Unlike most things in the world of consumer hardware and software, new graphics processing units are not released according to an annual upgrade cycle. Several years sometimes pass between major generation refreshes, and so each new series tends to introduce significant changes and improvements in performance. It’s now been almost exactly two years since Nvidia launched its GeForce RTX 20 lineup, and we’ve only just seen the launch of its successors, the GeForce RTX 3090, the GeForce RTX 3080, and the GeForce RTX 3070. As anticipated, these models are based on a whole new architecture, codenamed ‘Ampere’, which Nvidia says represents its “greatest ever generational performance leap”.
Nvidia of course has plenty to say about its new architecture — the main points are that RTX ray tracing is here to stay, power efficiency is improved, and prices are not much higher considering what you get. The first GPU in the GeForce RTX 30 series that we can get our hands on is the middle of the three, the GeForce RTX 3080. It’s aimed at high-end gamers who want to push the highest quality settings and all the effects turned on at 4K, probably while streaming everything.
As usual, Nvidia will sell its own Founders Edition graphics cards based on these new GPUs, and this is the version of the GeForce RTX 3080 that we’re reviewing today. The company has come up with a unique new hybrid cooling system and has also continued to refine its industrial design. Cards from partner brands including Asus, Zotac, Gigabyte, MSI, Palit, Galax will also soon be available, with their own coolers and custom tweaks.
There’s a lot about the GeForce RTX 3080 GPU and the Founders Edition graphics card to cover here, so let’s get started with a quick rundown of what’s new.
Nvidia GeForce RTX 3080: Ampere architecture overview
Last year’s GeForce RTX 20 series, codenamed Turing, introduced hardware ray tracing and AI-accelerated DLSS (Deep Learning Supersampling). For a while following the launch, these features weren’t widely adopted by games and seemed to offer little in the way of substantive improvements but have both since become more commonly supported over the past two years. Nvidia says that Ampere can process some operations such as shader algorithms over twice as fast.
We now have second-generation RTX ray tracing cores and third-generation Tensor cores for AI processing, plus faster cache memory and GDDR6X RAM. Performance improvements are said to be between 1.7X and 2.7X, plus new features have been added to leverage AI, such as live video and audio enhancements for streamers.
AI still primarily comes into play with DLSS, which is said to intelligently upscale frames so that the GPU can target a lower rendering resolution and work faster, but quality isn’t ultimately compromised. Nvidia even says that DLSS can now help deliver smooth and crisp gameplay at 8K, or even produce sharper visuals than rendering at a native resolution.
As for power efficiency, Nvidia says the same workload consumes just over half as much power as it would on an equivalent Turing-based GPU. That means less cooling is needed and less noise is generated.
Nvidia is introducing a few other new and updated standards. Ampere GPUs support PCIe 4.0 for additional bandwidth with compatible CPUs now and in the future. AMD’s Ryzen 3000 series supports PCIe 4.0 and Intel is expected to adopt it with its next generation. HDMI 2.1 allows for 8K 60Hz video through a single cable, and AVI hardware decoding allows for streams at that bandwidth in real time.
RTX IO will leverage Microsoft’s upcoming DirectStorage API for Windows 10 to allow huge game assets to be transferred directly from a fast PCIe SSD to a GPU’s VRAM without going through the CPU or system memory. This is potentially a very big deal, changing how games themselves are designed and allowing for huge environments to be streamed from storage rather than loaded and buffered. It’s similar to things we’ve been hearing about how next-gen consoles will handle storage, which makes it all the more interesting.
The GeForce RTX 3080 GPU specifically has been configured with 8,704 execution units called CUDA cores, organised into 68 “streaming multiprocessor” clusters, compared to 3,072 and 48 respectively for its direct predecessor, the GeForce RTX 2080 Super. You get 272 third-gen Tensor cores and 68 second-gen RTX cores as well. Interestingly, this GPU might reignite interest in cryptocurrency, which was once responsible for tying up the world’s entire stock of graphics cards but has died down of late thanks to diminishing returns on previous-gen hardware.
On the software side, there are new capabilities as well. Nvidia Reflex which can work with compatible monitors and mice to help optimise frame delivery and reduce input latency. AI capabilities are also harnessed by Nvidia Broadcast to swap virtual backgrounds, automatically frame streamers, and cancel out background noise in real time. Finally, Omniverse Machinima targets content creators, allowing game engines and assets to be used to easily create high-quality animations, if developers support it.
Nvidia GeForce RTX 3080 Founders Edition design
For the past few years, Nvidia has been competing with its partners and selling its own graphics cards, called “Founders Editions”. These have typically been extremely polished and stylish, but don’t bother with factory overclocking or tricked-out RGB LED effects. With this generation, Nvidia has really upped its game when it comes to design. Previous efforts truly have felt premium in a sea of plastic coolers and shrouds, but the RTX 30 series takes that to a whole new level.
The GeForce RTX 3080 Founders Edition is absolutely beautiful. It feels heavy and expensive, like a piece of finely crafted machinery, with its exposed heatsink fins and metal frame. There’s no RGB lighting here but you do get an illuminated GeForce RTX logo on the top and white accents lining the middle of the frame. It all feels simultaneously industrial yet sophisticated. The design really stands out, and that’s before we even get to the completely novel fan placement.
This is a large, chunky card. It’s longer than usual but actually isn’t taller or wider, occupying a relatively modest two slots. The main frame is a dark metal strip that’s rounded at the corners. It forms a sort of figure-8 with an elongated mid-section, with nearly the entire surface area that it encloses covered in exposed heatsink fins. This is also a very heavy card, and you can really feel the weight of the cooling system when you pick it up.
You might be surprised to see only one fan on the front, and that’s because there actually is a second one on the rear. Nvidia has come up with a completely unique twin-chamber design, in which the upper fan acts more like a traditional blower and exhausts air through the back of your case, while the lower one has its own airflow path that expels hot air upwards, to be sucked out by a case fan or radiator.
Looking closely, you can see that the actual circuit board of the RTX 3080 Founders Edition ends before the second fan, and you can look right through the fins and out the other side of the metal enclosure. Nvidia says a huge amount of engineering went into reducing board size to allow for this kind of airflow, and it was all necessary to achieve the kind of performance as well as acoustics that it targeted. The only downside we see is potential dust accumulation in the huge expanses of exposed fins.
Everything has been packed in denser and redesigned to reduce electronic interference. One of the changes required is a new 12-pin power connector which is mounted perpendicular to the board and angled forwards. Nvidia includes a short adapter with the RTX 3080 Founders Edition. You’ll need two 8-pin PCIe power connectors, and unfortunately cable routing might be a challenge, which somewhat takes away from the striking look of the card itself. It doesn’t appear that many of Nvidia’s partners have adopted the same connector, and so it might be a while before power supplies adopt it natively. Some PSU brands already have their own adapters for sale, or will send them to verified owners for free.
The rear of Nvidia’s Founders Edition card has one HDMI 2.1 and three DisplayPort 1.4a connectors. The VirtualLink Type-C port for VR headsets seen on previous-generation cards has been dropped, since the standard has failed to catch on. For this and non-overclocked third-party offerings, the reference GPU base clock is 1440MHz and the boost speed is 1710MHz. All cards will feature 10GB of GDDR6X memory running at 9500MHz on a 320-bit bus, for a total 760GBps memory bandwidth. PCIe 4.0 is supported. Total board power is rated at 320W which is quite high – previous-gen cards were rated for 225-250W.
Nvidia GeForce RTX 3080 Founders Edition performance
The test setup for our GeForce RTX 3080 Founders Edition is unchanged compared to what we’ve used to test previous generations. We’ve got an AMD Ryzen 2 2700X CPU, a Gigabyte Aorus X470 Gaming 7 Wifi motherboard, 2x8GB of G.Skill F4-3400C16D-16GSXW DDR4 RAM, a 1TB Samsung 860 Evo SSD, and a Corsair RM650 power supply. The monitor is an Asus PB287Q, which has a 4K resolution but unfortunately doesn’t support HDR or variable refresh rates. We used Windows 10 with all the latest updates and patches. Nvidia supplied a pre-release driver, version 456.16.
Starting with 3DMark as usual, we saw the GeForce RTX 3080 handily defeat the Zotac GeForce RTX 2080 Amp by 45-60 percent. The new DLSS Feature test in 3DMark indicated performance at 50.33fps with it off and 69.41fps with it on. The Port Royal ray tracing test put up a score of 10,721. We also ran the standalone Unigine Superposition test, which managed 13,708 points using the 4K Optimised preset.
AMD currently doesn’t have any offering that can post such numbers, but that might change soon with the impending launch of the Radeon RX 6000 series, based on its “big Navi” GPU. The Radeon 5700XT is in a much lower price and performance class, so we’ve left it out of this comparison table.
Nvidia GeForce RTX 3080 Founders Edition | Zotac GeForce RTX 2080 Amp | |
---|---|---|
3DMark Port Royal | 10,721 | NA |
3DMark Time Spy | 13,996 | 9,505 |
3DMark Time Spy Extreme | 7,191 | NA |
3DMark Fire Strike | 25,461 | 19,461 |
3DMark Fire Strike Extreme | 17,787 | 12,215 |
3DMark Fire Strike Ultra | 10,520 | 6,498 |
Unigine Superposition (4K Optimised) | 13,708 | NA |
Moving on to gameplay and game engine benchmarks, we decided to focus on performance at 4K, based on Nvidia’s claims and projections. Starting with Shadow of the Tomb Raider, we set it to run at 4K using the Highest quality preset and TAA enabled. For this first run, RTX ray tracing and DLSS were turned off. The in-game benchmark showed intermittent tearing with its pans through expansive scenery, but motion was otherwise crisp, and we got an average of 66fps. With RTX DLSS turned on, textures were still great, and no huge differences were immediately distinguishable. However, the average rose only to 69fps.
Middle Earth: Shadow of War also has an in-game benchmark and can be pushed pretty far. With the resolution at 4K, graphical quality set to Ultra, TAA enabled, and all other settings at Ultra or High, we got a comfortable and very playable average of 86fps.
Far Cry 5 pushed out an average of 73fps at 4K using its Ultra preset, and everything looked smooth. We took the resolution down to 2560×1440 to see the difference that would make, and got a nice bump up to 95fps. Metro: Last Light Redux managed to push out 58.79fps at 4K at Very High quality with SSAA on and 4X AF.
The Witcher 3: Wild Hunt pushed out a very consistent 85fps average with little variance, as measured by FRAPS in a manual runthrough. The resolution was set to 4K, with graphics quality at Ultra and post-processing set to High. Next up, we tried the zombie shooter World War Z at 4K using its Ultra preset. The built-in benchmark ran at a super-smooth 110fps without breaking a sweat.
Assassin’s Creed Odyssey is fairly recent and demanding. We measured 47fps at 4K with the graphics quality set to Ultra High. Taking that down a notch to Very High pushed our score up to 53fps. A lower resolution will be necessary to run this game at a smooth 60+ fps.
To put the GeForce RTX 3080’s ray tracing and DLSS 2.0 capabilities to the test, we fired up Control. The resolution was set to 4K and display quality to High. Starting with both features disabled, we noted an average of 77fps in a specific action sequence using Nvidia’s own overlay. Turning Ray Tracing on and choosing the High quality setting immediately pulled our average down to about 47fps. DLSS should compensate for this, and you can now choose your rendering and target resolutions independently. With the latter set to 2560×1440, we actually noted a jump up to 80fps. In terms of quality, things are looking good for DLSS – everything was sharp and we had no complaints about image quality. This is the experience we had in just one game, and that too on a very high-end GPU, but it does show that there are more ways to improve quality than just brute force rendering.
Just for fun we fired up Quake II with Nvidia’s RTX renderer. This ancient title really shouldn’t look so good, but it’s capable of using ray tracing for lighting effects which makes it feel somewhat fresh. We pushed the resolution all the way up to 4K and managed an average of 42fps.
All the engineering that went into the GeForce RTX 3080 Founders Edition card seems to have paid off – it hardly makes any sound, even under very heavy load. The “dustbuster” designs of the past are long behind us – you’ll only hear a dull whoosh when running heavy games or tests. However this card does push out a lot of hot air, and does so quite forcefully. On our open test bench, we were able to feel it even two feet away from the rear vents. The metal frame also got too hot to touch for more than a few seconds. This card will likely run hotter and louder within a small, cramped desktop enclosure, and it will be interesting to see how more traditional triple-fan blowers on partner cards fare.
Verdict
The Ampere architecture seems to be yet another feather in Nvidia’s cap, continuing a long string of successes and putting some severe strain on AMD. Of course we mustn’t forget that Intel is also coming up with an enthusiast-class discrete GPU early next year, so things are about to get very, very interesting for PC gamers. The GeForce RTX 3080 delivers excellent performance in today’s games – it’s more than enough for 1440p and you should be able to push smooth 4K visuals without much trouble.
The GeForce RTX 3080 Founders Edition is a phenomenal piece of hardware in terms of design. This card and the GPU it’s based around are however not without some flaws – heat is likely to be an issue, power consumption is quite high, and 10GB of VRAM isn’t as much as many people were expecting this generation to have. Performance is excellent, and much better than what you could get for this price a year or two ago.
All of this means that while Rs. 71,000 is not a bad price at all for the performance that you get, it’s still way out of reach for the vast majority of gamers and enthusiasts in India. Excellent performance at 4K is tantalising, but this is still a very niche target resolution here, and we know that people would much rather see how Nvidia will use its technology to serve the lower end of the market. The GeForce RTX 30 series models announced so far are also still too expensive to take ray tracing and DLSS mainstream.
There’s no telling yet what Nvidia has in store for lower-priced models — we don’t know whether the GeForce GTX series will continue to serve budget-conscious buyers or if there will now be a full spread of RTX models covering all tiers. We’ll have to wait at least a few months for the inevitable GeForce RTX 3060 and whatever else follows.
Nvidia GeForce RTX 3080 Founders Edition
Price: Rs. 71,000
Pros
- Excellent performance at 1440p and 4K
- DLSS is useful. insupported games
- Looks incredible
Cons
- Runs quite hot
- Relatively high power requirement
Ratings (out of 5)
- Performance: 4.5
- Value for Money: 4
- Overall: 4.5