‘Run AI models for as low as $5,000’: plucky CPU startup that claimed 99% saving on AI costs now wants to sell you an AI workstation — with an unbelievable price tag and 1TB RAM

CPU startup Tachyum has previously said that one of its Prodigy Universal Processor units can rival dozens of Nvidia H200 GPUs. The 192-core 5nm processor delivers 4.5 times the performance of the best processors for cloud workloads, and claims to be six times more effective than GPUs for AI.

Now the firm has announced the Prodigy ATX Platform, an AI workstation that promises to run advanced AI models for just $5,000. The ambitious system, which features an impressive 1TB of memory, is designed to make sophisticated AI models accessible to a wider audience.

The Prodigy ATX Platform is built around a 96-core Prodigy processor, which is designed with only half of its die enabled, a strategy aimed at reducing power consumption and enhancing yield, thereby lowering costs and making the platform more accessible.

Effective for LLMs

The system is expected to come equipped with 1TB of DDR5-6400 SDRAM using 16 memory modules, offering a peak bandwidth of 819.2 GB/s. The system’s design includes three PCIe x16 5.0 slots, three M.2-2280 NVMe slots with a PCIe 5.0 x4 interface, and SATA connectors for SSDs and HDDs.

Tachyum says its Prodigy ATX Platform is particularly effective for running LLMs which are notorious for their high memory capacity requirements. 

A single Prodigy system can reportedly run a ChatGPT4 model with 1.7 trillion parameters, which would require 52 NVIDIA H100 GPUs to run at a significantly higher cost and power consumption. 

Despite the impressive specifications, there are doubts about the economic viability of the Prodigy ATX Platform for Tachyum. As Tom’s Hardware points out, the total cost of the system components, excluding the Prodigy processor, is estimated to be around $4800. However, Tachyum’s CEO, Dr. Radoslav Danilak, remains optimistic, stating that the platform’s powerful AI capabilities will enable organizations of all sizes to compete in AI initiatives.

“Generative AI will be widely used far faster than anyone originally anticipated,” Dr. Danilak said. “In a year or two, AI will be a required component on websites, chatbots and other critical productivity components to ensure a good user experience. Prodigy’s powerful AI capabilities enable LLMs to run much easier and more cost-effectively than existing CPU + GPGPU-based systems, empowering organizations of all sizes to compete in AI initiatives that otherwise would be dominated by the largest players in their industry.” 

The Prodigy ATX Platform’s launch has been delayed multiple times, with the latest plan pegging the processor’s launch for the second half of 2024. It remains to be seen whether Tachyum can deliver on its promises and revolutionize the AI landscape.

More from TechRadar Pro

https://www.techradar.com/rss

waynewilliams@onmail.com (Wayne Williams)

February 25, 2024 11:11 am