A review like this doesn't come along often. Sure, we review graphics cards and their graphics processing units (GPUs) several times a year. But this time around, we are reviewing not only an all-new Nvidia flagship graphics card, namely the Nvidia GeForce GTX 1080 (Founders Edition) ($699), but one that's based on an entirely new architecture, as well, involving a shrink in the manufacturing process used to make the chips.
This only happens once in a while. In fact, the last time we saw this was way back in 2012 with the Nvidia GeForce GTX 680, in parallel with Nvidia's launch of its "Kepler" architecture. We're not here to talk about Kepler, though. We're here to discuss the new kid in town. That's Nvidia's brand-spanking-new chip architecture, named "Pascal."
The GeForce GTX 1080 is the first Pascal-based GPU aimed at gamers. It's built on a 16-nanometer (nm) process using "3D" FinFET transistors, and according to Nvidia it's "the new king" of the GPU world. That's certainly a bold claim, but as our benchmarks show, it is, indeed, the fastest gaming GPU we've ever tested.
Similar Products
Nvidia GeForce GTX 1070 (Founders Edition)
Nvidia GeForce GTX 980
MSI GTX 980 Gaming 4G
EVGA GTX Titan Black
The king-making claim also isn't entirely surprising, since every time a new GPU with a new architecture launches, it's generally faster than what preceded it. But this time around it's almost shocking just how fast it is, and how big the leap. You'll have to see the benchmark tests we ran to believe it...so let's get started.
How We Got Here
Before we dive into Pascal, let's take a brief walk down memory lane to understand how we arrived at this juncture in GPU history.
Back in 2012, Nvidia launched its "Kepler" architecture, which marked the company's leap from a 40nm silicon fabrication process to a much smaller 28nm process. (This figure in nanometers is essentially the size of the transistors used in the graphics-processor silicon.) When this number goes down, it's a big deal, and this event is known as a "die shrink." As the size of the transistors gets smaller, thanks to advancements in manufacturing technology, it allows Nvidia (or its competitor AMD) to pack a lot more transistors into the same size die it was using previously. That allows it to improve performance without being lazy and just creating a bigger silicon die.
Once Kepler had finished its run in the marketplace, everyone expected Nvidia to do what it had always done in the past, which is to shrink its fabrication process again. (Nvidia doesn't actually make the silicon itself, but just stick with us here.) That expected drop would have been from 28nm down to 20nm, and the expectation was that the company would announce a brand-new GPU using the new, smaller process. That didn't happen, and an official explanation was never given as to why. But instead of shrinking its process, Nvidia soldiered on with the same 28nm die it used to make Kepler cards for its successor series, which was named "Maxwell."
Interestingly, competitor AMD did exactly the same thing, continuing on with its 28nm GPUs for a second go-around, too. So there must have been diminishing returns at 20nm that made it a fool's errand to pursue. Or perhaps there was a problem with the silicon manufacturers themselves (TSMC, in the case of this card) being able to keep up with demand, alongside the push for smartphone chips and other in-demand devices. Though Nvidia's Maxwell architecture was indeed a lot faster and more efficient than Kepler, it was due to savvy engineering and the maturity of the 28nm process. Maxwell was excellent for its time, but it never quite achieved the same breakthroughs in performance that usually accompanies a die shrink.
Now that Maxwell has finished its run, Nvidia has gone ahead and moved to a new fabrication process, this time going all the way down to 16nm. As a two-generation leap from 28nm, 16nm is allowing Nvidia to make radical improvements in both performance and efficiency. The first 16nm gaming card to hit the market is the GeForce GTX 1080 we are reviewing today. It's the first card to land in our test bed that was made using the all-new process and Nvidia's "Pascal" architecture.
Inside Pascal
Nvidia's Pascal GPU not only uses a smaller 16nm process than Maxwell, but it also uses a newer type of transistor, named FinFET. With FinFET, transistors get stacked in 3D space, as opposed to laying them down side-by-side. Just like the 3D NAND we've seen in Samsung SSDs, such as the Samsung SSD 850 Pro, this allows Nvidia to pack more transistors into a smaller area than it ever has before. And more transistors in the same space typically equals better performance.
Just how many more transistors? In the case of the GeForce GTX 1080, it's two billion more...yes, with a "B." The previous-generation equivalent to the GeForce GTX 1080 card, the GeForce GTX 980, sported 5.2 billion transistors. The GeForce GTX 1080 has 7.2 billion, even though the size of the die remains unchanged.
What's amazing is that despite having that many more transistors, the thermal design power (TDP) rating for the GeForce GTX 1080 is only 15 watts higher than that of the GeForce GTX 980. (It's rated as consuming 180 watts, compared to the GeForce GTX 980's 165 watts.) What's even more incredible is that the GeForce GTX 1080 should be remarkably more powerful than the GeForce GTX 980, with very little "cost" in terms of the power required.
Further proof of Pascal's energy efficiency is the fact that the GeForce GTX 1080 requires just one eight-pin PCI Express power connector off your system's power supply, as opposed to the dual six-pin connectors that the GeForce GTX 980 demands. This is the first flagship GPU we've ever seen that uses only one power connector.
Another benefit of Nvidia's transition from 28nm transistors to the new 16nm FinFET process is a dramatic increase in possible clock speeds. The Kepler and Maxwell cards' clock speeds typically hovered around 1GHz or so, with the occasional chip able to hit 1.3GHz or 1.4GHz. But with the GeForce GTX 1080, the base—mind you, base—clock speed is a staggering 1.6GHz, with a boost clock able to hit 1.73GHz under certain conditions. Nvidia is always a bit conservative with citing the ceiling of these baseline clocks, too, so we can expect the cards to go much higher than this. At the launch of the card Nvidia said squeezing an additional 300MHz or so out of a GTX 1080 would be no big deal, and it showed a card running at 2.1GHz on the stock cooler as well. We weren't quite able to eke out such high stable clock speeds with our review card, but getting to 2GHz wasn't much of a problem. (More on that later in the Performance section, below.)
One "feature" that's also new with Pascal is the introduction of a "Founders Edition" version of the card, which is replacing the term "reference card" in the video-card vernacular. Nvidia's typical practice was to create a reference design of a new GPU to send to its card-making partners so that they could use it to build their own version of the cards, typically with fancier coolers than the stock versions and higher clock speeds. Not so anymore, as the version Nvidia builds, and which will now sell directly to consumers for the life of the product, is a premium edition, not a sample baseline model. The catch is that the Founders Edition of the GeForce GTX 1080 costs $100 more than the "standard" version, which will sell for $599. That means the Founders Edition, which we tested, costs $699.
We're not sure how the GeForce GTX 1080-based cards from partners such as EVGA, Asus, Gigabyte, and MSI will differ from the GTX 1080 Founders Edition. We hadn't seen any of them as of press time. But Nvidia took pains during the course of its product announcement to let everyone know it had spent considerable engineering resources designing its GPU. Not only is it highly overclockable, the company told us, but it runs cool and quiet, and it has premium components throughout, including vapor-chamber cooling and even a backplate, which is rare among Nvidia-designed cards.
The GTX 1080 also features a blower-style design that exhausts heat out of the chassis, which is favored by Nvidia in its stock-card GPU designs. Most, if not all, of Nvidia's partners typically offer card designs with coolers that exhaust heat inside the case as opposed to expelling it from the rear edge. So there is a benefit to Nvidia's blower design, especially for those using it in a cramped or small-form-factor chassis. The blower design is also useful when you've got several cards installed in an SLI arrangement, as one card doesn't transfer the bulk of its heat to the adjacent card by blowing hot air straight on it.
The New Shape of SLI
Speaking of SLI, Nvidia is also introducing a new connector with the GeForce GTX 1080 that doubles the bandwidth available, compared to the previous SLI solution. It's named the "SLI HB Bridge," the "HB" being for "High Bandwidth," and uses a two-lane configuration, as opposed to the single lane used by the previous connector.
The downside to this approach is that the GeForce GTX 1080 supports only two cards in SLI, as it uses both connectors on the card. In the past, the first connector was used to go to the first pair of cards, and the second connector was used to go to another card in the chain. You could keep going all the way down the chain with up to four GPUs. Nvidia states that three and four-way SLI is theoretically still possible using DirectX 12 features if developers include them, but it's "not recommended."
Also, triple and quad SLI are no longer possible with cards straight out of the box. Nvidia says it will be setting up a site by the time of the May 27 launch of the GTX 1080 to allow dedicated SLI hounds to "unlock" the three-way and four-way SLI feature of the GTX 1080 card. It will involve running an app to identify your card and trigger a request for a software "Enthusiast Key" from Nvidia, which will allow for SLI beyond two cards.
This change in SLI policy isn't that big of a deal, really, as we imagine the number of people running four top-end video cards could nearly fit inside our company's lunchroom. Besides, scaling drops considerably beyond two cards in SLI, and you also run into the issue of the performance being bottle-necked by the CPU.
It's also possible to run two cards in SLI and use a third just for PhysX, which is how Nvidia had PCs configured at the GTX 1080's press event.
Specs and Design
Let's take a look at the card itself first, then we'll get into some new technologies that Nvidia announced at the same time as the GeForce GTX 1080. Some of them are exclusive to this new GPU.
Like all previous high-end GeForce cards designed by Nvidia, the GeForce GTX 1080 is a dual-slot card. (A few third-party "beast" GTX 980 Ti cards, among them the blistering Zotac GeForce GTX 980 Ti Amp Extreme, take up three slots.) The GTX 1080 is 10.5 inches in length, and is exactly the same height, width, and length as the GeForce GTX 980 (and in the same ballpark as other high-end Nvidia GPUs). So nothing is new in the physical-layout department. Nvidia is supposed to be switching from GDDR memory to Generation 2 of its High Bandwidth Memory (HBM) next year, possibly with an ostensible consumer version of its "Big Pascal" card, which will allow the memory to move from next to the GPU to directly on the die. This will free up a lot of space on the PCB, and allow for the GPUs to be a lot shorter, like we saw with AMD's Radeon R9 Fury. But for now, it's business as usual.
The biggest change we can see in the GeForce GTX 1080 compared to all the prior generations is the lone eight-pin PCI Express power connector. That is a surprise, unheard of on a late-model flagship video card. Every recent GPU we've reviewed from Nvidia, aside from mainstream and entry-level ones, has required two PCIe connectors. So the fact that this is a flagship GPU reportedly "faster than a Titan X" and with just one power connector shows you the gains in efficiency the move to 16nm FinFET has produced (along with a little engineering by Nvidia, of course).
Out back, the card sports five outputs, including three DisplayPort connectors that are certified for DisplayPort 1.2 but "ready" for the 1.3 and 1.4 upgrades, according to Nvidia. Though DisplayPort 1.2 is sufficient to run a 4K (3,840 x 2,160) monitor at 60Hz, DisplayPort 1.3 ups the bandwidth on the channel enough to support a 4K monitor at 120Hz, and version 1.4 ups it even more, to handle nascent 8K at 60Hz.
Out of the box, the GeForce GTX 1080's clock speed is an incredibly high 1,600MHz, with a boost clock of 1,733MHz. This is about 700MHz higher than what we're used to seeing on a high-end Nvidia GPU, so it's a massive jump, and that doesn't even include overclocking numbers.
The card is designed to run cool and quiet too, just like the previous generation. It sports 8GB of super-fast Micron-made GDDR5X memory, which is a first in the GPU world. The surprisingly low TDP of 180 watts is 70 watts below the GeForce GTX 980 Ti, GeForce GTX Titan X, and previous flagship GPUs.
New Technologies
Let's take a brief look, in turn, at some of the new technologies that the GTX 1080 is bringing along with it.
Ansel
Ansel is designed to allow for more creative control when taking in-game screenshots, which if, you're like us, probably wasn't something you considered needed improving, or thought much about. But it's a thing.
Dedicated gamers have been taking very creative screenshots for a while now, but they are, of course, limited by where the camera can go, and the resolution of the images. Ansel solves both of these problems by allowing for a free-ranging camera in any game that supports it, and by letting you capture massive high-res screenshots.
In a demo Nvidia ran for the press before the launch of the GTX 1080, it captured a scene at 20x resolution, and the resulting file was 3GB in size and 46,000 pixels across, or "46K" to use the popular nomenclature. The Ansel software, named after famed photographer Ansel Adams, also lets you apply filters to your photos, in semi-Instagram fashion. You can also rotate the horizon and make other changes.
Ansel will be a feature on games that decide to include it in their feature set, so it won't be something that is available in every game by default. You activate it with a key combination, which pauses the game and causes an overlay to appear that allows you to move the camera, make adjustments, and ultimately capture the frame. Nvidia says this feature is supported on Pascal and Maxwell cards.
Simultaneous Multi-Projection (SMP)
People who use three monitors, or who are into virtual reality (VR), will be excited about this technology, as it could mark a big leap forward for both of these usage cases. What SMP does is allow the GPU to project into 16 "viewports" simultaneously and in stereo. What this means for the VR world is a massive increase in rendering speed, as the previous generation of cards had to render each eye in sequence.
The GTX 1080 can do both displays in one pass, however, which is why you heard things from Nvidia about how it's "twice as fast as a Titan X..." and then off to the side of the PowerPoint presentation it said "...in VR." It will also let you game on several monitors at once, and if you run three monitors with the side monitors angled towards you, it's able to reduce the distortion that occurs on objects and more accurately project one image across all three monitors. You can watch the demo here...
...and as it states, the original projection is correct if all three monitors are side-by-side. It's when you pull the side ones toward you that things get wonky. SMP fixes this issue, and it looks great too.
Fast Sync
We all know what familiar old V-Sync is: It syncs the frame-rate output of the GPU with the refresh rate of the monitor (typically 60Hz or 60 frames per second) or an even divisor of it. This produces gaming that is free from tearing, but it locks the top frame rate at 60fps, which is not ideal for a lot of e-sports competitors (apparently, we're not them).
The solution, then, is to turn off V-Sync, which lets the GPU run at full speed. But when you are running a really high frame rate, you can experience latency, which is also bad for e-sports. To fix both of these problems, Nvidia has developed a new syncing mode named "Fast Sync," and it's only advisable to use it in scenarios with extremely high frame rates.
About Our Tests
Things are a bit in flux these days when it comes to testing GPUs, as there are two emerging technologies that this card was built for that are difficult to test. The first is DirectX 12 (DX12), which is just now coming on the scene. There are very few real-world benchmarks for it. Still, DX12 will likely be the standard graphics API in the future, and this card was designed to last for a few years, if not longer. So it is important to know if a card can handle DX12 well before buying.
We tested the GTX 1080 with all the newest DX12-capable games we had on hand, including Hitman (the 2016 edition), Rise of the Tomb Raider, and Ashes of the Singularity. We tested a load of games using DirectX 11, too, because that API will still be in wide use for at least another year, and probably much longer.
The second technology that's difficult to test at present is virtual reality, or VR for short. The GeForce GTX 1080 was built to run VR twice as fast as its predecessor, and in all the launch presentation documents Nvidia specifically referred to the card's VR performance, as that was what the company wanted to highlight. However, there are two major competing VR technologies, in the form of the Oculus Rift and the HTC Vive, with more coming to market soon, and it's difficult to establish a lone test that is applicable to all scenarios.
Steam has its own VR benchmark, but at the time of this writing, it didn't output a score. Instead, it just indicated whether or not your PC was ready to handle games using an HTC Vive. Since the baseline recommendation for both the Vive and Oculus is a Core i5 processor and a GTX 970 graphics card, the GTX 1080 and the Core i7 CPU in our test bed would easily pass this test.
Futuremark is also working on an upcoming VRMark test, but it was only in beta when we wrote this, and it refused to run on our test bed, even after making sure the benchmark was up to date. We'll have to wait for future, finalized VR benchmarks. But if you're considering buying a GTX 1080 primarily for VR, you can rest assured that current VR-ready games and those launching in the near future will run on this card just fine. It well exceeds the minimum recommendations.
3DMark (Fire Strike)
We started off our testing with Futuremark's 2013 version of 3DMark, specifically the suite's Fire Strike subtest. Fire Strike is a synthetic test designed to measure overall gaming performance, and Futuremark has expanded Fire Strike nowadays into three subtests. In the past, we used the basic test (known simply as "Fire Strike"), as well as the more demanding Fire Strike Extreme test. But these GPUs are so powerful that we had to move up to the most punishing test, Fire Strike Ultra, which is geared toward simulating the stresses of gaming at 4K.
As the chart shows, the GeForce GTX 1080 dominated right out of the gate, besting every card that stepped up to it, including the former champ, the GeForce GTX 980 Ti. It's quite remarkable when you consider that compared to the GeForce GTX 980 Ti, the GeForce GTX 1080 is using 70 watts less power, has fewer CUDA cores, and uses a narrower memory bus (though faster GDDR5X memory). This just goes to show the benefits that moving to a smaller process, refining the architecture, and enabling insane clock speeds can provide.
Tomb Raider (2013)
Here, we fired up the 2013 reboot of the classic title Tomb Raider, testing at Ultimate detail and three resolutions.
In our first "real world" test, we can see the GTX 1080 is going to be a force to be reckoned with, as it was more than 60 percent faster than the GeForce GTX 980 GPU that it replaces, which is an incredible leap in performance from one generation to the next.
In the past, we usually saw jumps in the neighborhood of 20 to 25 percent from, say, a GeForce GTX 680 to a GeForce GTX 780, so this is one heck of an increase. We also see in our tests at 3,840x2,160 that the GeForce GTX 1080 will likely be the first single GPU capable of hitting 60 frames per second at 4K, which was not possible at all with the previous generation of GPUs.
Sleeping Dogs
Next, we rolled out the very demanding real-world gaming benchmark test built into the title Sleeping Dogs...
In Sleeping Dogs, we once again see the GeForce GTX 1080 towering over the competition, improving on the performance laid down by the GTX 980 by 60 percent once again. It was a smidge slower than the AMD Radeon R9 Fury X in this test at 4K, however, but by only 2fps, which is close enough to call it essentially a tie. The Radeon R9 Fury X has always been a beast at high resolutions due to its ultra-fast HBM memory, but the GTX 1080 roughly matches it at 4K here and, at lower resolutions, the R9 Fury X and other cards in this group are simply no threat.
Bioshock Infinite
The popular title Bioshock Infinite isn't overly demanding, as recent games go, but it's a popular one with stellar good looks. In its built-in benchmark program, we set the graphics level to the highest preset (Ultra+DDOF)...
This was a hard-fought battle, as the results in this test were closer than other tests, most likely due to the relatively undemanding nature of this game. Not surprisingly, the GeForce GTX 1080 still came out on top, and at 1440p performed around 50 percent faster than its predecessor, which is quite impressive. It was only 10fps faster at 4K than the GeForce GTX 980 Ti, but that's still a 20 percent advantage. The GTX 1080 was also able to hit 69fps at 4K in this benchmark, allowing for silky-smooth super-high-resolution gaming.
Hitman: Absolution
Next up was Hitman: Absolution, which is an aging game but still pretty hard on a video card.
By now, it's no surprise that the GeForce GTX 1080 topped the charts. The only question is by how much? The test that blew our minds this time around is the 1440p run, where the GTX 1080 ran more than 80 percent faster than the GeForce GTX 980. This is the kind of performance leap we haven't seen in several years, certainly not from a single generation to the next.
Far Cry Primal (Ultra)
Ubisoft's latest open-world first-person hunting game is one of the most demanding titles we use, thanks to its lush foliage, detailed shadows, and otherwise incredible environments.
This test very clearly illustrates the benefits of upgrading from a GeForce GTX 980 to a GTX 1080, as we see that 4K gaming was not really playable at top settings on the GeForce GTX 980 at 25fps, but is much more acceptable (at 41fps) with the GTX 1080. That's a big improvement, and one that can justify an upgrade, as opposed to gaining "just" a 20 percent or so improvement in certain titles.
Like the previous titles, we saw the GTX 1080 post top scores across the board, and while its performance of 41fps at 4K is a lot better than 25fps, as we previously noted, this falls short of the 60fps required for gameplay that's silky-smooth. Perhaps an eventual GeForce GTX 1080 Ti will be able to handle it.
Ashes of the Singularity
Oxide's Ashes of the Singularity is a bit of a departure as a benchmark, as it's a real-time strategy title, rather than a first-person shooter or a third-person action title. Due to the planet-scale nature of its battle scenes, with hundreds of onscreen tanks, ships, and other implements of future warfare, it can be extremely demanding at high settings. And because of the plethora of rendered units, this game is also more CPU-bound—especially at high settings and resolutions—than most other recent games.
Due to the CPU-heavy nature of this title, the scores here are much closer than they were anywhere else in our testing. Still, the GTX 1080 still dominates here, edging out the previous-generation GTX 980 Ti. At 4K resolutions, the new card looks its best, adding several frames per second over what any other card offered up.
Grand Theft Auto V
As one of the most popular game franchises on the planet, Grand Theft Auto really needs no introduction. The fifth installment took a lot longer than many expected to land on the PC. But when it finally did in early 2015, it brought with it a number of graphical improvements and tweakable visual settings that pushed the game far beyond its console roots.
Once again, the GTX 1080 dominated here, easily besting the competition from AMD and Nvidia at all resolutions. Most impressively, it tacked on a full 15fps over the Radeon R9 Fury X and GTX 980 Ti at the highest resolution (4K).
Rise of the Tomb Raider
Lara Croft rises once again in the early 2016 iteration of Square Enix's long-running action franchise. As our hero works to unfold an ancient mystery (and reveal the secret to immortality) ahead of the ancient and deadly Order of Trinity, she traipses through a slew of complex atmospheric environments, from arid tombs to the frigid Siberian wilderness. A dynamic weather system, and the complexities of Lara's wind-tousled hair, add to the game's visual complexity.
The performance trend continued here, with the GTX 1080 particularly mopping the floor with AMD's current high-end R9 Fury X. But even compared to the GTX 980 Ti, the GTX 1080 added nearly 10fps at the highest resolution, making for smooth game play on today's high-end monitors.
Hitman (2016, Under DX11) The newest game in the Hitman series finds Agent 47 turning over a new leaf, and embarking on a journey of self-discovery as a teacher at a school for underprivileged children. Just kidding; he kills a bunch of people in this one, just like the rest. It does offer gorgeous graphics in both DX11 and DX12 varieties, though. We'll tackle the former first.
The AMD R9 Fury X managed to best the GTX 1080 for the first time here, at all resolution settings. As this is the first time we've actually used this benchmark, we're entirely not sure why. But AMD's High-Bandwidth Memory might have something to do with it. Regardless, at 4K, the GTX 1080 is close enough to nearly call it a tie. And Nvidia's flagship easily bests its previous-generation cards on this test.
It's tough to get any real sense of DirectX 12 performance at this point. When we wrote this in mid-May 2016, only a few titles were available with DirectX 12 support. And running these games, anecdotally we saw no graphical differences between the titles running at DX11 versus DX12 settings. In some instances, titles running under DX12 offered performance gains, but elsewhere we saw lesser performance. Also, under DX12, the 2016 Hitman title locked up more than once in our testing. And we saw a few instances in Rise of the Tomb Raider in which large chunks of the world failed to render at all. We noted missing trees and chunks of wall more than once in testing.
In other words, you should take the below results with a Gibraltar-size hunk of salt. DirectX 12 is still in its extremely early stages, and those developers who have implemented it have yet to throw on the spackle and smooth over the cracks. We'll have to wait at least a few more months to say for sure how much of an advantage DX12 offers, and whether it sways things in favor of AMD or Nvidia in any substantive way. Still, because this is a cutting-edge card and DX12 is cutting-edge tech, it's worth taking a look at what the GTX 1080 and its competition can do with Microsoft's latest gaming API today.
Rise of the Tomb Raider (Under DX12)
This sequel to 2013's Tomb Raider is one of the first AAA titles to offer DirectX 12 support. We used the preset labeled Very High for testing.
It's notable that the only GPU that could run this game at 4K with a playable frame rate and some overhead was the GTX 1080. The GTX 980 Ti was playable at that setting as well, but only just. This game simply crushed the other GPUs at super high-res. The GTX 1080 also posted an astonishing 84fps at 1440p, which is almost double what the GTX 980 was capable of, and four times what the Radeon R9 Fury X turned in at the same setting here.
Hitman (2016, Under DX12)
The newest Hitman title also offers up a DX12 graphics option in its benchmark which, like Rise of the Tomb Raider, looked identical to our eyes to the DX11 version. Or at least it did when it wasn't locking up, which never happened under DX11.
Though the GTX 1080 topped the charts once again (no surprise there), it did so by the smallest margins we've seen yet, and the AMD R9 Fury X, which was stomped in our previous DX12 test, was extremely competitive here.
Ashes of the Singularity (Under DX12)
The strategy title Ashes of the Singularity was among the first to offer DirectX 12 support, even when it was still in beta. Perhaps unsurprisingly, then, it was also the most stable DX12 test we ran, never once crashing, locking up, or noticeably glitching when we ran it multiple times on a handful of high-end cards.
Here again, the GTX 1080 was the leader, but again the AMD R9 Fury X wasn't far off. At least most of the cards here posted solid gains in frame rates over the DX11 version of the test. DX12 performance is still rather unsettled turf, and it remains to be seen what type of gains we'll see when several developers implement a wide aspect of features, and which cards will respond best to the new API.
Overclocking
We didn't have as much time as we'd like to test overclocking. But using a beta version of EVGA's Precision X utility, we were able to push our review card's clock speed up just above 2GHz, which is a threshold unheard of until this card, and a healthy 300MHz or so above the 1,733MHz top Boost Clock speed of the card.
That translated to a 3fps bump in FarCry Primal at the Ultra setting and 4K resolution, and a 4fps boost to Hitman under DirectX 11 at 4K and high settings. Perhaps with more time, we could have eked out a higher stable overclcock, but considering the card comes clocked much, much higher out of the box than any previous card, the fact that it could be pushed even higher is impressive. And of course, overclockability often varies from card to card, so your pixel boost may vary.
It's also worth pointing out that it was only when overclocking, with the side popped off of our test bench, that we really noticed any fan noise on the GTX 1080 at all. Running at stock speeds, the card was remarkably quiet, even under load. The fan was spinning, of course, so it wasn't completely silent. But the card wasn't noticeably noisier than the 120mm fan on our Noctua CPU cooler, which is impressive for any high-end card. Considering the GTX 1080 is far and away the most powerful single-chip card we've tested (and it's not liquid cooled like the AMD R9 Fury X), its silent performance is doubly impressive.
The Fastest Graphics Card
A lot of times when writing conclusions to reviews of video cards like this, we have to resort to qualifiers such as "It's a great card if..." or "It's really fast, but you might want to wait..." or even "It's the best card at this price, but there's a card that is less expensive that is almost as good." That is not the case this time around.
The GeForce GTX 1080 is unquestionably the fastest single GPU in existence at this time, and by a wide margin, too. Not only does it offer as much as 80 percent more performance than a GeForce GTX 980, but it also spanks the GeForce GTX 980 Ti by about 20 to 30 percent, and offers new technology that neither card supports. We're talking, of course, about the Simultaneous Multi Projection, new VR capabilities, and high-bandwidth SLI, to name a few. If you're looking for the fastest GPU available, there is no question whatsoever it's the GTX 1080, as it also pummeled AMD's $650 Radeon R9 Fury X in nearly every benchmark we ran.
In fact, it seems like the only upcoming GPU that might suppress sales of the GeForce GTX 1080 is the also-announced GeForce GTX 1070, which is a GPU we know very little about at this time. We know it's going on sale on June 10, and that it will offer less performance than the GTX 1080 at a price of $379. (A Founders Edition version will be $449.) That's a much more compelling price point for the vast majority of gamers out there. It remains to be seen what kind of performance the GTX 1070 will offer, but we're certain a lot of gamers are holding off on pulling the trigger until that launch occurs.
As far as AMD goes, it's all silent on the Western Front, as they say. We know the company is planning on unveiling some details about its upcoming Polaris architecture at Computex 2016, but those GPUs will reportedly be targeted at mobile and "mainstream," so it's doubtful any of them near-term will be able to lay a glove on the high-end Pascal GTX 1080 or GTX 1070. AMD has spoken of an upcoming "enthusiast" GPU, code-named "Vega," but it's rumored to be arriving later this year, giving Nvidia free reign to dominate the upper echelon of the gaming market for the summer of 2016, if not longer. Since AMD has also moved to a much smaller 14nm FinFET process for Polaris and Vega, hopes are high that the company can challenge Nvidia just as capably as it did with its Fiji and Hawaii GPUs. But only time will tell. The longer the lead AMD lets Nvidia have with its impressive new high-end card, the more market share it will likely lose, at least to well-heeled enthusiasts.
The only other consideration at this point in time is the fact that there's still the looming threat of a GPU based on "Big Pascal," which would be the full-size die of this architecture. You might not have realized it, but the GTX 1080 is actually a cut-down version of the full-size chip, which is the norm for Nvidia launches. The company has previously launched a smaller version of its die first, then later on out a more powerful full-size chip, which has been in the past the GTX Titan variant. In this case it might be named the GeForce GTX 1080 Ti or perhaps GeForce GTX 1180. But again, there's no telling when (or if) that card will arrive in a form aimed at gamers.
Our only real reservation with the GeForce GTX 1080 is the $100 "Founders Edition" price premium that Nvidia is charging with this version of the card. It's hefty, on a card that's already expensive, and it's unclear when cards will arrive at the promised $599 MSRP. We would urge the frugal and patient among you to wait and see. But with high-end cards like this, there's often limited stock and pent-up demand (even when there isn't nearly this much of a performance boost), which drives up prices. That, combined with the fact that AMD doesn't seem ready to launch a rival to the GTX 1080 anytime soon, means it may be months before we see non-Founders Edition versions of GeForce GTX 1080 cards selling at $599. Board partners like EVGA and Asus will certainly have no reason to drop the prices that low if demand remains high.
So, if you want a high-end card for 4K gaming and VR future-proofing, and can afford the hefty outlay, there's little reason not to plunk your credit card down. No other single card is likely to come close to the GTX 1080 soon, and we don't expect prices to drop much quickly. The GTX 1070 will almost unquestionably be a better value for large numbers of shoppers. But, then again, gamers playing in the high-end video-card market, who want the absolute highest frame rates possible, never seemed to care a whole lot about that. We don't expect them, with all that the GTX 1080 brings, to start now.
The GTX 1080 is far and away the fastest single GPU we have ever tested, and is the first GPU that allows for true 60fps 4K gaming (and improved VR, too). It's a game-changing GPU in every sense.
Like What You're Reading?
Sign up for Lab Report to get the latest reviews and top product advice delivered right to your inbox.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.
Thanks for signing up!
Your subscription has been confirmed. Keep an eye on your inbox!
Sign up for other newsletters
