A country against AMD, NVIDIA and Intel: this is China’s gaming GPU

China scares everyone, regardless of who looks at it and with the eyes that look at it. It has become the owner of rare minerals, electronics and now develops all kinds of its own components under its companies to try to compete and in the near future dominate the world with its business vision. You will like it more or less, but the US does not have the leadership right now and only certain companies stick their heads out of the country, but maybe for a short time, since if AMD, NVIDIA and Intel promised them very happy in their oligopoly, Be careful with what China has just presented.

Jingjia Micro JM9271 and JM9231, the surprise as a GPU for gaming

From the military to the civilian to end in gaming, it makes sense if we see the figures that the video game and graphics card sector is beginning to move based on the income of NVIDIA and AMD. China is clear about it and the sample is the new JM9271 GPU, which although it does not compete with the last of the three main manufacturers, begins to scare by the great advance in such a short time.

Two models have actually been presented, the JM9231 and the JM9271, the first as an entry range and the second and discussed is the top model. Interestingly, the company is not focused on implementing support for DX12, so they remain in OpenGL 4.5 and OpenCL 1.2 for the JM9231, while for the JM9271 we talk about OpenGL 4.5 and OpenCL 2.0.

The data and hardware specifications are very scary, where the lower-end model reaches up to the 1500 MHz with support for PCIe 3.0 x16 with a pixel rate of 32 GPixel / s, a performance of 2 TFLOP, support for H.265 at 4K and 60 FPS and surprisingly, with a memory 8GB GDDR5 for a bandwidth of 256 GB / s.

Superior performance than a GTX 1080

JM9231 GTX 1050 JM9271 GTX 1080
API Support OpenGL 4.5, OpenCL 1.2 OpenGL 4.6, DX12 OpenGL 4.5, OpenCL 2.0 OpenGL 4.6, DX12
Boost Clock Rate > 1,500 MHz 1,455 MHz > 1,800 MHz 1,733 MHz
Bus Width PCIe 3.0 PCIe 3.0 PCIe 4.0 PCIe 3.0
Memory Bandwidth 256 GB / s 112 GB / s 512 GB / s 320 GB / s
Memory Capacity / Type 8GB GDDR5 2GB GDDR5 16GB HBM 8GB GDDR5X
Pixel Rate > 32 GPixel / s 46.56 GPixel / s > 128 GPixel / s 110.9 GPixel / s
FP32 Performance 2 TFLOPs 1.8 TFLOPs 8 TFLOPs 8.9 TFLOPs
Output options HDMI 2.0, DisplayPort 1.3 HDMI 2.0, DisplayPort 1.4 HDMI 2.0, DisplayPort 1.3 HDMI 2.0, DisplayPort 1.4
Video Encoding H.265 / 4K 60FPS H.265 / 4K 60FPS H.265 / 4K 60FPS H.265 / 4K 60FPS
TDP 150W 75W 200W 180W

Your sister gets a frequency of 1800 MHz with support for PCIe 4.0 and a FP32 performance over 8 TFLOPS with a pixel rate of no less than 128 GPpixel / s. The most impressive thing is that you get 512 GB / s of bandwidth thanks to the implementation of 16GB HBM, which is double that of the RX Vega 64.

The consumptions are not as good as those that NVIDIA had for example at the time with the GTX 1080 and GTX 1050, since they obtain respectively 150 watts and 200 watts, but they do not have a node as advanced as the one that AMD and Huang had. The number of shaders has not been revealed, mainly because it is practically certain that these GPUs are manufactured at 28 nm and not at 14 or 16 nm like the mentioned rivals, so it has even more merit.

The first time we talked about them was two years ago when they were still on paper. Jingjia Micro has complied with everything that was said at the time, so although they are 5 years behind in technology, this is a very narrow gap for the sector where we are, especially seeing that they started from nowhere and now have a actual product on the table.

The use of HBM in a process and custom PCB, since they have not collected Intel or TSMC technology, which indicates that the stacking of stacks is also real and since it is not easy at all, it shows the huge amount of resources, time and money that they are dedicating thereto. Will we see in 10 years a real competitor for Intel, NVIDIA and AMD? At the rate they are going is not ruled out.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *