Skip to main content

Nvidia RTX 3090 vs. RTX 3080: Here’s how they stack up

Nvidia’s next generation of GPUs is here. The new Ampere architecture promises a significant performance improvement, and Nvidia is making some ambitious claims about its potential for gamers.

The RTX 3090 and RTX 3080 sit at the top of the stack, and are both extremely powerful cards. But they’re also some of the most expensive mainstream cards Nvidia has ever produced. From performance to features to price, here are our thoughts on how the two compare.

Recommended Videos

Pricing and availability

RTX 3090
Image used with permission by copyright holder

Both the RTX 3090 and 3080 made their worldwide debut on September 1, 2020, when Nvidia announced slightly staggered release dates and drastically different pricing. The RTX 3080 will release on September 17 priced at $699, while the 3090 will go on sale on September 24 at $1,499.

Get your weekly teardown of the tech behind PC gaming
Check your inbox!

These prices and release dates are for Nvidia’s Founders Editions of the cards only. A number of add-in-board partners, including MSI, Powercolor, Palit, Gainward, and Inno3D, have already announced their own versions of the RTX 3090, 3080, and 3070. With a number of options for each of these top-tier GPUs, it could be that the third-party alternatives launch alongside, or very soon after Nvidia’s own Founders Edition options. Their pricing is likely to be a mix of slightly cheaper, and vastly more expensive cards, depending on their cooling and clock speeds.

Specs and performance

RTX 3090 RTX 3080 
GPU TBD TBD
Interface PCIe 4.0 PCIe 4.0
CUDA cores 10,496 8,704
Tensor cores TBD TBD
RT cores TBD TBD
Base clock 1,395MHz 1,440MHz
Boost clock 1,695MHz 1,710MHz
Memory 24GB GDDR6X 10GB GDDR6X
Memory speed 19.5Gbps 19Gbps
Bandwidth 936GBps 760GBps
Memory bus 384-bit 320-bit
TDP 350w 320W

Although Nvidia hasn’t addressed the possibility directly, it seems the 3090 is set to replace the Titan RTX. The specs certainly line up with that. Nvidia’s top-line Ampere GPU comes with a massive 24GB of GDDR6X memory, matching the Titan RTX. The RTX 3080 is more conservative with 10GB of GDDR6X, but both cards offer improvements in memory speed and bandwidth over the last gen.

Nvidia hasn’t provided any reference benchmarks for the 3090 yet, but the 3080 is set to overtake the 2080 Super and 2080 Ti. Digital Foundry had some hands-on time with the 3080 and found it could offer between a 60 and 90% performance improvement over the RTX 2080 (with a mild factory overclock) in a variety of Nvidia-chosen games, including Control, Doom Eternal, and Borderlands 3. That was both with and without ray tracing enabled, at 4K resolution, and with all settings at their maximum. DLSS was still needed to hit a comfortable 60 frames per second at 4K with ray tracing at its highest setting, but even that is a staggering achievement and shows how far the 3080 has come over its Turing predecessors.

The 3090 should be even better. How much more? It’s hard to say. It has 20% more CUDA cores, 23% more memory bandwidth, and almost 2.5 times the memory capacity of the 3080. According to Nvidia, this enables the 3090 to deliver 60 fps at 8K resolution with DLSS on.

That means it’s likely that the RTX 3090 will be able to offer 100 fps at 4K without ray tracing, or 60 fps at 4K without the need for DLSS.

Image used with permission by copyright holder

With all that said, we’ll need to wait to see how the 3090 performs in independently-tested benchmarks and games before drawing a hard conclusion.

Power and cooling

Nvidia

The RTX 3080 and 3090 both have higher power draw compared to the last generation. The cards demand 320W and 350W at stock, respectively, on the reference design, supported by Nvidia’s new 12-pin power connector. Third-party alternatives seem to do just fine with dual eight-pin connectors, so the necessity for Nvidia’s new connector design has yet to be determined.

Both power-hungry cards warranted a new Founders Edition cooler design, which involves a dual push-pull fan configuration and a V-shaped PCB. This new design should help air more easily pass through the case, and Nvidia claims it allows the cards to operate at much lower noise levels than previous generations, despite the additional performance.

The new cooler makes the 3080 quite a chunky card — almost an inch longer than the 2080 Ti and slightly taller, too. The 3090, described by Nvidia as a “BFGPU,” measures 12.3-inches long and 5.4-inches high, a full inch greater in each dimension than the 3080. That raises some questions over clearance issues in smaller cases. It’s also a triple-slot design, so don’t expect to fit this into a Mini-ITX case or motherboard.

Despite these enhancements, expect third-party alternative cooler designs to be more efficient and quieter — and possibly even bigger.

Ray tracing, DLSS, and more

Image used with permission by copyright holder

Ray tracing and DLSS are at the core of the 3080 and 3090. Nvidia’s redesigned RT cores deliver up to 2 times the performance of Turing with ray tracing turned on. Compared to last gen, both cards deliver higher frame rates with ray tracing, and, combined with DLSS, higher resolutions.

To take advantage of either card’s ray tracing abilities, you’ll need to play one of these games with ray tracing.

Although 4K is the focus right now, both cards support 8K with three HDMI 2.1 ports and a lone DisplayPort 1.4a connection. Nvidia’s benchmarks show the 3090 reaching over 60 fps in games like Apex Legends and Destiny 2 at 8K with ray tracing and DLSS.

Nvidia Reflex is new to RTX 30 series GPUs, too. With an RTX 3080 or 3090 and a G-Sync monitor, you can use Reflex to reduce system latency in competitive games. Nvidia’s benchmarks show a decent improvement in overall system latency with Reflex turned on, and a drastic improvement when paired with a high refresh rate display and a 30-series GPU.

A new RTX I/O is also intriguing, as it could leverage the GPU to handle data decompression from storage, helping to better take advantage of PCIExpress storage drives for even faster load times and in-game asset loading for reduced pop-in. That will be available on all Nvidia GPUs since the GTX 10-series, however.

The 3080 and the 3090 are almost the same, with the most significant difference being that the RTX 3090 supports dual-GPU setups through SLI. In contrast, the RTX 3080 doesn’t support that, and so it’s likely to have a greater number of RT and tensor cores.

A BFGPU with a price tag to match

The RTX 3090’s impressive specs tempt us to recommend it over the RTX 3080, but its price means we can’t do so in good conscience. In terms of replacing the Titan RTX, the pricey 3090 card is designed more for media creators and data scientists than gamers. This card will be the quickest GPU around for gaming once it becomes available in stores, but $1,499 is more than most people pay for their entire rig, let alone a graphics card.

With such a high price, only the most invested gamers are likely to be interested in the RTX 3090.

The RTX 3080 is a capable 4K gaming GPU and a better value at $699. With all the RTX 3080 offers and its reasonable price, this card will most likely be gamers’ ultimate high-end choice. There’s a reason Nvidia has called it its flagship GPU.

Topics
Jacob Roach
Lead Reporter, PC Hardware
Jacob Roach is the lead reporter for PC hardware at Digital Trends. In addition to covering the latest PC components, from…
Nvidia’s RTX 5090 may not be such a power-hungry beast after all
The RTX 4090 graphics card on a table alongside a set of cables held in hand.

The RTX 5090 will undoubtedly claim its spot atop the rankings of the best graphics cards when it's here, but that kind of power comes with a lot of ... well, power -- or more specifically, a high power draw. To that end, we've heard a lot of speculation about the RTX 5090 potentially decimating your power supply and needing over 600 watts of power. However, Segotep, a China-based PSU manufacturer, weighed in on the matter, and it seems that enthusiasts can hold off on buying a new monstrous PSU for a while yet.

It all started with claims that the RTX 5090 may feature not one, but two 16-pin power connectors. We've already seen some high-end PSUs sporting dual 12V-2x6 power connectors (made by manufacturers such as MSI), which could potentially power a GPU that draws far more than 600 watts.

Read more
Bad news for AMD? Nvidia might fast-track the RTX 50-series
Two RTX 4060 cards side by side

Things are finally about to start heating up for some of the best graphics cards. Although we're still in the dark about final release dates, both AMD and Nvidia are said to be launching new GPUs in the first quarter of 2025. However, a new leak tells us that Nvidia might try out a different approach with the RTX 50-series, and that's bound to put some pressure on AMD at the worst possible time.

What's new? We've already heard that Nvidia is likely to announce the RTX 5090 and the RTX 5080 at CES 2025, with its CEO Jensen Huang scheduled to hold a keynote during the event. However, the release dates for the rest of the lineup remained a mystery. Now, a previously reliable source sheds some light on the matter with potential details about the planned launch dates for the RTX 5070, RTX 5070 Ti, RTX 5060, and RTX 5060 Ti.

Read more
Nvidia may give the RTX 5080 a sweet consolation prize
The back of the Nvidia RTX 4080 Super graphics card.

Nvidia's best graphics cards are due for an update, and it seems that the RTX 5080 might get an unexpected boost with faster GDDR7 memory than even the flagship RTX 5090. That might be its sole consolation prize, though, because the gap between the two may turn out to be even bigger than in this generation.

First, the good news. Wccftech cites its own sources as it reports that the RTX 5080 will get 32Gbps memory modules from the get-go -- a significant upgrade over the RTX 5090 with its 28Gbps. The best part is that such a memory upgrade would bring the RTX 5080 to a whopping 1TB/s of total bandwidth, marking a huge improvement over the RTX 4080 Super, which maxes out at 736GB/s.

Read more