“People with very high expectations have very low resilience. Unfortunately, resilience matters in success,” he later said. “Greatness is not intelligence. Greatness comes from character.” And character, in his view, can only be the result of overcoming setbacks and adversity. To Jensen, the struggle to persevere in the face of bad, and often overwhelming, odds is simply what work is. It is why, whenever someone asks him for advice on how to achieve success, his answer has been consistent over the years: “I wish upon you ample doses of pain and suffering.”
“I don’t actually know anybody who is incredibly successful who just approaches business like, ‘This is just business. This is what I do from 8 to 5, and I’m going home, and at 5:01, I’m shutting it down,’ ” Jensen has said.15 “I’ve never known anybody who is incredibly successful like that. You have to allow yourself to be obsessed with your work.”
Jensen’s work ethics, resilience and obsession with his work are simply just inspiring.
“It turns out that by having a lot of direct reports, not having one-on-ones, [we] made the company flat, information travels quickly, employees are empowered,” Jensen said. “That algorithm was well conceived.”
This view sparked reflection on the value of 1:1 meetings I have had. Throughout my career, I’ve had varied experiences with them—ranging from laughably useless sessions dominated by small talk, to project heavy updates and discussions, to career coaching. Considering a world without recurring 1:1s is intriguing: unproductive chit chats vanish—I do not have to hear my manager talking about their gardening for the better part of the meeting. Work-focused conversations can shift to a broader audience, and career development can move to a quarterly cadence. Ultimately, the primary benefit of 1:1s seems to be providing a safe space for sensitive concerns. However, competent reports should proactively manage their own career growth and raise issues as they arise, instead of waiting for the next recurring meeting. The Nvidia Way seems like a compelling model that I’d like to experiment with in the future.
Another promising path was to modify the level of precision of the math that Nvidia’s GPUs could perform. At the time, the company’s GPUs supported 32-bit (single float, or FP32) or 64-bit (double float, or FP64) mathematical precision; either math type was a requirement for many scientific and technical fields. But deep-learning models didn’t need to be that precise. Models only required the GPUs to perform 16-bit floating-point calculations, because the networks were resistant to calculation errors during training. In other words, Nvidia GPUs did math that was overly precise—and therefore much slower—for deep-learning models. To make the GPUs run faster and allow these models to run more efficiently, in 2016 Dally implemented FP16 support on all Nvidia GPUs.
Never did I have the realization that GPU is on a totally level of bits than CPU. And the fewer bits the better for deep learning? That is just wild. I somehow assumed GPU also uses 64-bit architecture, which is a false impression influenced by how CPU works.
But Jensen’s wife, Lori, didn’t want him to leave LSI until Malachowsky also left Sun—and Malachowsky’s wife, Melody, didn’t want him to leave Sun until Jensen committed.
This anecdote made me laugh. At the end of the day, the ladies are the ones helping their husbands make the best career choices, and the best intent from both sides led into a dead lock.
Jensen had come up with DLSS on the spot. He had seen the promise inherent in one technology and transformed that promise into a new feature with a better business case. Now, if DLSS worked, the company’s entire product lineup, from the low end to the high end, would become more proficient, and thus valuable, allowing Nvidia to charge higher prices. “The researchers had invented this amazing thing, but Jensen saw what it was good for. It wasn’t what they had thought,” Luebke said. “It shows what a leader Jensen is and how technical and smart he is.”
As a gamer, I have a deep appreciation for DLSS. Games I’ve recently finished, such as Clair Obscur: Expedition 33 and the Diablo 4 DLC, would have run significantly less smoothly on my 3080 without it. I had no idea this technology was contemplated as far back as 2018 at SIGGRAPH, following Jensen’s push. It’s amazing to see such visionary leadership, and the time gap between the technology’s inception and when it finally reaches an ‘average Joe’ like me is fascinating.
Welcome to Shaobo's Blog! 