Realtime News

Original Chinese language 7nm GPU opponents Nvidia and AMD for efficiency


Shanghai Tianshu Zhixin Semiconductor

(Image credit ranking: Shanghai Tianshu Zhixin Semiconductor)

Shanghai Tianshu Zhixin Semiconductor Co’s contemporary 7nm GPU is up and working. The contemporary chip is a beast, packing no fewer than 24 billion transistors. For context, Nvidia’s GA102 GPU, as chanced on in the RTX 3080 and 3090 graphics boards, contains 28 billion transistors.

Identified as Astronomical Island (BI), the contemporary GPU can crank out as a lot as 147 TFLOPS of FP16 efficiency. As Tom’s Hardware points out, that’s with out a doubt aggressive with the same products from Nvidia and AMD. Nvidia’s A100 and the AMD Instinct MI100 are edifying of 77.97 TFLOPS and 184.6 TFLOPS of FP16 efficiency, respectively, though the Nvidia A100 additionally has Tensor cores with a extra 312 TFLOPS of FP16.

As for production skills, Shanghai Tianshu Zhixin Semiconductor says BI is fabbed on 7nm. The crucial points supplied appear to align with TSMC’s 7nm node, but the foundry partner is no longer identified and it’s unclear if China has home chip production skills in that class. Images of the contemporary GPU indicate a multi-chip equipment with the GPU itself housed with onboard memory or cache.

For certain, the BI chip along with the AMD and Nvidia competition mentioned are all GPUs namely made for endeavor capabilities, no longer desktop PCs or gaming. Any GPU for gaming will want a huge vary of hardware beyond the computational cores that crank out these TFLOPS, including a paunchy pipeline of texture objects, pixel processing and render outputs. 

It’s no longer determined if the BI GPU contains any of these functions and, indeed, there’s miniature rationalization for his or her inclusion given the remit. So, the true assumption is that Shanghai Tianshu Zhixin Semiconductor’s contemporary GPU poses zero risk to the established competition in gaming graphics. 

That said, the easy indisputable fact that a spacious, ultra-complex GPU with connected efficiency thru floating point efficiency to the established players has been produced is necessary. It’s a tall step from there to a gaming GPU. However it’s with out a doubt a initiate.