David and Goliath of the AI world: Graphcore outperforms Nvidia
Just a few years ago, it would take days and sometimes even weeks to train AIs. However, thanks to billions of dollars invested in new computing startups, performance and efficiency are now unprecedented. Nvidia has been the undisputed leader in the category, but Graphcore, a startup developing chips and systems to accelerate AI workloads, is taking on the industry Goliath.
Founded in 2016, Graphcore makes Intelligence Processing Units. The company believes that its IPU technology will become the worldwide standard for machine intelligence compute and transform several industries, including medical research, robotics, or autonomous cars. The IPU’s unique architecture lets AI researchers undertake entirely new types of work, not possible using current technologies, to drive the next advances in machine intelligence.
With the newest edition of the MLPerf training benchmark results, there’s clear evidence that the investments were worth it.
Nvidia worked closely with Microsoft on the benchmark tests, and, as in previous MLPerf lists, Nvidia GPUs were the AI accelerators behind most of the entries, including those from Dell, Inspur, and Supermicro. Nvidia itself topped all the results for commercially available systems, relying on the unmatched scale of its Selene AI supercomputer.
But there is a new major player now. Graphcore says its IPU-POD16 server easily managed to outperform Nvidia’s DGX-A100 640GB server. Specifically, when systems were tested to train computer vision model RESNET-50, Graphcore’s unit did the job almost a minute faster. It took 28.3 minutes to train the model, while DGX100 took 29.1 minutes. This represents a 24% jump over the previous MLPerf results. The company says it was able to achieve these results thanks to software optimization.
This positions Graphcore as a major rival for Nvidia, especially after the UK-based startup began moving into North America. It is taking aim at Nvidia’s data center GPU business with its IPUs and purpose-built AI systems, so it’s recruiting solution provider powerhouse Trace3 and other channel partners to ramp up sales in North America. “We’re growing both the company and the partner network at a really rapid pace, and that’s due to the fact that the market is absolutely hungry for options that are different than GPUs,” says Graphcore’s channel chief.
Graphcore’s go-to-market and research and development efforts received a significant boost during its latest funding round when it raised $222 million, bringing its valuation to $2.77 billion.
Graphcore has made it clear that it’s going after Nvidia’s fast-growing data center GPU business, claiming that its IPU architecture is better suited for AI applications because it was built from the ground up for “fine-grained parallelism” while also coming with 900 MB of on-board ultra-high-speed RAM, allowing the IPU to hold large AI models inside the memory. The company has also claimed that its M2000 systems provide more performance-per-dollar than Nvidia’s DGX systems.
“Graphcore is an amazing company that we are buying at Sakal Ventures. Artificial intelligence and machine learning will be revolutionary in the world. Let’s take the finance and trading sector, for example. Graphcore’s Intelligent Processing Unit is a completely new processing architecture designed for machine intelligence, capable of running advanced financial models up to 26x faster. It is no wonder that JP Morgan, D. E. Shaw & Co. and Citadel are among the company’s financial business customers,” says Kris Bort, founding partner of Sakal Ventures, a fund that focuses on late-stage private investing.
As part of our strategy at Sakal Ventures, we identify industries that are overdue for a meaningful digital transformation, developing a unique understanding of the growth trajectory of a business that is miscalculated in the marketplace. We provide an avenue for individual investors into the private market.
Right now, you have a unique chance to connect with the founder or head of capital development at our venture capital fund. Take this opportunity to talk to some of the brightest minds and best investors in the late-stage private space: visionaries, MIT and NYU Stern School of Business graduates. Book a FREE CONSULTATION with Sakal Ventures.