Lunar Lake also excels. It uses the Intel Foveros packaging process to combine the computing chip and the platform controller chip to achieve higher integration and lower energy consumption. At the same time, Lunar Lake also supports high-speed PCIe 5.0 and USB 4.0 interfaces, providing users with faster and more convenient data transfer and expansion capabilities.
Intel said that the Lunar Lake chip is expected to be officially unveiled in the fall of 2024. The release of this chip will further promote the development of the AI PC market and bring users a more intelligent and efficient experience. At the same time, the cooperation between Intel and TSMC will also bring new opportunities and challenges to the semiconductor industry.
