[Podcast Summary] Nvidia: The GPU Company (1993–2006) on Acquired
- Title: Nvidia: The GPU Company (1993–2006) — The Complete History & Strategy of NVIDIA: Part 1
- Author: Acquired Podcast
- Link: https://www.acquired.fm/episodes/nvidia-the-gpu-company-1993-2006
- Reading End Date: Jan 1st, 2023
Jensen Huang — school life
Huang was born in Tainan, Taiwan. His family emigrated to the US when he was 9 years old. At first, they found a boarding school that is cheap enough that they can afford. It is called Oneida Baptist Institute and it is in eastern Kentucky. It turned out that OBI (Oneida Baptist Institute) was so cheap was it’s actually not a prep school. It’s a reform school, and he made lots of friends there (and workout, which made him jacked).
Huang received his undergrad degree in EE from Oregon State University, and his master’s degree from Stanford University (which took him 8 years).
Jensen Huang — career
After college he was a director at LSI Logic, and before that, he was a microprocessor designer (more like PM) at AMD. The reason he went to AMD was he thought he could do it all. But later he realized that it was not cool. His passion lies in building a platform that allow everybody to make chips, without worrying about the tools. He wants to facilitate others to focus on a specific part.
LSI Logic
LSI Logic is a company that makes custom chips (first ASIC company) that accelerated storage and networking in data centers. The custom design chips would be for a very specific function to be integrated into other systems, e.g., Lockheed Martin’s defense tools.
LSI Logic made Don Valentine the largest venture return in an IPO in history. Sun Microsystems was one of their biggest customers to build chips on workstations.
3D Graphics Pitch
Two of Jensen’s buddies who he’s become close with at Sun, Chris Malachowsky and Curtis Priem, pitched to Jenson an idea — they see 3D graphics are really becoming a thing.
The way 3D graphics are done, you need SGI workstations. You need very high-end, very expensive stuff, only something with the budget of either the military or a Jurassic Park (which came out at 1993) can afford to do this, but people love it.
The pitch is about making graphics card for gaming. A chip that can accelerate the graphics of a normal PC to enable 3D graphics, just like SGI is doing with professional workstations.
Jensen decided to resign, and he approached to the LSI logic CEO, Wilf, who later introduced Don, founder of Sequoia to Jensen. Jensen messed up the pitch, but still got some funds. “Wilf says to give you money. Against my best judgment, I’m going to give you money. But if you lose my money, I’ll kill you.” Classic Don line.
Get Adrian Hsu’s stories in your inbox
Join Medium for free to get updates from this writer.
However, the business did not work. They weren’t the only people in Silicon Valley that saw that kids want to play games on PCs. With Doom, Microsoft saw that there are graphics cards companies out there now that are doing this. Microsoft decided to create standards.
Developers want to do 3D graphics directly into Windows without any of this crufty middleware from some no-name company like NVIDIA out there. Hence, DirectX/Direct3D were created.
Compete on Performance — RIVA 128
Jensen were tended to set the standard, the NV1, but unfortunately Windows came into play, and they set the standard successfully. NV1 was a failure. The only thing NVIDIA can do is standardize on the same Microsoft Direct3D as everyone else. They still have a final bet, RIVA 128 — designed to accelerate Direct3D to the utmost extent possible. It was built to render within the Direct3D and OpenGL.
Jensen figured that the only shot is to compete on performance among all GPU providers, and try to become the best chip out there in this now sea. His co-founders don’t want to do this because it was not exciting. Also, they think great performance does not guarantee anything.
They have to lay off 70% of the company. Everybody who’s staying knows we now have to design from scratch, and ship a new chip before our runway runs out, which is 9 months. A normal chip design cycle will not work. NVIDIA is not using TSMC at this point because they can’t. TSMC only works with the best and NVIDIA is not the best. They’re using secondary foundries. That process takes a long time.
They were running out of time. One startup built some new machines out there that enable emulation of chips, and Jensen decided to use their product to emulate the graphics chip. It’s all in software. If they want to ship something in nine months, they don’t have time to actually have it execute on the hardware. They decided to go right to manufacturing without ever producing a physical prototype and ship that.
The competitors are not going to be able to get them out for 18 to 24 months. NVIDIA is going to get the same generation of design out in six months. This chip is called the RIVA 128. This super unique chip development process beats all the competitors.
Convince customers
RIVA 128 is powerful, way more powerful than others. However, because they built this thing in such a manner, it barely works. Essentially, Direct3D at the time had 24 different ways and techniques (blend modes). The RIVA only works about two-thirds.
Jensen was actually going into these developers trying to convince them to do this. He talked to them and being like, “come on, what do you really need more than these 8 out of 24 for? What are you really going to do where you need to use that fancy stuff?”
Because the only thing that matters is performance, not the number of features. Consumers are going to buy hardware and games based on the quality of the graphics. This is being discovered for the first time. People are willing to make a lot of compromises in service of performance. They sell one million units of the RIVA 128 within four months.
NVIDIA naming
Nvidia initially had no name and the co-founders named all their files NV, as in “Next Version”. And then they had, “invidia”, the Latin word for “envy”. It became NVIDIA at the end.
TSMC
After the RIVA 128 hits and has become a big success, Jensen writes a physical letter and addresses it to Morris Chang in Taiwan. Jensen picks up the phone. Morris is like, “hello, this is Morris Chang at TSMC. I got your letter”. And then he hears Jensen yelling, everybody shut up. Morris Chang is on the phone. That’s how TSMC became the manufacturer of NVIDIA chips. It’s a landmark deal for both companies.
GeForce 3 — get out of commodity business
The GeForce 3 introduces programmable shaders and lighting on the GPU. The GPU massively parallel can light all these pixels, but previously, it’s essentially just taking instructions that are pre-hardcoded and baked in on what the lighting is going to look like. In GeForce 3, developers can program for these GPUs to make dynamic lighting and 3D graphics that are calculated.
The way to think about it is those “GPUs” were fixed-function graphics accelerators. This is different than a graphics card. NVIDIA develops in conjunction with this. They call it CG, something they would later do with CUDA. Literally, they extend the C programming language with graphics. One can directly program graphics, lighting, and shaders for the GPU.
This is how Jenson get out of this commodity business and do something unique and different. This is game-changing. They created an industry. They create the APIs, the SDK, the interface with it — do all this. Moreover, they’re doing it with Microsoft this time, instead of against Microsoft.
