- Flash -based memory replacement for Syndis AI adds Top Top Computing Minds to List
- HBF Memory supports RISC and GPU leaders to promise high bandwidth and massive capacity
- Join the Sandesk to guide the extension of flash memory beyond Patterson and Kodori HBM limits
Sandsk has set two prominent personalities in computing to help the AI work load LIGH helps to form the direction of its high -capacity memory tech.
Professor David Patterson and Raja Kodori have joined the Sandesk’s new Technical Advisory Board to provide strategic and technical inputs to the High Bandothy Flash (HBF), which is a flash -based replacement by the High Bandothy Memory (HBM).
Patterson has been supported by low -developing instruction set computing (RISC) and inexpensive discs (RAID), and will lead to the advisory board. The Kodori graphics is known for its lead in the architecture, which has monitored the AMD and the GPU design in Intel.
You can like
Decades of experience
Simultaneously, they bring decades of experience in computing, memory systems and massive architecture.
“We have the honor of joining our two prominent computer architects experts on our technical consultation board,” said Alcabbar, Executive Vice President and Chief Technology Officer of Sandesk.
“Their collective experience and strategic advice will play an important role in creating HBF as a future memory standard for the AI industry, and confirming that we not only exceed the expectations of our customers and partners.”
Patterson said, “HBF shows the promise of playing an important role in the data center AI, providing unusual memory capacity in high bandwidth, which enables the burden of work on a scale far more than today’s obstacles. It can reduce the costs of new AI applications that are currently unbearable.
Kodori added, “HBF is ready to bring about revolution in Edi AI by equipment with memory capacity and bandout capabilities that will support locally -driven sophisticated models in real time. This development will open a new era of intelligent edge applications, which is mainly transformed.”
HBF is designed to meet HBM bandout, while 8 times more capacity is offered at the same price.
BICS Flash, CBA Weifer Bonding, and proprietary stacking, which allows 16 to die in the package, offers a new way to increase GPU memory without fully relying on expensive drums.
Although HBM is not a direct alternative, HBF shares the same electrical interface and requires only the least protocol changes.
Sandesk had previously shown how only HBM users could support AI GPU 192GB memory, but by connecting it with HBF, that number could reach 3TB.
In a sequence using only HBF, the memory capacity can measure up to 4TB.
This technology first appeared in the Sandesk’s future FWD 2025 Investor event in February 2025, as well as its roadmap for future HBF generations.
These updates show the increase in capacity and bandout over time, with some trade in energy saving.
By creating a consulting board and achieving open quality development, the Sandesk is trying to avoid closing the market in proprietary solutions.
This can help to gain traction against rivals like Samsung and SK Hanks, who both invest in HBM space.
(Image Credit: Sandsk)
