Intel has given up on the AI race with NVIDIA & is now looking to launch a cost-effective solution with its new Gaudi 3 AI offering.
Intel Gaudi 3 AI GPUs Will Feature Industry-Leading Performance Value Per Dollar, But Won’t Compete With NVIDIA As Blue Team Shifts Target Market
It looks like Team Blue has finally realized that competing in the “computing power” race with NVIDIA is not the right path for them to have a sustainable business.
Instead, the company is now tapping into a relatively unpopulated segment of the AI business, which is a cost-effective implementation of AI accelerators that will likely entertain most of the industry. According to a CRN report, Intel is offering its latest Gaudi 3 AI GPU as a value proposition, coming with some of the best price-to-performance ratios available on the market.
Nanduri said that while the Gaudi 3 “can’t catch up” to Nvidia’s latest GPUs from a head-to-head performance perspective, the accelerator chip is well suited to enable economical systems to run both task-based and open source models on behalf of the company, where the company has “traditional strengths.” ”
via CRN (Dylan Martin)
Intel claims that the Gaudi 3 lineup offers performance on par with NVIDIA’s popular H100 AI accelerator, particularly in inferring workloads, which saw massive improvements following the debut of its “reasoning-focused” LLM models. In terms of actual numbers, Intel claims that the Gaudi 3 AI accelerator delivers 80% better performance per dollar value when compared to the NVIDIA H100, and when benchmarking on Llama-2, the performance per dollar difference increases by 2x. , which is impressive indeed.
The company is simply offering the new AI lineup as the best solution for small-scale startups and individuals looking to gain the computing power of AI. However, when tested in floating-point operations, the Gaudi 3 AI GPU felt short of NVIDIA’s alternatives, suggesting that robust AI performance isn’t Intel’s best option for now.
Team Blue realized that they couldn’t compete with NVIDIA in terms of hardware dominance. Surprisingly, the company claims that it is not trying to capitalize on demand from mainstream market players. In the long term, they believe that the smaller LLM model will be more widely adopted once the AI craze and the madness behind large-scale data centers fades.
The world we’re starting to see is people questioning that [return on investment]costs, energy, and others. This is where—I don’t have a crystal ball—but the way we think about it is, do you want one giant model that knows everything.
We feel like we’re at where we are with the product, the customers involved, the problems we’re solving, that’s our swim lane. The bet is that the market will open up in that space, and there will be lots of people building their own inference solutions.
– Intel’s Anil Nanduri via CRN
Intel Gaudi 3 AI solutions have been well adopted by the industry, most notably by IBM Cloud, Hewlett Packard Enterprise, and even Dell in their respective data center products for those markets. We can’t say that the Gaudi 3 didn’t catch the market’s attention, but Team Blue’s bet on the AI market hasn’t panned out, with the company’s previous CEO Pat Gelsinger calling NVIDIA’s CUDA a moat, but it’s turned out to be the dominant compute stack.
It will be interesting to see how the AI situation pans out for Intel in the future as the company is in dire need of help considering the financial conditions they are in.
