Lenovo, which is in attendance at the Intel Innovation 2023 event here in San Jose, made a very significant announcement of their own, beefing up their AI portfolio on multiple fronts at an event where AI was centre stage.
Today, Lenovo is announcing a series of edge AI services and solutions designed to enable mass deployment of remote computing capabilities that will empower new AI applications for any business. The majority of the offerings are Lenovo TruScale for Edge and AI, to give customers a pay-as-you-go model at the edge. In addition to the TruScale announcements, Lenovo introduced the new Lenovo ThinkEdge SE455 V3, which becomes the top of the line in their edge server portfolio.
“We have been focusing on AI workloads since 2017, so we have been in this for about six years now,” said Robert Daigle, Director of Lenovo’s Global AI Business, who leads the global AI business for Lenovo’s edge infrastructure group. Our first edge AI box was in 2019, and we keep expanding the portfolio based on what we hear from customers, We have gone from one purpose-built box to five systems specifically purpose-built for the edge.”
Much, but not all, of the announcements relate to TruScale, Lenovo’s infrastructure as-a-service offering.
“We are bringing the goodness of TruScale to edge computing,” Daigle said. “A lot of software partners that we work with sell SaaS, and many customers now want it all consumption-based, and they want it all on one bill.”
Daigle said that customer requests for edge offerings in TruScale led to Lenovo making some modifications to the core TruScale design.
“These requests for edge on TruScale really encouraged us to design TruScale to be more suitable for edge AI,” he indicated. “It’s very different from the data centre, which TruScale typically worked with before.”
These TruScale solutions, are, however, being targeted at a fairly broad audience.
“Companies with five employees use ChatGPT while Fortune 100s tend to train LLM in house,” Daigle said. Lenovo TruScale for Edge and AI provides organizations with immediate access to edge AI deployment and a one-stop connection to Lenovo’s 150+ turnkey AI solutions.
“Historically, AI has been very use case driven,” Daigle indicated. “What has been shifting with Generative AI, is LLM for multiple use cases. Most companies wont train their own Generative AI models, but they will take one of about 25 open source models and run inferences on it. That takes a lot less compute and a lot less skills needed in house, so it reduces barrier to entry.
“When we launched AI in 2019, we put something out there and listened to our customers, such as their wanting more GPU performance, or wanted the solutions to be quieter,” Daigle added. “By leveraging our turnkey solutions, we significantly reduce the cost of deploying and managing edge infrastructure solutions. That makes us a one-stop-shop for end-to-end, edge-to-cloud AI solutions.”
The new server, the Lenovo ThinkEdge SE455 V3, is the new top of the line model, designed to meet the needs of new LLMs for more processing power.
“This is now our highest performance edge box,” Daigle said. “AI workloads have often remained CPU-bound, even when GPUs are involved. This server uses the EPYC 8004 series processors to ramp up edge performance while still being quiet, as well as expandable. In addition, some companies still use rack servers in the back of the store for autonomous shopping, with leads to noise, thermal issues and security concerns. The SE455 has tamper detection and dust filtration covers – at a lower price point than rack servers.”
Daigle also highlighted the impact on these new offering of the Lenovo AI Innovators Program, which was launched a year ago and beefed up this summer with new financial resources.
“We have qualified over 45 AI companies in this,” he said, referring in articular to what they call their tee-shirt sized solutions.
“These are bundled reference architectures for different solutions, designed in Small, Medium and Large breakpoints,” Daigle sated. “It helps customers get started a lot quicker.”