For 25 years, the Texas Advanced Computing Center (TACC) has been one of the world's premier hubs for open scientific computing. It has grown from a group of 12 people with a single loaner system into an operator of some of the largest open-compute clusters in the world, several of which have supported Nobel Prize-winning research. In this video, TACC shares why it chose to run SambaStack and what that means for the researchers who depend on it.
TL;DR
- A 25-year open-science leader, TACC runs SambaStack to accelerate AI inference for researchers across every scientific and engineering field.
- SambaStack keeps high-bandwidth memory close to the processing and is purpose-tuned for AI workloads, delivering low energy use, low footprint, and low latency.
- The platform integrates cleanly into TACC's existing infrastructure, including its Splunk-based logging and other services.
- TACC values the SambaNova partnership as much as the technology: A team that works to understand and solve its challenges.
AI becomes science's new companion
Across every scientific and engineering discipline, the way research gets done is changing at a fundamental level. Generative AI has become a new companion for researchers, giving them faster ways to move through enormous volumes of complex data and surface the insights that matter. For a center whose mission is to help people solve hard problems, that shift makes efficient inference a first-class priority.
Why TACC runs SambaStack
“We're always interested in new ways to make computing go faster. That's our job.”
What drew TACC to SambaNova was the clever engineering behind both the hardware and the software. SambaStack pairs high-bandwidth memory closely with the processing itself, and it is tuned specifically for AI workloads rather than acting as a general-purpose tool. The result is a system that gives TACC the flexibility it needs while staying low in energy use, low in footprint, and low in latency: a highly efficient platform the team describes as the best of both worlds.
Built to fit existing infrastructure
Adopting a new platform only pays off if it fits the environment around it. TACC highlights how straightforward it has been to get in, work with SambaStack, and integrate it exactly the way the center needs, connecting it into the broader infrastructure alongside services like its Splunk-based logging. Rather than forcing TACC to build around the system, SambaStack slots into the operations already in place.
Not just hardware, but people
“It's been a very, very good experience working with the team at SambaNova.”
Beyond the technology, TACC points to the partnership itself. SambaNova does more than come in to fix and configure a system. The team works alongside TACC to understand its challenges and how best to meet its needs. As TACC puts it, it is not only the hardware, it is the people, and that combination is what defines any successful facility.
FAQs
TACC was drawn to the engineering behind both SambaNova's hardware and software. SambaStack pairs high-bandwidth memory closely with the processing and is tuned specifically for AI workloads rather than serving as a general-purpose tool. For a center whose job is to make computing go faster, that combination delivers the flexibility TACC needs while keeping energy use, footprint, and latency low.
TACC uses SambaStack to bring fast, efficient AI inference to scientific research. As generative AI becomes a new companion across every science and engineering discipline, TACC's researchers rely on it to move through enormous volumes of complex data and surface the insights that matter, integrating AI directly into their existing research workflows.
TACC was able to get in, work with the system, and integrate it exactly the way the center needed, connecting SambaStack into its broader infrastructure alongside services like its Splunk-based logging. Rather than forcing TACC to rebuild around a new platform, SambaStack slots into the operations already in place.