Last May, Intel released the tenth-generation Core processor, codenamed Ice Lake. Intel says each of the next-generation Core processors comes with a new Sunny Cove microarchitecture with Intel AVX-512-Deep Learning Boost that accelerates automatic image enhancement, image retrieval, media postprocessing, and other AI tasks. They are also configured with the Intel GNA Private Engine and DynamicS Tuning 2.0.
On the eve of CES 2020, Intel introduced some vague information about Ice Lake’s successor, Tiger Lake, and AI capabilities will be significantly improved.
Intel briefed the media on some new news on the eve of CES 2020, including the upcoming Comet Lake-H processor for laptops, which will have a main frequency of more than 5.0 GHz and has up to eight cores and 16 feature threads, as well as Tiger Tiger Some vague details about the performance of Lake processor AI, as well as the first official news of the NUC 9 Extreme “Ghost Canyon NUC”. Intel will also announce the expansion of Athena in tomorrow’s keynote address
Intel provided more information about its AI program, including a vague statement that the AI performance of its Tiger Lake processor will improve due to improvements in Intel’s three-engine AI subsystem. Tiger Lake processors will have faster DL Boost and inference engines that run in the execution kernel and the Xe graphics engine, respectively.
Intel says the AI engineembedded in the chipset will receive updates to the new second-generation solution. However, no performance indicators or hard data were given to support this claim. Intel will announce more details in tomorrow’s keynote address.
Intel’s transition to a 10nm process has been difficult as it strives to create high-performance nodes with high yields to be produced cost-effectively. Because of these challenges, the Ice Lake chip has a lower main frequency than the previous generation, and Intel has made up for the reduction in clock speed with architectural improvements, increasing the IPC by an average of 18 percent.
The new 10nm Tiger Lake chip should have an updated architecture and a higher clock rate based on the 10nm process (10nm? According to Intel’s official account, the Tiger Lake processor will have up to 25W of power. Based on recently obtained information, Tiger Lake-U’s L3 cache capacity will increase by 50%, from 8MB to 12MB, 3MB per core. The chips also support the AVX-512 and are expected to appear alongside the Willow Cove kernel and Gen12 Xe GPU.
Intel once again explained its progress on AI and described the growing software ecosystem that supports AI acceleration, such as Adobe Photoshop and Lightroom, CyberLink, Topaz Labs, and Cyber Labs. Not to mention Intel’s own software solutions. The extensive support for AI acceleration software marks a huge step forward compared to the only applications that supported AI acceleration software when Ice Lake was originally released.
Intel says the Tiger Lakes Gen12 Xe image processor has hardware for inT8 acceleration, which is the preferred format for low-precision AI. The graphics engine is primarily used for continuous and recurring workloads, which will free up more performance from the graphics engine. Unspecified enhancements to the DL Boost engine will provide AI performance improvements for burst loads on the CPU core. Of course, chipsets will still solve low-power audio processing tasks, and the improved AI accelerator engine will deliver higher performance.
As we expected, Intel insists that AMD’s chips do not have hardware AI acceleration. AMD’s lack of optimization is not surprising, as AMD claims to be willing to be a quick follower of certain types of instructions, such as avX-512, allowing all types of new use cases and associated software ecosystems before hardware is supported. As we’ve seen, Intel is still leading aDI on desktop and laptop processor platforms, but the software ecosystem will take some time to mature.