Broadly speaking, there are three major segments that AMD and Nvidia make graphics cards for: gaming, professional applications, and the datacenter. But there used to be a fourth segment that's been long forgotten, and while this part of the market was fairly niche, it nevertheless got lots of attention from both AMD and Nvidia: prosumer.

A portmanteau of professional and consumer, prosumer cards offered a middle ground between gaming GPUs and professional, workstation GPUs. Nvidia's legendary Titan brand was for prosumers, and AMD also made lots of prosumer cards though not under one consistent brand. But it's been five years since we last got a Titan, and just shy of five years since we got a prosumer card from AMD. The prosumer GPU may have died for a good reason, but it has ultimately left a gap in the graphics market.

The rise and decline of the prosumer GPU

There isn't a strict definition for what it takes for a GPU to be "prosumer," but generally a prosumer card will be optimized for both gaming and professional applications like SolidWorks and Maya. These GPUs also usually came with the biggest or nearly the biggest graphics chip available as well as lots of VRAM. In their heyday, many of them would have been considered some of the best GPUs, though their ultra-high price tags meant they were relegated to the top-end of the top-end.

But prosumer GPUs didn't just use high-end hardware, they also offered less locked-down hardware. Because gaming and professional GPUs tend to use the exact same chips, AMD and Nvidia artificially lockdown and lower performance of gaming GPUs in professional applications. Prosumer cards were less locked down when it came to compute workloads (like those that rely on FP64 performance), but they didn't get as many driver-level optimizations as the professional cards and often had inhibited FP64 performance.

The first modern prosumer GPU was Nvidia's GTX Titan from 2013, and both its incredible gaming and compute performance was praised. At $1,000, the Titan was also by far the most expensive consumer graphics card to use one chip, and unlike today, $1,000 was a lot of money to ask for back in 2013. But compared to the $3,500 you'd have to spend on an Nvidia Tesla GPU using the same graphics chip, the Titan was a bargain.

Nvidia mostly ruled the prosumer segment with its Titans from 2013 to 2018, but AMD also tried its hand several times. The Radeon Pro Duo in 2016 combined two Fury X GPUs into a single card, though despite the "Pro" in its name, the Pro Duo really was made for gaming more than anything else. The Vega Frontier Edition in 2017 was the first truly prosumer AMD card, and compared to the RX Vega 64, it had 16GB of HBM2 rather than 8GB and supported Radeon Pro drivers. In 2019, the Radeon VII was much the same but went a step further and had less cut-down FP64 performance.

But after the Titan RTX in 2018 and the Radeon VII in 2019, we've not seen a prosumer card since. While the Titan brand was certainly very successful in the minds of those who were into PC during the time of the Titan, it was only a thing for a very short amount of time. Nvidia ostensibly replaced the Titan with the RTX 3090, and AMD quietly abandoned prosumer after the failure of the VII.

Prosumer GPUs offered too good of a deal

When it comes to gaming, I can understand why these cards are no longer offered. While the marginally better performance of the Titan was somewhat justifiable at $1,000, its gaming prestige started falling apart with the $1,200 Titan X Pascal and Titan Xp, partly due to the naming confusion as well as the higher price. With the Titan V and Titan RTX, people simply stopped viewing them as gaming GPUs, and Nvidia felt comfortable marketing the RTX 3090 as a Titan successor, and it was for gamers since they never cared about the professional stuff anyway.

It was really the professional features of prosumer GPUs that made them undesirable for Nvidia and AMD after a point. Nvidia very quickly stopped offering good FP64 performance on Titan cards (except for the Titan V, the most expensive one) and while AMD wanted to launch the Radeon VII with pretty low FP64 performance, the company ended up increasing it to presumably help it sell better given how uncompetitive it would have been otherwise. It's pretty likely Nvidia and AMD felt they were making less money than they could otherwise.

Nvidia and AMD just want people to buy their multi-thousand-dollar professional GPUs, and it's kind of scummy, but that's just business I guess. It's not terrible that businesses have to pay a little more for good GPUs, because they can afford it, but it's a shame that the death of the Titan and AMD's various cards have squeezed the hobbyists, the DIYers, and the aspiring professionals out of this sort of stuff. The Titan at least offered something people could work with despite it still being locked down, and the Radeon VII in particular offered great FP64 performance for its price.

The prosumer segment is probably done for good

The possible comeback of prosumer GPUs has been tossed around for the past few years, and some rumors indicate that it might happen or could have happened. There was a leak for a four-slot Titan RTX GPU about a year ago, though it's hard to say whether it's real or fake, and if it was real, then it's possible that Nvidia canned it. At the very least, Nvidia hasn't released a Titan based on the RTX 40-series' Ada Lovelace architecture.

While the discussion surrounding this rumored Titan certainly shows there's still a little reverence for the Titan brand and prosumer GPUs in general, it's hard to imagine Nvidia and AMD offering such a product again so long as it cuts into their margins. Maybe there's a path for Nvidia and AMD to offer cards that appeal to users of very specific applications, but there's very little chance we'll ever see a return to the crazy bang for buck that many prosumer GPUs had when it came to compute performance and professional apps.