More on this book
Community
Kindle Notes & Highlights
by
Tae Kim
Read between
February 24 - March 23, 2025
its internal code name and dubbed it the RIVA 128, which encapsulated the chip’s ultimate purpose: RIVA stood for Real-time Interactive Video and Animation Accelerator, and “128” was a nod to the 128-bit bus, which would be the largest ever included on a single chip—another first for the consumer PC industry.
Standard chip development usually spans two years, involving multiple revisions to identify and fix bugs after a chip “tape-out,” when a finalized chip design is sent for prototype manufacturing. The NV1, for example, had three or four physical tape-outs. Nvidia could afford just one physical tape-out for the NV3 before the company had to send it to production. To shorten the timeline, Nvidia would have to shorten the testing cycle.
“Hire someone smarter than yourself.”
As he saw it, he needed to prevent the kind of internal rot that he observed at other companies, where employees often manipulated their projects to provide steady and sustainable growth that would advance their individual careers, when in reality they were making only incremental improvements that actually hurt the company in the long term. The “Speed of Light” notion ensured that Nvidia would never tolerate such sandbagging.
“What’s the main limiting factor in getting a graphics card to market?” Diercks responded that software drivers—the specialized programs that enable the operating system and PC applications to interface with and use the graphics hardware—were the primary obstacle, because they needed to be completely ready by the time the chip was prepared for mass production. In traditional production processes, the first step was to build a physical prototype of the chip. Once it was complete, software engineers could begin work on the drivers and fix any bugs they encountered. Then the chip’s design was
...more
To save time, Jensen decreed that Nvidia would have to develop the driver software for the RIVA 128 before the prototype chip was completed—a reversal of the customary process. This would shave nearly a year off of the production timeline, but it would require the company to find a way to bypass the step of testing the software on physical chips. That was why Nvidia invested $1 million in its Ikos emulator, even though every dollar was precious: it would allow them to approach the “Speed of Light.”
instantaneous “Mycelium Spore Drive.”)
Nvidia had software drivers ready at the beginning of chip production: the drivers would already have been tested across all the important applications and games and to ensure compatibility with prior Nvidia chips. This approach became a significant competitive advantage for Nvidia, whose rivals had to develop separate drivers for different chip-architecture generations.
“Graphics drivers are perhaps the most challenging piece of software in the PC after the operating system,” he said. “Every app touches it, and every app release or update can potentially break it.”
“There may be people smarter than me,” Jensen once told his executive staff, “but no one is ever going to work harder than me.”
the company applied focused ion beam (FIB) technology, which can modify chips at the microscale level. The FIB instrument looks like an electron microscope but does not use electrons: it uses ions to modify chip prototypes. The modified chips worked, saving Nvidia’s RIVA series from instant obsolescence.
“The first time we came in second place, Jensen sternly told me: Second place is the first loser,” Logan said.12 “I never forgot it. I realized I’m working for a boss who believes we have to win at everything. It was a lot of pressure.”
the RIVA 128 would apply dithering—a form of intentional noise that was designed to break up or obscure obvious visual irregularities—to certain types of renders, such as smoke or clouds.
“The biggest joy I get out of this job was to see my customers grow, make money, and succeed,”
To sweeten the deal, he structured the loans as convertible notes that, when the company introduced its IPO, could convert into equity at 90 percent of the eventual IPO price—which would give Nvidia’s potential creditors a far higher potential upside than that of standard loan interest.
“Oh my God, we got here and we thought there was going to be a secret sauce,” one engineer said.3 “It turns out it’s just really hard work and intense execution on schedules.” It was Nvidia’s culture, in other words, that made the difference.
