In brief: Tianshu Zhixin Semiconductor, a fabless joint venture between Via Solutions and the Shanghai municipal investigation, has revealed what may look like China’s first 7nm GPGPU, a gargantuan chip surely supposed to compete with the likes of N-vidia A100 and AMD MI100 in the data center.
China is on a quest for “technological self-sufficiency , ” and the exchange hand techinques war with the US would have only accelerated that procedure. China is capable of covering 20+ percent of the chips was looking for in the local industry, but the lawmakers plans to increase that number to 70 percent by 2025. A sample of that is using aging chips while its semiconductor current market catches up, turning to Japan for etching models that would otherwise be considered outdated.
Last year, we saw any arrival of an impressive x86 PC courtesy of government-backed Zhaoxin, which showed simply how much progress has been made on which usually front. We’ve also heard of Huawei working hard to develop server GPUs , but things are already relatively quiet ever since those tech giant lost unrestricted entry to key suppliers such as TSMC.
Tianshu Zhixin was known to be production progress on the first general-purpose GPU designed in China, after that “Big Island. ” The continued chip shortage only accepted the company to do a paper launch in March, sadly what’s interesting about these dings is that they’re supposedly free to compete with Nvidia and AMD’s GPU at the highest part.
Big Island GPUs are usually technically designed to go against RTX and Radeon gaming gpu cards, but instead they’re aimed at machine learning, high performance research, medical research, and safety and security. This means they’re meant to not in favor of the likes of Nvidia’s A100 and AMD’s Instinct MI100, both of which are gargantuan french fries designed for the data center and display an order of variation the performance of prior generation architectures, while capturing lot less space not to mention power to operate.
Immense Island was developed between 2018 and 2020, built by making use of TSMC’s 7nm process node and 2 . 5D CoWoS packaging. They also utilize TSMC’s 65 nm silicon interposer, and feature no less than 24 billion transistors. Just like AMD and as a consequence Nvidia, Tianshu Zhixin maintains equipped its GPGPU because of 32 GB of HBM2 physical memory (1. 2 TB for each and every second bandwidth) and made the software compatible with the PCIe look at. 0 standard.
The claimed performance of the Chinese GPGPU seems to be impressive. When it comes to FP16 (half precision math) execution, it is capable of 147 teraflops, which sits right among Nvidia A100’s 78 teraflops and AMD MI100’s 184. 6 teraflops — as it should be noted the A100’s Tensor cores are capable of 312 FP16 teraflops.
When it comes to FP32 workloads, Big Island is required to be capable of up to 37 teraflops, which is higher than both the A100 and MI100, which offer 21. 5 teraflops and 3. 1 teraflops, respectively. Impact consumption is rated using 300 watts, and Tianshu Zhixin says the new GPU will offer a better price-to-performance gesunder verstand than competing solutions.
Experience will tell if this is without a doubt true and China has got hit such an important motorola milestone in its quest to reduce its reliance on foreign semiconductors.