On March 10, 2017, the NVIDIA launched the GeForce GTX 1080 Ti for US$699, with the promise of being 35% faster than the GTX 1080 for the same launch price as its predecessor. Jensen Huang called her “The Ultimate GeForce”, faster than the $1,200 Titan X. This balance between performance and pricing would never be repeated in the company’s history. In this article, we will revisit this model, which, without a doubt, is one of the most iconic graphics cards in history.
Are you ready?
Jensen Huang took the stage in San Francisco on the eve of GDC 2017to announce the GTX 1080 Ti. “Gaming warriors, welcome! Are you ready?” declared the CEO before revealing that the PC gaming market had reached 200 million GeForce players globally, with the Steam base growing 4x in just four years.
The event started with two pieces of news: the GTX 1080 would be US$100 cheaper, and five lucky people in the audience would win Titan X Pascal. So, after demonstrating GameWorks DX12 running fluid and smoke simulations in real time, Jensen released the statement: “It’s time for something new… something that’s 35% faster than the GTX 1080… something new that’s faster than the Titan X.”
The audience erupted as Jensen held up the GTX 1080 Ti case, declaring, “Ladies and gentlemen, the GeForce GTX 1080 Ti — the ultimate GeForce.” The specifications confirmed on stage were impressive: 3584 CUDA cores, 11 GB of GDDR5X memory at 11 Gbps and bandwidth of “almost 500 gigabytes per second”. The GP102 chip, manufactured using a 16nm process using the Pascal architecture, consumed 250W with a seven-phase current delivery system.
The Engineering that doubled the memory bandwidth
Jensen spent minutes explaining how Pascal achieved such a brutal performance. “How is it possible that, straight out of the box, the 1080 Ti is 35% faster than the 1080?”, he asked, before revealing two proprietary technologies of the architecture.
The first was frame compression: “When we render to the frame buffer, we do so in a compressed form… each individual pixel is represented by fewer bits.” This multiplied the effective bandwidth without increasing the physical speed of the GDDR5X memory.
The second era block cacherevealed publicly for the first time at that event. “We took the screen and divided it into lots of small blocks… if we could put all of that on the chip, we could render at a much higher rate”. Jensen explained that the chip’s internal cache ran at over 1.5 GHz, while the external frame buffer was bandwidth limited. In scenes with intensive translucency, smoke, and transparency effects, the blocky cache shined.
“We effectively… between bandwidth compression and block caching, we effectively more than doubled the raw bandwidth”. This combination turned the theoretical 484 GB/s into something close to 1 TB/s effective in real loads — an engineering feat that explains why Pascal has aged so well.
11 GB of Memory: preparing for the future
Jensen justified the choice of 11 GB with concrete data. “With each generation, our frame buffers get bigger… games are getting prettier, textures are getting bigger, worlds are getting bigger”. The screen displayed graphics showing Total War, Call of Duty and Grand Theft Auto V consuming more and more video memory year after year
The decision to equip the 1080 Ti with 11 GB was prophetic: in 2026, mid-range cards still struggle with 8-12 GB, while the 1080 Ti already offered this space nine years ago.
The demo: over 100 Frames per second
To prove the performance, Jensen ran Epic Games’ Paragon character viewer — the same demo used at the launch of the GTX 1080 in May 2016, but which barely managed to deliver performance that didn’t occasionally drop frames on a 60Hz monitor. On the 1080 Ti, it was a different story.
“Over 100 frames per second, 35% faster than 1080… beautiful shaders, depth of field effects, beautiful lighting, beautifully shadowed,” narrated Jensen while the demo ran smoothly. The numbers in the corner of the screen confirmed it: 2 GHz frequency, temperature of just 66°C with the original cooler.
“This is the GTX 1080 Ti straight out of the box, stock cooler… with the overclocking margin we provide, you can get another 15, 18, 20%.”
The price that shocked the industry
Then came the decisive moment. “The ultimate GeForce, 1080 Ti, 35% faster than the 1080, faster than the Titan X, and now overclocked to over 2 gigahertz… all for $699”. The audience exploded in applause. Jensen laughed: “Don’t throw your wallets up here!”
The GTX 1080 Ti cost exactly US$699, the same launch price as the GTX 1080 a year earlier. This parity had never occurred before: when the GTX 980 Ti hit the market in June 2015, it cost US$649, representing a US$100 increase over the GTX 980. NVIDIA was, for the first time, delivering a Ti generational leap without increasing the price.
The Pascal generation in the context of 2017
The GeForce 10 series, codenamed Pascal, represented the 16th line of GeForce products released in May 2016. At the time of the 1080 Ti’s release, AMD’s competition was limited to the RX 580 (released in April 2017), a modest upgrade of the RX 480 that competed in the $200-250 range. AMD’s Fury cards were already aging, and the Vega line would only arrive months later.
Unlike 2026, when NVIDIA directs 90% of its efforts to data centers and artificial intelligence, and earns 12x more with this segment than with cards aimed at gamersin 2017 the focus was on players. Jensen opened the presentation celebrating: “PC gaming is thriving and growing… one of the leading phenomena in entertainment… the number of people watching others play video games on YouTube and Twitch has now reached 600 million, growing 50% annually.”
From GTX to RTX: changing priorities
The GeForce 10 series was the last major generation to exclusively carry the “GTX” prefix. In September 2018, NVIDIA introduced the “RTX” nomenclature with the 20 (Turing) series, marking the arrival of dedicated cores for ray tracing and artificial intelligence acceleration.
When the RTX 2080 Ti arrived in September 2018, it cost $999-1,199, up to 70% more expensive than the 1080 Ti at launch. The RTX 2080, with performance equivalent to the 1080 Ti, often cost $100 more than Pascal cards still available. The main feature, real-time ray tracing, had literally zero games supported at launch.
By 2026, NVIDIA’s transformation is complete. The company reports projected data center revenue of $170 billion for the fiscal year, driven by Blackwell and H200 GPUs for artificial intelligence training.
“You’re crazy if you don’t use AI to do everything,” says NVIDIA CEO
The Pascal architecture of the GTX 1080 Ti did not have dedicated cores for artificial intelligence, an item that became essential in NVIDIA’s strategy with DLSS. In 2017, machine learning was a niche area of academic research; in 2026, it is the main business that transformed Jensen’s company into one of the most relevant companies on the planet.
A plaque with a story to tell
The GTX 1080 Ti represents a unique moment in graphics card history, when NVIDIA prioritized delivering maximum value to gamers. Jensen closed the 2017 presentation by saying:
“I hope you all stick around so we can all enjoy together and play with the latest, greatest, ultimate GeForce”.
Source: https://www.hardware.com.br/artigos/imortal-historia-da-gtx-1080-ti-placa-de-video/
