Artificial intelligence company Cerebras has once again unveiled the world’s largest processor this year. The company has now raised the bar to the 5nm level. The WSE-3 (Wafer Scale Engine-3) processor, which has set a record in the number of cores, comes with 900,000 artificial intelligence cores and 4 trillion transistors. Cerebras receives the processor from TSMC at the wafer level. Thus, giant-sized processors emerge.
The world’s largest processor WSE-3 and its features
Cerebras is a venture that produces processors used in artificial intelligence studies. Previously, the company had taken the title of the world’s largest processor with the 16nm WSE-1 and 7nm WSE-2 processors. Now, the company, which has developed the same title, is introducing a much more powerful processor.
The processor to be used for artificial intelligence training is in a huge size of 46,225 mm2. Being at the wafer level, the processor provides much higher performance and power efficiency. Since the venture produces these chips for supercomputers and artificial intelligence servers, it has a more flexible area in terms of processor sizes. In this sense, the venture takes advantage of the ability to produce processors at the wafer level.
The WSE-3, which has the same power consumption as its predecessor WSE-2, is presented with double the performance. The processor has 900,000 artificial intelligence cores, along with 44 GB of chip-on-top SRAM and 125 FP16 PetaFLOPS peak performance.
Intel Gaudi2 vs NVIDIA H100 AI chips face off in a performance test. Which model was faster?
The world’s largest processor WSE-3 also powers Cerebras’ supercomputer CS-3. This supercomputer can be used to train artificial intelligence models with up to 24 trillion parameters. According to Cerebras, the CS-3 supercomputer has the capacity to train artificial intelligence models 10 times larger than Gemini and ChatGPT.
This supercomputer can support external memory sizes of 1.5 TB, 12 TB, or 1.2 PB. This allows it to store massive artificial intelligence models on its own without partitioning. In this sense, the CS-3 supercomputer has a significant performance advantage for training artificial intelligence.
Additionally, Cerebras is aiming to break the record for the world’s largest artificial intelligence supercomputer with the G42. Previously, the Condor Galaxy-1 and Condor Galaxy-2 supercomputers held this record. The newly developed Condor Galaxy-3 supercomputer will bring together 64 CS-3 systems and have a structure with 57.6 million cores. Currently, the Condor Galaxy-1 and Galaxy-2 supercomputers have a total power of 8 exaFLOPS. The Condor Galaxy-3 supercomputer, on the other hand, will have a power of 16 exaFLOPS on its own.
