[rhdjsjebshjffn]

I'm excited about this too, but it's a little concerning there's a brand in the title. There's no shortage of those from ati, intel, amd, apple, ibm, the game gaggle, etc to interview. The fact that nvidia succeeded where others failed is largely an artifact of luck.

[jandrese]

I would say nVidia made their own luck. A lot of their success can be attributed to their management never losing sight of the fact that the software is just as important as the hardware. Both drivers and CUDA are key to nVidia's success. ATI and nVidia would trade places on quality of hardware, but there was never a question on the software side.

[rhdjsjebshjffn]

I'm not sure how nvidia's driver track record would have helped them, but drivers nor linux nor software in any way has ever really been nvidia's strong-suit. but even with the popularity of it CUDA cannot explain nvidia's success alone; you also need the demand of butcoin and the secondary-but-farcical imitation of LLMs but also the inexplicable lack of awareness of alternatives that need explaining...

[BobbyJo]

I worked on CUDA and OpenCL in the 2010-2014 timeframe, well before buttcoin and LLMs were profit centers, and Nvidia was already well ahead in the "GPUs as general compute" area. Literally everyone doing highly parallel HPC wanted to use Nvidia, despite AMD having higher throughout for some workloads. It was better, easier to use software.

[WJW]

I'll add to that: even though it is true that "drivers nor linux nor software in any way has ever really been nvidia's strong-suit", as GP put it, their software was still miles ahead of its competitors. In the land of the blind a one-eyed man is king, and all that.

[MichaelZuo]

I definitely recall Nvidia cards were consistently priced higher than AMD cards of the equivalent hardware as early as 2011.

So judging by actions, not words, they had a clear software/firmware advantage by then already.

[rhdjsjebshjffn]

I would agree if I didn't associate nvidia with incomprehensibly bad support for drivers. Idk, maybe this is a linux-only thing, but it's hard to imagine a vendor with a worse reputation for delivering functional software. Only perhaps microsoft itself has tried to be even more anti-consumer in their approach to support.

Cuda is ok, but it's the sheer mass of people targeting the platform that makes it more interesting than eg opencl. It hasn't don't anything clearly better aside from aggressively trying to court developer interest. It will pass and we will have stronger runtimes for it.

What really sets nvidia apart is its ability to market to people. Truly phenomenal performance.

[WJW]

Yea sure I see what you mean. So can you tell me what reputation AMD has for CUDA support on Linux? Or any of the other GPU providers?

They have none because their driver support is nonexistent. That's why everyone under the sun uses Nvidia despite their abysmal software support: it's still better than everyone else.

[FabHK]

The author worked at Nvidia and explains some of the engineering decisions made at the time. Why shouldn’t the brand be in the title?

[rhdjsjebshjffn]

Because the author is far more interesting than the brand? I'm not sure what would justify branding your personal observations outside of bandwagoning onto hopeful investors.

[mrandish]

> largely an artifact of luck.

I disagree with "largely". Luck is always a factor in business success and there are certainly some notable examples where luck was, arguably, a big enough factor that "largely" would apply - like Broadcast.com's sale to Yahoo right at the peak of the .com bubble. However, I'm not aware of evidence luck was any more of a factor in NVidia's success than the ambient environmental constant it always is for every business. Luck is like the wind in competitive sailing - it impacts everyone, sometimes positively, sometimes negatively.

Achieving and then sustaining substantial success over the long run requires making a lot of choices correctly as well as top notch execution. The key is doing all of that so consistently and repeatedly that you survive long enough for the good and bad luck to cancel each other out. NVidia now has over 30 years of history through multiple industry-wide booms, downturns and fundamental technology transitions - a consistent track record of substantial, sustained success so long that good luck can't plausibly be a significant factor.

That said, to me, this article didn't try to explain NVidia's long-term business success. It focused on a few key architectural decisions made early on which were, arguably, quite risky in that they could have wasted a lot of development on capabilities which didn't end up mattering. However, they did end up paying off and, to me, the valuable insight was that key team members came from a different background than their competitors and their experiences with multi-user, multi-tasking, virtualized mini and mainframe architectures caused them to believe desktop architectures would evolve in that direction sooner rather than later. The takeaway being akin to "skate to where the puck is going, not where it is." In rapidly evolving tech environments, making such predictions is greatly improved when the team has both breadth and depth of experience in relevant domains.

[jjtheblunt]

Nvidia’s Cg language made developers prefer their hardware, I’d say.

[pjmlp]

Which influenced HLSL, with their close collaboration with Microsoft on DirectX.

[voidspark]

Concerned about a "brand in the title"? What do you expect?

The author is David Rosenthal, who was employee #4 at Nvidia (Chief Scientist).

He's not some random historian or interviewer. That's his life experience.