Tachyum’s First CPUs Are Late, But Still Planning 20 ExaFLOPS Supercomputers with Next-Gen CPUs
Tachyum has launched its Prodigy common processor for 3 years with guarantees that will probably be higher than CPUs and GPUs from AMD, Intel and Nvidia. Nevertheless, the chip remains to be not in manufacturing regardless of being manufactured. initially expected to be in full volume production in 2021. Now the corporate appears to be wanting forward and Prodigy highlights 2-based supercomputer design It guarantees 20 ExaFLOPS efficiency in 2025~2026.
Tachyum supercomputer design The corporate mentioned it’ll ship 20 FP64 vector ExaFLOPS and 10 AI (INT8 or FP8) ZetaFLOPS efficiency inside an influence goal of 60 MW in an space of 6,000 sq. ft by 2025. The machine will use 64 Prodigy 2-based cupboards and 16 storage racks, however Tachyum made what number of Prodigy 2 processors it could must ship that a lot efficiency.
Tachyum additionally mentioned it may construct a supercomputer that provides 24.9 FP64 vector ExaFLOPS and 13.27 AI ZettaFLOPS consuming 73.8MW. To place the quantity in context, AMD’s new El Capitan supercomputer, powered by Intuition MI300 information heart APUs, is ready to ship roughly 2 FP64 ExaFLOPS.
Final 12 months, the US Division of Vitality expressed its curiosity in buying 20 ExaFLOPS supercomputers with an influence consumption of 20MW-60MW by 2025. Tachyum has submitted a proposal to construct such a system by 2025, however didn’t go into particulars. For the reason that firm’s authentic Prodigy did not meet Tachyum’s efficiency targets (which is why it sued Cadence), it was cheap to imagine that the corporate would use its 2nd Era Prodigy for the machine (which the corporate talked about earlier).
Unveiled by Tachyum this week, this supercomputer design basically complies with DoE’s necessities, though the corporate hasn’t introduced the efficiency ranges it expects from the Prodigy 2.
Tachyum says its Prodigy processor is the primary common processor able to dealing with a variety of demanding computing workloads. The unique Prodigy processor accommodates 128 proprietary 64-bit VLIW cores with two 1024-bit vector items and one 4096-bit matrix unit per core. The flagship Prodigy T16128-AIX processor was anticipated to supply a most of 90 FP64 teraflops for top efficiency computing (HPC) and as much as 12 ‘AI petaflops’ for AI inference and coaching (with INT8 or FP8 precision). Moreover, every chip is envisioned to have an influence consumption of as much as 950W and use liquid cooling.
Tachyum’s Prodigy 2 is predicted to extend the core depend and add HBM3, PCIe 6.0 + CXL and built-in networking capabilities.
“Tachyum provides pioneering options from silicon to finish methods to fulfill the ever-increasing calls for for each HPC and synthetic intelligence,” mentioned Radoslav Danilak, founder and CEO of Tachyum. “Tachyum-designed supercomputers are pushing the forefront of HPC efficiency as they break the zetta-scale barrier for AI and rework information facilities into common computing facilities.”
#Tachyums #CPUs #Late #Planning #ExaFLOPS #Supercomputers #NextGen #CPUs