Here are the top 5 Nvidia GPUs after 20 years
Top Team Green GPUs.
Nvidia, founded 1993, rarely loses. The quality of its products has bankrupted and driven out many competitors. It has promoted deep learning and AI hardware and made many of the best graphics cards. Still, its latest cards’ expensive RTX 40-series GPUs and questionable Frame Generation features are flaws.
Let’s forget the sad present and remember Nvidia’s exciting past. Since GeForce was founded 24 years ago, Nvidia has made many great GPUs, most of which can compete with AMD’s. Think these are Top 5 Nvidia GPUs, including families and cards.
1. RTX 3060
RTX 30-series looked promising in 2020. Architecture, ray tracing, tensor cores, raw performance, and price were better. Prices were far below expectations. A 30-series GPU or any graphics card was rare near MSRP.
After the 30-series launched in late 2020, Nvidia added lower-performance models. The midrange was crucial because the RTX 20-series’ 2060 and 2060 Super weren’t good GTX 1060 successors. Nvidia didn’t have to work hard on a new Top 5 Nvidia GPUs in 2021 because anyone could buy a pulse GPU, but the RTX 3060 surprised everyone.
Midrange Not all Nvidia products are good, but the RTX 3060 was. Double the VRAM (12GB) of the RTX 2060, it performed better. It had more VRAM than 10GB RTX 3080.
That AMD’s midrange cards, usually powerful, were weak was even more surprising. The RX 6600 and 6600 XT had decent horsepower but 8GB VRAM, FSR 1.0 upscaling instead of DLSS, and came out months later. VRAM capacity was usually better for AMD than the 30-series, except for Navi 23 cards.
Even after Nvidia’s first “self-owned” attempt to lock out Ethereum mining, the RTX 3060 had a GPU shortage. The shortage ended in 2022, making the 3060 a cheap GPU. The 6600 and 6600 XT dropped below $300, while the 3060 never reached $329. FSR 2.0’s quality and performance improvements made it more competitive with DLSS, reducing the 3060’s advantage.
Nvidia’s midrange GPU RTX 3060 remains strong. Great card that challenged AMD despite pricing and availability issues. The 30-series was usually too expensive or had too little VRAM, but this was the best. Unfortunately, the RTX 4060 undid most of that progress.
2.GTX 680
Nvidia rarely makes major mistakes, but Fermi was one. Fermi, introduced in the GTX 480 in mid-2010, only improved performance over the 200-series and consumed tons of power, not what Nvidia needed. Despite the dire situation, Nvidia released a second Fermi and GTX 500-series before 2010, improving product efficiency.
Fermi likely changed Nvidia’s strategy. Throughout the 2000s, Nvidia trailed Radeon (ATI and then AMD) in nodes. Newer nodes were more expensive and had “teething pains.” but had better efficiency, performance, and density. By the mid-2000s, Nvidia’s strategy was to make big Top 5 Nvidia GPUs on older nodes, typically placing GeForce first.
After the traumatic Fermi experience, Nvidia joined AMD at 28nm in early 2012. The first 28nm GPU, Kepler, differed from Fermi and previous architectures. The latest process made its largest version lean at just under 300mm2 and efficient. Nvidia and AMD flagships competed differently in 2012.
Nvidia released the Kepler-powered GTX 680 three months after AMD released the HD 7970. AMD’s HD 4000- and 5000-series GPUs were great, but the 680 was faster, more efficient, and smaller than the 7970. Nvidia only led these metrics slightly, but it led all three, which was rare or unprecedented.
Nvidia lost performance to the HD 7970 GHz Edition and better AMD drivers, but it led power and area efficiency. AMD faced another Kepler revision, which powered the GTX 700-series and forced the launch of a hot and power-hungry Radeon R9 290X. Fermi-like R9 290X beat GTX 780, but GTX 780 Ti reclaimed the title.
GTX 680 isn’t remembered, but it should be. AMD’s strategy helped Nvidia beat Fermi. Perhaps a better Top 5 Nvidia GPUs did the same thing and overshadowed it.
3. GTX 980
GeForce and Radeon battled for generations from the early 2000s to the early 2010s. Nvidia won most of the time, but ATI (later AMD) was close. One side was defeated only when ATI’s Radeon 9700 Pro defeated Nvidia’s GeForce 4 Ti 4600. Nvidia nearly repeated this scenario several times.
The stars must align for Nvidia by mid-2010s. Chipmakers like Nvidia and AMD’s GPU partner TSMC struggled to move beyond 28nm. Nvidia could make big Nvidia GPUs on old nodes without AMD responding with a new node. Top 5 Nvidia GPUs AMD was near bankruptcy, but Nvidia had more resources.
The perfect storm occurred when these two factors coincided. Kepler did well in the GTX 600- and 700-series, but Maxwell in the GTX 900-series (and GTX 750 Ti) was different. From 28nm, performance, power efficiency, and density improved.
The 980 was 15% faster, twice as efficient as the 290X, and cut 40mm2 of die area. The 980 was 40% more efficient, 10% faster, and 160mm2 smaller than the 780 Ti.
Though not as impressive as the Radeon 9700 Pro, this was a huge win. Like AMD’s HD 5870 attack on Nvidia. AMD had no Nvidia GPUs deterrent to Nvidia. AMD’s Radeon 200-series barely survived 2014.
AMD tried again in 2015, but only high-end. From low to high midrange, it refreshed the Radeon 200-series as the 300-series and used its new Fury lineup for the top. With a larger Maxwell GPU, Nvidia’s GTX 980 Ti defeated AMD’s R9 Fury X. The 980 Ti was preferable to the 4GB Fury X (a decent card) because it had 6GB of memory.
Though Nvidia won, the GTX 900-series changed gaming graphics cards forever. The Fury X was AMD’s last competitive flagship until the 2020 RX 6900 XT because AMD stopped making them every generation. AMD makes flagship Nvidia GPUs again (knock on wood), but Maxwell crippled Radeon for years.
4.8800 GTX
Modern graphics cards emerged in the early 2000s as Nvidia and ATI improved. Nvidia’s GeForce 256 introduced hardware-accelerated transform and lighting, while ATI’s Radeon 9700 Pro showed GPUs should be big and have more computational hardware. After losing big to the 9700 Pro in 2002, Nvidia made bigger and better GPUs.
Though ATI started the arms race, Nvidia was determined to win. In late 2006, Nvidia and ATI had GPUs as large as 300mm2, but Nvidia’s Tesla architecture scaled to nearly 500mm2 with the G80 chip. That size is typical for a flagship Nvidia GPUs today, but not then.
Tesla’s GeForce 8800 GTX hit AMD in late 2006 like the Radeon 9700 Pro did Nvidia four years earlier. Size helped the 8800 GTX beat ATI’s flagship Radeon X1950 XTX, which was almost 150mm2 smaller. Because it was fast and power hungry, the 8800 GTX normalized GPUs with 150+ watt TDPs, even though it seems quaint now.
BFGPU creator ATI couldn’t compete with 8800 GTX. HD 2000-series, at 420mm2, couldn’t compete with G80 chip and Tesla architecture. ATI shifted to smaller, more efficient Nvidia GPUs with higher performance density. The HD 3000-series flagship HD 3870 was surprisingly small at just under 200mm2, and the HD 4000- and 5000-series followed with similar die sizes.
Nvidia now replaces powerfu lNvidia GPUs with even more powerful ones to show AMD who’s boss, but not then. The Tesla architecture was so good that Nvidia used it again for the GeForce GTX 9000 series, a faster 8000 series. Despite being almost half the price of the 8800 GTX, the 9800 GTX was boring.
The 8800 GTX’s age belies its modernity. It had two 6-pin power connectors, an aluminum-finned cooler, and a high-end GPU die size. Despite being a modern GPU, it only supported DirectX 10, which didn’t survive.
5. GTX 1080 Ti
The Nvidia GTX 600-, 700-, and 900-series 28nm GPUs were hot. Every win over AMD’s cards was bigger. After the Fury X, AMD stopped making flagship Nvidia GPUs, leaving Nvidia alone for the next generation.
AMD left as TSMC’s 16nm node entered volume production. AMD used GlobalFoundries, which licensed Samsung’s 14nm, to make graphics cards because TSMC was too expensive. Note that TSMC’s 16nm node was better.
Despite its great GTX 900-series architecture, Nvidia went 28nm to 16nm. Nvidia outperformed AMD in process nodes for the first time in a while with TSMC’s 16nm. A shrunken Maxwell, the 16nm Pascal architecture introduced few new features, mostly for VR, which Nvidia (and AMD) were disappointed by. Only Rise of the Tomb Raider used Voxel Ambient Occlusion (VXAO), but it improved ray tracing.
GTX 10-series shrank without losing power. The 2016 flagship GPU GTX 1080 outperformed the GTX 980’s 700-series improvements. The 1080 was 30% faster, twice as efficient, and half the size of the 980 Ti. Pascal’s 16nm shrink optimized Maxwell after 28nm.
AMD’s old R9 Fury X, regular R9 Fury, and compact R9 Nano competed with the GTX 1080 and 1070 in 2014. Instead, AMD introduced the RX 400 series, starting with the 480. Nvidia’s GTX 1060 was good and efficient like the 1080. AMD’s specialty was midrange GPUs, and driver updates and the 480’s 2GB VRAM kept it competitive.
The GTX 1080 won 2016 and 2017 when AMD’s RX Vega flagship GPUs debuted. AMD performed similarly to the 1080 and 1070 but was less efficient. The GTX 1080 Ti, a larger Pascal GPU, got ahead of AMD three months earlier. Fury X mostly matched the 980 Ti, but Vega 64 couldn’t match the 1080.
GTX 10-series performance, efficiency, and price are remembered. Since the RTX 20-, 30-, and 40-series lack the 10-series’ diverse product stack, it may be Nvidia’s last great graphics card series. Nvidia released good $100 GTX 1050 and $700 GTX 1080 models over time. The GTX 10-series coincided with peak PC gaming.
GeForce RTX 4090
They debated including the GeForce RTX 4090, which is absent from this list. Great hardware and fast GPU, but complicated. They concluded with an honorable/dishonorable mention. The GPUs were best and worst.
Like the 980 Ti and 1080 Ti, the 4090 is the fastest gaming GPU. The 4090 outperforms AMD’s RX 7900 XTX in ray tracing and DLSS over FSR despite having similar horsepower. Though smaller than the 1080 Ti’s, the 4090’s performance lead is impressive.
The 4090’s two main drawbacks are noticeable. The 4090 uses the problematic 12VHPWR connector and consumes over 400 watts, making power issues worse. The power plug has been discontinued and revised, but many 4090 cards made before the redesign are unaffected. Fermi didn’t burn, so 4090 was likely a disaster.
PC gaming has become more expensive in the past five years, as shown by the 4090. At $1,600, it’s as expensive as a good gaming PC. The 3090 was designed as a Titan-class luxury GPU, but the 4090 merely follows it in price. The 3080 was slower than the 3090, but the 4080 was much slower.
A direct successor to the 980 Ti and 1080 Ti is unlikely due to the 4090’s high price. Launched at $649 and $699, the 980 Ti and 1080 Ti were affordable. The cheapest 4090s cost $2,000 or more, so its ridiculous $1,599 MSRP is unattainable.
Despite its strengths, the RTX 4090 is not Nvidia’s best GPU. Not as good as its predecessors. Instead of melting connectors and costing $1,000, it could have been the next GTX 1080 Ti. They await a champion.
