Tachyum has announced its new and open-source TDIMM memory standard, offering big uplifts in bandwidth & high capacity per module.
Tachyum's Open-Source TDIMM Memory Standard Boasts Big Bandwidth & Capacity Uplifts For AI Data Centers
Following its 2nm Prodigy and Prodigy Ultimate announcement, Tachyum has now laid out the plans for a brand new memory standard called TDIMM, which essentially refers to Tachyum DIMM. This new memory design is fully open-source and offers big uplifts in bandwidth and capacity.
According to Tachyum, the DDR5 TDIMM standard increases the bandwidth by 5.5x, upping the numbers from 51 GB/s on standard RDIMMs to 281 GB/s. The memory standard is designed around various form factors, each offering a diverse range of capacities. The standard designs feature 256 GB capacities, taller DIMMs will get 512 GB capacities, and then there's an Extra Tall height which maxes out the capacity to 1 TB.
Comparing DDR5 RDIMMs to DDR5 TDIMMs:
- RDIMM: 64-bit data, 16-bit ECC, 40 DQS, 14 C/A, 16 control signals, 288-pin connector
- TDIMM: 128-bit data, 16-bit ECC, 36 DQS, 14 C/A, 16 control signals, 484-pin connector
The first thing to make from this is that TDIMM will require a new connector layout with similar physical dimensions as DDR5 RDIMMs and doubles the data rate. At the same time, Tachyum claims that TDIMM memory will only increase signals by 38% but offer double the bandwidth, they will also require 10% fewer DRAM ICs, leading to 10% lower cost.
In the future, the company expects evolutionary changes to the TDIMM standard, which would increase the bandwidth to 27 TB/s, a massive increase versus the DDR6 standard, which would jump from 6.7 TB/s on DDR5 to 13.5 TB/s on DDR6. These changes are proposed for 2028.
Tachyum today announced details about how its TDIMM™ is bringing the future of AI and computing, enabling AI models with parameters many orders of magnitude greater than those of any existing solution at a fraction of the cost. TDIMM is key to reducing the estimated cost of OpenAI data center $3 trillion and 250,000 megawatts of power to $27 billion and 540 megawatts.
“The TDIMM is key in reducing the cost of AI systems trained on all the knowledge from $8 trillion and 276 gigawatts to $78 billion and 1 gigawatt in 2028,” said Dr. Radoslav Danilak, founder and CEO of Tachyum. “The TDIMM ushers in the era of affordable AI trained on all written knowledge produced by humanity, accessible to many companies and nations.”
via Tachyum
Once again, Tachyum has made some big claims, and we will have to wait and see if the TDIMM memory project moves beyond an announcement.
Follow Wccftech on Google to get more of our news coverage in your feeds.
