Intel introduced Rialto Bridge and Falcon Shores on Tuesday – new GPU platforms that the company says will increase performance and efficiency for high-performance computing needs in the data center market.
The semiconductor power announced new chips during the ISC 2022 event in Germany. Rialto Bridge focuses on artificial intelligence (AI) and high-performance computing (HPC) workloads and promises significantly increased density, performance and efficiency compared to the already impressive Ponte Vecchio – the company’s first GPU data center. Rialto Bridge is set to be issued in mid-2023.
Intel plans to follow Rialto Bridge to Falcon Shores, which will combine x86 with GPU technology into a single socket, which the company says will lead to a fivefold increase in performance over current GPU platforms. Falcon shores said it is scheduled for release in 2024. Falcon Shores’ unique architecture will allow multiple tiles to be used in a single drawer, which, according to Jeff McVeigh, vice president and general manager of Super Compute Group at Intel, will give customers more flexibility in maximizing performance. for their own needs.
“I think it’s a revolutionary change,” McVeigh told reporters during a news conference. “What’s different here is to make sure you can select and place different tiles and interfaces – so you can adapt over time.”
GPUs have become indispensable for supercomputer tasks, which include AI training, drug research, financial modeling, and medical imaging. While CPU-driven tasks are still powered by x86 technology (and Intel’s Xeon chips), the GPU is strongly dominated by Nvidia, and AMD is closely following. In the teraflop speed race, Ponte Vecchio placed among the latest offers from AMD and Nvidia on the 45TFLOPS.
Kent Tibbils, vice president of marketing for Intel ASI Corp. He said the race of three giants to dominate data center data centers is great for all players. “Industry always benefits when competing forces chase each other. I think there is a lot of interest in Intel’s new offerings. But it will have to be good. But Intel seems to be putting it on a level playing field for the GPU. “
However, data center partners are just as excited to introduce Intel’s unified development software, called oneAPI, which automatically selects the right Intel silicon for the job. Tibbils said Nvidia’s CUDA toolkit will be a tough competitor, especially for older users. “CUDA has been on the market for decades,” he said. “But there are really strong opportunities for new use cases – something where someone does something new that requires this level of computing.”
Tibbils said Intel will have to be competitive with both price and attractive specifications. “They may be a little slow in reacting on the price side,” he said. “I think they will have to get there and be really competitive at a price.”
Dominic Daninger, vice president of engineering at Nor-Tech, based in Burnsville, Minnesota, also pointed to CUDA’s position in the data center market. “I think Intel will have a real challenge to crowd out Nvidia in much of the GPU accelerator market, because Nvidia has such an edge with CUDA. We have a lot of customers who have invested a lot of work in CUDA and most of them are reluctant to move on to something else, “he told CRN US.
Roger Kay, technology analyst and founder of Endpoint Technologies Associates, said the GPU plan is still a bit out of focus – the company is likely to slowly release details of its chip plan as the game develops. “They’ve looked at the upcoming products and will probably start talking about the details,” he said, adding that Intel’s attempt to snatch the market from Nvidia would take years, not months. “In this game, Intel is playing catching up,” he said. “This transition is likely to be a ten-year period.”
Kay said companies will be fighting for the future of supercomputers for a long time. “Intel doesn’t have to win every battle,” he said. “He just has to win a little.”
This article originally appeared on crn.com
#Intels #GPU #data #center #questions #Nvidias #dominance