Intel plans to expand its developer cloud so that its customers can test out new chips up to a year ahead of mass production.
The Santa Clara, California-based company is making many of its most advanced chips for data centers, including its latest “Sapphire Rapids” Xeon Scalable CPUs, available through a new cloud service called Intel Developer Cloud. Thus, some select pre-qualified customers and partners can make and fully test apps remotely, instead of having to buy physical silicon that’s likely still several months out due to ongoing delays.
At the Intel Innovation event held in Silicon Valley, the company said it is expanding the early-access cloud with its coming Xeon CPUs that integrate high bandwidth memory (HBM) and embedded Xeon D processors, based on its Ice Lake microarchitecture.
Intel is also unlocking access to its “Flex” family of data-center GPUs and high-performance Xe GPU code-named Ponte Vecchio. Both are key components in its effort to challenge NVIDIA’s dominance in the market.
This could aid developers who are at the heart of a software strategy that’s becoming a central focus for the U.S. semiconductor giant, as it looks to hook new customers and revive its market leadership.
CTO Greg Lavender said Intel is in full swing with a “software-first strategy” and reinforced its commitment to open software ecosystems at a time when it’s trying to raise its profile with a new generation of developers.
"In some ways, Intel is the largest software company that you've never heard of," said Nick McKeown, head of the network and edge business at Intel, in a briefing with reporters and analysts ahead of Innovation.
"What we’ve done in the past is to do the software that enables the low-level silicon to shine, but it's kind of invisible,” he noted. Intel, which touts itself as employing more than 20,000 software developers, said it hopes to make its developer cloud easier to use over time.
Hardware in the Cloud
Many developers of software running on Intel’s hardware have used public cloud services for years now. But to develop and test software, they typically had to wait for the cloud-computing firms to buy physical silicon and then install it in their data centers.
Intel plans to make new chips available in the Developer Cloud up to a year in advance. Doing so will allow companies writing software on top of Intel’s new server processors and accelerators to improve their offerings and test new features even faster while avoiding the hassle of buying the hardware.
The new offering resembles cloud services from the likes of Amazon AWS and Google in some ways. But it stands out by providing access to hardware and software primarily for test and development purposes.
With it, the company said customers can get a head start on product pre-qualification and building Intel-based solutions. In some cases, programmers will be able to get access before new processors are even publicly announced.
Over time, there will be a portfolio of cloud services to let customers perform application performance benchmarking, optimization, and troubleshooting on a broad range of Intel-based hardware and software. Among the other cloud-accessible chips will be Intel’s new Habana “Gaudi” accelerators for deep learning.
At launch, Intel said the Developer Cloud will have bare-metal servers and virtual machines (VMs). A wider range of configurations will be added throughout 2023, when it will start migrating other cloud services and development tools to the cloud.
Persistent Delays
The new cloud offering comes as CEO Pat Gelsinger tries to rebuild the company’s competitiveness after losing its leadership in chip manufacturing to TSMC and falling behind in performance to AMD and other rivals.
But Intel’s recovery has been hitting some obstacles. It continues to push out volume production of one of its most important new chips: the next-generation Xeon Scalable CPUs based on the Sapphire Rapids architecture.
Intel has repeatedly pushed out shipments of the data-center CPU. While the CPU was supposed to be ready by late 2021, the company last year said it wanted more time to improve the performance and warned it would start producing the server chip in early 2022 instead. The release date was further delayed to the first half of this year, then by the end of the year, and now likely by early 2023.
The delays have opened the door for rivals including AMD, Arm, and NVIDIA, as well as startups such as Ampere Computing, to stake out a stronger claim in the data-center CPU market that it has long dominated.
Intel is not withholding the CPUs from everyone. The company said it has begun shipping both the Ponte Vecchio GPU and HBM-equipped Sapphire Rapids CPUs to Argonne National Laboratory for use in the lab’s Aurora supercomputer.
For everyone else, they will likely have to wait longer to plug the new processors into their own data centers. But Intel is hoping the ability to test chips in the cloud is the next best thing for developers.
Lisa Spelman, corporate vice president of Intel’s server-chip business, said ahead of the Innovation event that she is looking forward to companies getting to test-drive the new Xeon CPUs.
"And this is a unique and rare opportunity to do so before you've even hit launch or had the opportunity to buy it and bring it into your own data center,” she said, citing Intel’s new try-before-you-buy cloud offerings.
Sapphire Rapids is one of the company’s most ambitious server processors to date, which is possibly why Intel is struggling to bring it to the market. The server chips will be the first in the Xeon CPU family to adopt its advanced 2.5D packaging technology with chiplets to cram in more transistors and, thus, CPU cores. Intel is also bringing PCIe Gen 5, CXL, DDR5, and other advanced features into the fold.
Senior fellow Ronak Singhal said Intel is putting quality front and center with its new generation of Xeon Scalable processors. “We will not compromise that quality even if it means delaying the product.”
He added, “We're going through a rigorous validation cycle, and when we find issues, sometimes it will push out the schedule. But we think that is the right tradeoff for our customers, and what they are expecting from us.”