Super Performance Intel Xeon 128-Core CPU Comes Out

In particular, CEO Radoslav Danilak, who has 25 years of experience in the semiconductor industry, was founded by SandForce, a once-popular SSD control giant. He also personally served as CEO, was later acquired by Seagate, and later founded Skyera, and continued to study SSD control. Technology and then acquired by Western Digital in 2014. Tachyum Prodigy claims to be the world's first "universal processor" (universal processor) because it integrates for high-performance computing, AI artificial intelligence, DML deep mechanical learning, and interpretable in a single silicon chip based on a parallel multi-processor environment and can simplify the programming model and environment.

The latest top model is Prodigy T6128, a single-channel single chip integrates 128 physical cores, out-of-order execution architecture, 4 instructions per clock cycle , supports 64-bit addressing, 512-bit vector operations, AI / ML vector and matrix multiplication acceleration, Virtualization, advanced RAS, operating frequency up to 4GHz.

In terms of cache, each core 32 KB first-level instruction cache and 32KB first-level data cache support ECC, and the last-level cache shares 64 MB and supports DECTED ECC. In terms of memory, it supports DDR4 and DDR5 with 12 channels, and the highest frequency is DDR5-4800, but each channel can only have one memory stick. The maximum capacity of a single stick is 512 GB and the total maximum is 6TB. It also supports advanced error correction and RAS. In terms of IO, it integrates up to 36 PCIe 5.0 controllers, up to 48 channels , and also integrates two sets of 400 G (400,000 Gigabit) Ethernet controllers.

What is even more amazing is that all these powerful specifications, with the support of TSMC's 7nm processor, the packaging area is only 85 × 85 square millimeters, which is slightly larger than the Intel LGA2066 Core, but smaller than the AMD SP3 thread reader . Tachyum did not disclose the specific architecture of Prodigy. I do n’t know if it is based on RISC-V, MIPS, ARM or self-research. It just said that whether it is single-threaded or multi-threaded applications, it has surpassed Intel Xeon, but it is more compact than ARM. According to reports, the Prodigy T6128 processor is suitable for large-scale supercomputers, big data, and large-scale AI applications. It can provide 262T Flops AI training inference performance and 16TF lops HPC high computing performance.

In terms of development environment, Tachyum also provides a series of tools, including FPGA simulators, software simulators, binary translators, C / C ++ / Fortran compilers, debuggers and configuration files, and TensorFlow compilers, all under the Linux operating system. If the customer does not need the 128-core specifications, Tachyum also provides different configurations such as 64/32/24/16 core.

There are two models with 64 cores, one is T864 , which supports eight channels of DDR4 / DDR5 memory, 72 PCIe 5.0 channels, two sets of 400G Ethernet, two sets of HBM3 (optional), 32 MB full consistency L2 / L3 cache. The operating frequency is 4GHz, the core voltage is 0.8 V, the thermal design power consumption is 180 W, the core area is 290 square millimeters, and the package area is 66 × 66 square millimeters. It can replace single / dual Xeon E7 and Xeon E5. The other is TH24, which is designed for AI / HPC, quad-channel DDR5 and/or 32GB HBM3. The latter can be used as a cache or independent memory, which requires high-precision water cooling. The 32-core model is T432, four-channel DDR4, 32 PCIe 4.0, and two sets of 100G Ethernet. The 16-core is T216, dual-channel DDR4, 32 PCIe 4.0, two sets of 50 G Ethernet. Both are small-sized packages, with low cost and low price, suitable for replacing Xeon E5, Xeon E3, Xeon D series. Of course, most of these products are still on paper. At present, only the 64-core T864 has been successfully taped, and it is expected to enter mass production this year.

access: