跳至主要内容

Chinese GPU chips are catching up with GTX 1080

In addition to CPUs, domestic companies are also beginning to catch up with international standards in the GPU field. A few days ago, Changsha Jingjiawei revealed that the new generation of GPU chips developed by the company is currently in the back-end design, and subsequent progress will be disclosed in regular reports.


Although Jing Jiawei did not disclose what the so-called next-generation GPU is, according to the company's previous information, the next-generation GPU chip should be the JM9 series, which has been developed since 2018.

On the GPU, Jingjiawei currently has two series, JM5 and JM7. Among them, the JM5400 series has been used in Chinese military aircraft to replace /ATI products, and the JM7200 series uses a 28nm process and has similar performance to NVIDIA’s GT640 graphics card. However, the overall power consumption is less than 10W, which is much lower than the latter's 50WTDP, and some orders have been obtained. The next-generation GPU is the JM9 series. In this prospect, Jiawei mentioned in the report that the JM9 series GPU will be replaced with a unified rendering architecture and increase the number of programmable computing modules, in line with the main trends of current graphics cards.

According to the official specifications, the performance of JM9231 can reach the level of low-end products in 2016, and the core frequency of JM9271 is not lower than 1.8GHz, supports PCIe4.0x16, with 16GB HBM memory, bandwidth 512GB/s, floating point performance up to 8TFLOPS, The performance is not lower than the level of GTX1080 and can reach the level of high-end graphics cards at the end of 2017.

Jingjiawei was established in April 2006 with a registered capital of 301 million yuan. It is mainly engaged in the research and development, production and sales of high-reliability electronic products. The products mainly involve the field of display and control, small-scale specialized radar, chip field and others.

评论

此博客中的热门博文

NVIDIA GeForce RTX 2080 Ti

The GeForce RTX 2080 Ti is an enthusiast-class graphics card by NVIDIA, launched in September 2018. Built on the 12 nm process, and based on the TU102 graphics processor, in its TU102-300A-K1-A1 variant, the card supports DirectX 12 Ultimate. This ensures that all modern games will run on GeForce RTX 2080 Ti. Additionally, the DirectX 12 Ultimate capability guarantees support for hardware-raytracing, variable-rate shading and more, in upcoming video games. The TU102 graphics processor is a large chip with a die area of 754 mm² and 18,600 million transistors. Unlike the fully unlocked TITAN RTX, which uses the same GPU but has all 4608 shaders enabled, NVIDIA has disabled some shading units on the GeForce RTX 2080 Ti to reach the product's target shader count. It features 4352 shading units, 272 texture mapping units, and 88 ROPs. Also included are 544 tensor cores which help improve the speed of machine learning applications. The card also has 68 raytracing acceleration cores. NVID...

AMD Ruilong 5000 APU "Cezanne" exposure: GPU still integrates Vega

Although Ruilong 4000 series desktop APU processors have not been officially released, news about the next generation Ruilong 5000 series has begun to gradually reveal. According to the device ID information captured by Komachi, the AMD Ruilong 5000 APU is code-named "Cezanne (Cezanne)" and the PCI ID is 1638. Previously, the PCI ID of the Ruilong 4000 Renoir APU was 1636. Interestingly, according to Komachi, the partner with Cezanne Zen3 CPU is still GFX9, which is the Vega (Vega) GPU. Of course, AMD will further tap Vega's potential and debug performance. Finally, there is good news. The Ruilong 5000 "Cezanne" APU will retain the AM4 interface. With the realization of Cezanne, we can also sort out Zen3 family products. EPYC server product code is Milan (Milan), desktop HEDT fever platform thread tearer code is Genesis Peak, desktop mainstream platform (Ruilong 4000) code Vermeer (Vermil). In addition, the Zen4 architecture EPYC codename Genoa (Genoa), the...

NVIDIA A100 4x GPU HGX Redstone Platform

While the 8x NVIDIA A100 GPU “Delta” platform with NVSwitch got a lot of airtime during the Ampere launch, it was not the only platform being launched today by NVIDIA. The 4x GPU “Redstone” platform is a smaller NVLink mesh platform that is designed to be a lower-cost option. The NVIDIA A100 “Redstone” HGX platform is important since it is a smaller and less complex version of the HGX A100 platform. The Redstone platform incorporates 4x SXM NVIDIA A100 GPUs onto a PCB. As we saw with the Tesla A100 overview, the new GPUs have 12x NVlinks per GPU. Each NVLink provides 50GB/s of GPU-to-GPU bandwidth for 600GB/s total. Redstone takes those 12 NVLinks and splits them into three groups. Instead of the NVIDIA NVSwitch solution we see on the HGX A100 platform, we get a mesh topology without switching. NVIDIA has offered both switched and non-switched systems for some time. This type of topology, NVIDIA has been using for years and is the basis for many important compute nodes. For ...