跳到正文
SiliconANGLE· Jonathan Anthony·· 4 天前AI 評分36

Dell:AI 運算開始利用各種零散電力,從商店插座到甲烷洩漏

From store outlets to methane leaks, every scrap of power is now fair game for AI compute

AI 導讀

Dell押注完整系統方案,支援規模不斷擴大的 AI 工廠及分散式推理,並將零散電力轉化為 AI 運算能力。Dell正出貨其稱為首款 100% 液冷網絡系統的產品,預先測試後以單一機架交付。Dell亦指出,客戶希望以單一控制平面連接最多 1,000 個小型站點,利用各處可用電力。

正文

From store outlets to methane leaks, every scrap of power is now fair game for AI compute

Avatar photo

AI factories are outgrowing the traditional data center as models get bigger and distributed inference moves from concept to deployment. The infrastructure behind those models now has to span racks, sites and even scattered pockets of power.

Dell Technologies Inc. is betting that the answer lies in building complete systems rather than components, from rack-scale management to fully liquid-cooled designs. The size of modern AI clusters dwarfs anything high-performance computing has handled before, according to Ihab Tarazi (pictured), senior vice president and chief technology officer for AI, compute and networking of Dell.

“The difference with AI is the scale is massive. Tens of thousands, hundreds of thousands of GPUs that are all on a single fabric. It may be multiple clusters, but they end up being a single fabric at the end of the day,” Tarazi said.

Tarazi spoke with theCUBE Research’s John Furrier and Rebecca Knight at SC26, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed Dell’s 100% liquid-cooled networking system, distributed inference and turning stranded energy into AI capacity. (* Disclosure below.)

Liquid cooling sets the stage for distributed inference

Dell’s latest move takes liquid cooling beyond the GPU. The company is shipping what it calls the first 100% liquid-cooled networking system, with switches, optics and cabling pretested in its factory and delivered as a single rack, Tarazi explained.

“The fact that it’s pretested as a single-rack system means, instead of having to assemble hundreds of pieces on-site and take maybe two weeks or three weeks to find the bugs, it all comes in, you install it [and] we turn it up in hours,” Tarazi said.

The networking launch follows Dell’s debut of liquid-cooled storage for AI racks. Compute, networking and storage now share the same cooling and power specs. That matters as operators design new facilities around water cooling, according to Tarazi.

“Instead of building two data centers — one for GPU, one network — being able to do one is a massive reduction in time and cost for us,” Tarazi said.

Software is the next lever for distributed inference. Customers want a single control plane that can work across different GPUs and CPUs and link as many as 1,000 small sites that each have some spare power, Tarazi noted. That approach turns stranded energy into AI capacity for companies that can’t build a massive data center quickly.

“You’ll be surprised what power availability means,” Tarazi said. “It doesn’t mean just that you have power in your store. It could also mean methane coming out — every source of power possible. People are going backwards to say, ‘How do I leverage that power for compute and enable inferencing in a distributed manner?'”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of SC26:

(* Disclosure: Dell sponsored this segment of theCUBE. Neither Dell nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE

A message from John Furrier, co-founder of SiliconANGLE:

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

  • 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more
  • 11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network

SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios — with flagship locations in Silicon Valley and the New York Stock Exchange — SiliconANGLE Media operates at the intersection of media, technology and AI.

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

來源:SiliconANGLE · siliconangle.com