In short: China’s very first homegrown GPUs are remarkable on paper, however the business that makes them does not appear too crazy about offering them to customers. Rather, it’s pitching them as power-efficient calculate accelerators for cloud systems, despite the fact that the spec sheet states they support DirectX and Vulkan.
In November, we discovered that a Chinese business called Innosilicon had actually been establishing a GPU that might be a comparable success story as that of YMTC and its 128- layer 3D NAND memory. The brand-new graphics card– called Fenghua 1– didn’t get a fancy statement like Nvidia and AMD are understood for, however what captured our eye was that it was developed to support the DirectX and Vulkan APIs utilized in lots of modern-day PC video games.
This week, TechNode’s Stewart Randall shared a brief video on Twitter of what seems evidence that China’s very first homegrown GPU isn’t simply a mirage. The video reveals a mini-tower PC geared up with Innosilicon’s Fenghua 1 graphics card running the GFXBench T-Rex standard. There’s no frames-per-second counter on the screen, and we do not get to see the test outcome, however because this is more of a mobile-oriented criteria, it would be unreasonable to draw contrasts to existing graphics cards from AMD and Nvidia.
At least on paper, the Fenghua 1 needs to match low-mid variety graphics cards. There are 2 variations, a Type-An efficient in providing up to 5 teraflops of FP32 efficiency and a Type-B that can double that thanks to a dual-chip architecture. They’re both based upon PowerVR IP from Imagination Technologies, which must make them fairly power effective in spite of being produced utilizing a 12 nm procedure node. They likewise support DirectX 11, DirectX 12, and Vulkan 1.2.
For referral, Nvidia’s GTX 1660 SUPER can provide a bit over 5 teraflops of FP32 efficiency, and the non-Ti RTX 2080 offers simply a hair over 10 TFLOPS. The specs offered by Innosilicon do not appear to include up. On the one hand, the Fenghua 1 cards utilize 19 Gbps GDDR6 and GDDR6X memory for approximately 304 GB per second of bandwidth, and the business states the pixel fill rate of the Type-A setup is a decent 160 GPixels per second. On the other hand, the common power intake ranking is a small 50 watts, and there’s no other way to inform how these specifications equate into video gaming circumstances.
Innosilicon states it will offer various versions with 4, 8, and 16 gigabytes of VRAM, and the marketing products suggest these are mostly targeted at the business market. Very little is understood about retail accessibility, however we might find out more about these mystical graphics cards next year.