Tachyum has created Prodigy T16128, a processor designed to carry out any type of workload, from general-purpose computing to high-performance computing (HPC) and AI. The chip, consisting of 128 64-bit cores with a maximum frequency of 5.7GHz, 16 DDR5 memory controllers (with a maximum capacity of 8TB per socket) and 64 lanes PCIe 5.0, is defined by the company as the first "processor universal "in the world, capable of running multiple computing applications on a single device that needs 1/10 the power compared to traditional hardware at 1/3 the cost.
Photo Credit: Tachyum Tachyum Prodigy T16128 supports multiple data types, including FP64, FP32, TF32, BF16, Int8, FP8 and TAI, and offers four times the performance of the fastest Intel Xeon on the market and three times that of the NVIDIA H100 in the HPC field, all with extraordinary efficiency. This incredible result was possible thanks to the design choice of having built Prodigy with matrix and vector processing capabilities from the beginning. Each core is capable of 2 1024-bit vector units, 4096-bit matrix operations, and 4 out-of-order instructions per clock. The chip also includes 128MB of L2 + L3 cache with error correction capabilities.| ); }
Photo Credit: Tachyum This could be a real breakthrough, especially in the field of data center, as companies may no longer need multiple components to perform the same tasks, increasing efficiency and reducing consumption and space. Tachyum Prodigy T16128 is the top model of the line, with the T864 (64 cores) and T832 (32 cores) occupying the medium and entry-level leaves. Production will start in 2023, so we'll have to wait until next year before we see full third-party testing.
Photo Credit: Tachyum Tachyum Prodigy T16128 supports multiple data types, including FP64, FP32, TF32, BF16, Int8, FP8 and TAI, and offers four times the performance of the fastest Intel Xeon on the market and three times that of the NVIDIA H100 in the HPC field, all with extraordinary efficiency. This incredible result was possible thanks to the design choice of having built Prodigy with matrix and vector processing capabilities from the beginning. Each core is capable of 2 1024-bit vector units, 4096-bit matrix operations, and 4 out-of-order instructions per clock. The chip also includes 128MB of L2 + L3 cache with error correction capabilities.| ); }
Photo Credit: Tachyum This could be a real breakthrough, especially in the field of data center, as companies may no longer need multiple components to perform the same tasks, increasing efficiency and reducing consumption and space. Tachyum Prodigy T16128 is the top model of the line, with the T864 (64 cores) and T832 (32 cores) occupying the medium and entry-level leaves. Production will start in 2023, so we'll have to wait until next year before we see full third-party testing.