The processor adopting risc-v architecture inadvertently appears in daily electronic products, which seems to have become more and more common: not only some typical MCU manufacturers begin to embrace risc-v. It is also reflected in some mature products. For example, at this year's China IC leaders summit, we had a dialogue with Silicon Valley digital analog, and learned that today's very mature TCON chip has joined the risc-v small core without showing mountains and dew; Some years ago, NVIDIA began to select risc-v as the controller in its GPU products; Western Digital also began to choose risc-v core in the whole product line. This makes us more and more interested in the processor microarchitecture of risc-v instruction set: last October, sifive released the first risc-v disordered CPU core, U8 series processor IP. Sifive mentioned in the publicity that U8 series is the highest performance risc-v instruction set core IP (it seems to be later than xuantie 910), "based on superscalar scrambled pipeline, and configurable pipeline depth and transmission sequence width".

       Sifive is still quite active in the micro architecture IP authorization of risc-v. as the high-performance core of the risc-v world, sifive's 7-series and 8-series IP cores are an important window for us to observe the current development of risc-v. although the micro architecture analysis can not show the characteristics of the instruction set itself, it is also an opportunity to understand the risc-v ecology through their comparison with arm. How does the highest performance risc-v core compare to arm? First of all, it is necessary for us to understand the positioning of different series of products within sifive. In an online conference called "embedded intelligence everywhere with sifive 7 series core IP" held by sifive in April this year, sifive once summarized its product line. Sifive's E-Series core focuses on 32bit embedded use scenarios; The 64bit s-core is biased toward the scenario with greater demand for computing power; The core of U-series is the positioning with the strongest performance and is oriented to high-end computing. If taking numbers as the segmentation dimension, sifive's 2 series is the most efficient and smallest processor IP among risc-v processor IP; The 3-Series and 5-series are more widely deployed, and are used in some multi-core configurations and scenarios requiring high real-time processing capabilities; Both the 7 series and the 8 series focus on high performance as described above.

       The combination of numbers and letters constitutes different products. For example, E3 provides 32bit performance for embedded applications oriented to midrange positioning; S7 is obviously a product focusing on performance; U8 realizes a high-performance core that can be configured in a large scale and achieves the highest performance per watt. Before the release of U8, sifive's high computing power products were mainly U5 and U7 - these two are still sequential architectures, and the target is the low-end and microcontroller core of arm. In the face of more complex loads and scenarios requiring computing power, they are still deficient. U8 series is a product to make up for this shortcoming. Sifive announced at that time that U8 would greatly expand the ecosystem viability of sifive and risc-v in terminal products.