Sometimes at TR, we do funny things for major product launches. We’ve asked chips to review themselves. We’ve had the NSA monitor our testing labs. With the advent of the GeForce GTX 1080 Ti, Nvidia’s fastest consumer graphics card ever, I felt like it was time to continue the tradition.
Once I got this card on the bench, though, I dropped that idea, because the GTX 1080 Ti is no joke. I’m also not that funny.
If you’re not already familar, Nvidia tends to launch its biggest graphics chip per generation aboard a Titan-branded card that sticks around for a while, after which it tends to send some folks in with the world’s tiniest chainsaws to produce a slightly cut-down yet similar-performing product it can sell for less money. When the Titan X Pascal launched in July of last year, we knew that a GTX 1080 Ti would likely follow. It just took a while.
In the case of the $700 GTX 1080 Ti, the tiny-chainsaw-wielders didn’t have much work to do with the GP102 graphics chip that’s shared by the Titan X Pascal. The GTX 1080 Ti comes with the same 3584 stream processors enabled as its $1200 forebear. To justify the $500 price difference, Nvidia buzzed off eight ROPs and narrowed the memory bus width to 352 bits, resulting in an unusual 11GB pool of GDDR5X RAM compared to the Titan X Pascal’s 12GB on a 384-bit bus. At the same time, the green team bumped up the boost clock 51 MHz and filled out that 11GB of RAM with new 11 GT/s memory. That means in some regards, the GTX 1080 Ti is actually faster than the Titan. Maybe some people will really want that black cooler.
|GTX 970||1050||1178||56||104||1664||224+32||7 GT/s||3.5+0.5GB||145W||$329.99|
|GTX 980||1126||1216||64||128||2048||256||7 GT/s||4 GB||165W||$499.99|
|GTX 980 Ti||1002||1075||96||176||2816||384||7 GT/s||6 GB||250W||$649.99|
|Titan X||1002||1075||96||192||3072||384||7 GT/s||12 GB||250W||$999.99|
|GTX 1080||1607||1733||64||160||2560||256||10 GT/s||8GB||180W||$499.99|
|GTX 1080 Ti||1480||1582||88||224||3584||352||11 GT/s||11GB||250W||$699.99|
|Titan X Pascal||1417||1531||96||224||3584||384||10 GT/s||12GB||250W||$1200.00|
Even folks who prefer black may want to go with the GTX 1080 Ti anyway, because this Founders Edition card has an improved design compared to the first FE coolers.
Most importantly, the DVI output is no more, so the 1080 Ti’s blower has more vent area to exhaust the heat the cooler wicks away from the 471 mm² GPU underneath. Those who still need a DVI connector will find an DisplayPort-to-DVI adapter in the 1080 Ti FE’s box. Even if you have to burn a DisplayPort this way, the 1080 Ti FE still offers two more DisplayPort 1.3 outs and an HDMI connector.
The cooler keeps the same 63-mm blower fan and a vapor-chamber heatsink similar to the ones we know from earlier Nvidia reference designs. We didn’t have time to strip down the 1080 Ti FE for forensic purposes, but the new card weighs 25g more than the GTX 1080 FE. We’re guessing not that much is different underneath the shroud.
The GTX 1080 Ti reference PCB. Source: Nvidia
At a board level, though, there are most definitely differences. With a 250W TDP, the Founders Edition card needs both six-pin and eight-pin PCIe power connectors to operate. Nvidia also says it’s beefed up the power-delivery subsystem of the 1080 Ti with a seven-phase “dual FET” setup capable of delivering 250A of power. This setup purportedly delivers cleaner power with less waste heat. Compare that to the five-phase design of the GTX 1080 FE. We’ll see how this setup translates to overclocking prowess a little later on.
Our testing methods
Most of the numbers you’ll see on the following pages were captured with PresentMon, a software utility that uses data from the Event Timers for Windows API to tell us when critical events happen in the graphics pipeline. Namely, we’re interested in the time between present() calls, which correlate with the frame times that we used to collect using Fraps.
As ever, we did our best to deliver clean benchmark numbers. Our test systems were configured like so:
|Motherboard||Gigabyte Aorus GA-Z270X-Gaming 8|
|Memory size||16GB (2 DIMMs)|
|Memory type||G.Skill Trident Z
|Memory timings||18-19-19-39 2T|
|Hard drive||2x Kingston HyperX 480GB
Corsair Neutron XT 480GB
|Power supply||Corsair RMx 850W|
|OS||Windows 10 Pro|
|Driver revision||GPU base
|Radeon R9 Fury X||Radeon Software 17.2.1||—||1050||1000||4096|
|EVGA GeForce GTX 1070 SC2||GeForce 378.78||1594||1784||2002||4096|
|GeForce GTX 1080 Founders Edition||1607||1733||2500||8192|
|GeForce GTX 1080 Ti Founders Edition||1480||1582||1753||11264|
Thanks to Intel, Corsair, Kingston, and Gigabyte for helping to outfit our test rigs with some of the finest hardware available. AMD, Nvidia, and EVGA supplied the graphics cards for testing, as well.
Some game settings (especially texture sizes) were not friendly to the Radeon R9 Fury X’s 4GB of RAM. Where necessary, we reduced these settings to prevent crippling performance issues. All other in-game settings remained the same between the Fury X and the Nvidia cards on the bench.
Unless otherwise specified, image quality settings for the graphics cards were left at the control panel defaults. Vertical refresh sync (vsync) was disabled for all tests. We tested each graphics card at a resolution of 3840×2160 and 60 Hz, unless otherwise noted.
The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Sizing ’em up
Take some clock speed information and some other numbers about per-clock capacity from the latest crop of high-end graphics cards, and you get this neat table:
|Asus R9 290X||67||185/92||4.2||5.9||346|
|Radeon R9 295 X2||130||358/179||8.1||11.3||640|
|Radeon R9 Fury X||67||269/134||4.2||8.6||512|
|GeForce GTX 780 Ti||37||223/223||4.6||5.3||336|
|Gigabyte GTX 980||85||170/170||5.3||5.4||224|
|GeForce GTX 980 Ti||95||189/189||6.5||6.1||336|
|GeForce Titan X||103||206/206||6.5||6.6||336|
|GeForce GTX 1080||111||277/277||6.9||8.9||320|
|GeForce GTX 1080 Ti||139||354/354||9.5||11.3||484|
|GeForce Titan X (Pascal)||147||343/343||9.2||11.0||480|
We won’t be testing every card in the table above, but we think it’s useful to see how far we’ve come from some popular graphics cards of the past. The GTX 1080 Ti and the Titan X Pascal both offer more of, well, everything than any other card in this table save memory bandwidth. As we’ll soon see, however, theoretical peaks alone don’t tell the whole story.
To see how these theoretical numbers play out, we turn to our trusty Beyond3D test suite.
In the first of our synthetic Beyond3D tests, the GTX 1080 delivers what will soon become a familiar sight throughout this review. That 130 GPixels-per-second figure likely falls short of our theoretical peak because of the complex way that Nvidia’s cuts to GP102’s resources likely affect its real-world performance.
Thanks to Pascal’s delta-color-compression mojo and the fact that GP102 is just a whole lotta chip, the GTX 1080 Ti excels at moving a whole lotta data with a 100% compressible black texture. Its incompressible-texture performance approaches the Fury X’s compressed performance. Good grief.
Likely thanks to GPU Boost doing its thing, the 1080 Ti happily beats its theoretical peak texture-filtering rates. It also happily beats up on every other card here.
Here’s a benchmark that’s posed a long-running mystery for us: how do Nvidia’s Maxwell-and-newer chips outstrip their theoretical polygon throughput? Well, we can finally explain what’s going on. TR has long surmised that Nvidia’s Maxwell cards have used a form of tiled rendering to achieve this counterintuitive result, and friend-of-the-site David Kanter proved it a while back. Nvidia admits as much now, and the technique explains the Pascal cards’ performance here.
The GTX 1080 Ti continues to excel in these tests of pure number-crunching ability. It’s hard to say too much about these numbers with my jaw on the floor. Let’s see how these startling benchmarks translate into real-world performance for the GTX 1080 Ti now.
id Software’s 2016 Doom revival is a blast to play, and it’s also plenty capable of putting the hurt on today’s graphics cards. We selected the game’s Ultra preset with 16X anisotropic filtering and 8X TSSAA and dialed up the resolution to 3840×2160 to see what the GeForce GTX 1080 Ti is capable of.
With Doom‘s OpenGL rendering path, the GTX 1080 Ti achieves average framerates unlike anything we’ve ever seen at 4K with ultra settings, and it delivers 99% of its frames within a picture-perfect 16.7-ms window. A+.
Our “time-spent-beyond-X” graphs can be a bit tricky to interpret, so bear with us for just a moment before you go rocketing off to the conclusion. We set a number of crucial thresholds in our data-processing tools—50 ms, 33.3 ms, 16.7 ms, and 8.3 ms—and determine how long the graphics card spent on frames that took longer than those values to render. Those values correspond to instantaneous frame rates of 20 FPS, 30 FPS, 60 FPS, and 120 FPS.
If even a handful of milliseconds start pouring into our 50-ms bucket, we know that the system is struggling to run a game smoothly, and it’s likely that the end user will notice severe roughness in their gameplay experience if time starts building up there. Too much time spent on frames that take more than 33.3 ms to render means that a system running with traditional vsync on will start running into equally ugly hitches and stutters. Ideally, we want to see a system spend as little time as possible past 16.7 ms rendering frames, and too much time spent past 8.3 ms is starting to become an important consideration for gamers with high-refresh-rate monitors and powerful graphics cards.
By those measures, the GTX 1080 Ti is practically perfect for gaming at 4K and 60 FPS with Doom. It spends just a sliver of time on frames that take more than 16.7 ms to render, and it spends 10 seconds less of our one-minute test run past 8.3 ms compared to the GTX 1080. That means a smooth and fluid gameplay experience throughout.
Aside from the API change, we didn’t change any settings to test Doom‘s Vulkan renderer. Here’s what happened.
Although the GTX 1080 Ti’s average frame rate doesn’t get any better with Vulkan on, its 99th-percentile frame time gets a smidge better. The Radeon R9 Fury X is the real beneficiary of this change, slicing its 99th-percentile figure nearly in half. It still can’t quite catch the GTX 1070, however.
Fury X aside, our “time-spent-beyond” graphs are practically identical to our OpenGL results. The GTX 1080 Ti maintains its considerable lead at 8.3 ms, and it delivers sterling results at the 16.7-ms mark.
Crysis 3 hasn’t failed as a test of the mettle of any graphics card we’ve thrown it at, and the GTX 1080 Ti is no different. We dialed in a 3840×2160 resolution and Very High settings all around, save for SMAA antialiasing instead of more demanding methods.
Crysis 3 still has teeth, and the GTX 1080 Ti doesn’t quite have enough power to defang it. Still, we’ve never seen better performance from this game with all the eye candy cranked at 4K. A bit of judicious dialing-back (or a bit of overclocking) could produce the magic 60-FPS average we want to see.
Our time-spent-beyond graphs back up that close-but-no-cigar feeling. The 1080 Ti spends just over six seconds of our one-minute test run on frames that take more than 16.7 ms to render. None of the other cards here can even come close.
Deus Ex: Mankind Divided
We couldn’t quite get Deus Ex: Mankind Divided running well at 4K with any graphics card in our suite, so we decided to switch things up and ran it at 2560×1440 instead with a blend of High and Very High settings with no AA.
Nvidia says the GTX 1080 Ti is about 30% faster than a GTX 1080 on average, and these 2560×1440 results bear out that claim. If you have one of the increasingly-common high-refresh-rate G-Sync panels out there, the Ti is a perfect pairing.
As we should expect, the GTX 1080 Ti spends an imperceptible amount of time past 16.7 ms on difficult frames, but the gap between it and everything else isn’t quite as pronounced in Deus Ex as it is at higher resolutions. Flip over to 8.3 ms, though, and the 1080 Ti asserts its unquestionable superiority once more.
Gears of War 4
Gears of War 4 is one of Microsoft’s first-party DirectX 12-only efforts, and it has plenty of PC-friendly graphics settings to make even the most powerful graphics card sweat. We dialed everything up to Ultra at 3840×2160, save for High shadows and screen-space reflections, before doing our usual test run in the early stages of the game.
Here’s another title where the 1080 Ti could likely be made to run a little faster with a couple judicious settings tweaks. Still, the overall experience from the Ti is fast and smooth.
I hate to sound like a broken record, but the 1080 Ti is the best thing going for minimizing the time spent beyond 16.7 ms on tough frames at 4K. Nothing else comes close.
Grand Theft Auto V
At 4K with maximum settings, Grand Theft Auto V is a worthy opponent for high-end graphics cards. We expected great things from the GTX 1080 Ti here, so we fired up our usual test run and got to it.
High FPS average? Check. Low (albeit not perfect) 99th-percentile frame time? Check. If you want to dial in stupid settings for GTA V at 4K, the 1080 Ti is your best way there.
Yep, still amazing.
Hitman (DirectX 12)
You know the drill by now. We maxed out Hitman at 4K and put our graphics cards to the test.
I’d love to say something interesting about these results, but the Ti defies me with another record-breaking performance. Perhaps it’s more interesting that the once-closely-matched GTX 1070 and R9 Fury X now diverge quite a bit with Hitman‘s DX12 renderer at 4K.
As usual, the 1080 Ti barely puts any time in the 16.7 ms bucket, and all our other cards trail it by wide margins. This is what Tiger Woods must have felt like at the 1997 Masters.
Rise of the Tomb Raider (DirectX 11)
Rise of the Tomb Raider is a tougher cookie than most games, and even the GTX 1080 Ti can’t quite manage a 60-FPS average with it at 4K. Still, we’re setting new bars for performance here: just look at everything else.
Although the 1080 Ti spends enough time past 16.7 ms that you’ll likely notice, all the other cards will make you notice a lot more.
Tom Clancy’s The Division (DirectX 11)
The Division has a reputation for being a toughie, so we set it up with a blend of high and ultra settings and stepped into the disaster-stricken landscape of New York to see whether it lives up to that billing.
The Division‘s open-world New York setting is a formidable challenge for all of these cards to render, and the 1080 Ti’s 99th-percentile frame time indicates that it has to work harder in spots than its commendable average frame rate would first suggest. We wouldn’t put too much stock in the Fury X’s numbers here—it seems our settings overtaxed the Radeon’s 4GB of RAM.
Even with some tough spots and a bit of spikiness, the GTX 1080 Ti once again spends just a handful of seconds past 16.7 ms in The Division. Nothing else is nearly as smooth.
The Witcher 3
The Witcher 3 remains as graphically-challenging a game as ever when the resolution climbs, so we set it up with Ultra settings and HairWorks off at 4K to see how that played in White Orchard.
If you want to enjoy the world of The Witcher in smooth and glorious 4K, the 1080 Ti is your ticket. We’re not sure what happened to the GTX 1070 here, although the result was repeated in all three test runs. We probably need to re-run those numbers for a more accurate picture of performance ASAP.
Ayup. That sure is an impressively small amount of time spent past 16.7 ms. You won’t find a smoother Witcher 3 experience at 4K from any other graphics card.
Watch Dogs 2
Here’s a new addition to our GPU-testing suite. We know from our recent Ryzen review that Watch Dogs 2 can challenge every part of a system, so we turned up the eye candy and walked through the forested paths around the game’s Coit Tower landmark to get our graphics cards sweating. Ignore the 1920×1080 resolution in the settings below.
Like The Division, Watch Dogs 2 is set in a huge open world, but in foggy San Francisco rather than post-apocalyptic New York. Those sweeping vistas full of fine detail put the hurt on any graphics card, and even the 1080 Ti can’t quite get this dog to heel.
Still, the 1080 Ti spends about a third as much time on tough frames that take longer than 16.7 ms to render as the GP104-powered cards do. If you need visuals that match the potential of a 4K screen, the 1080 Ti is the only way to go for smooth gameplay.
A quick turn at overclocking
Nvidia boasts that GTX 1080 Ti cards can reach 2000 MHz or so with some tweaking, so we decided to turn up the clocks in MSI Afterburner to see whether there was any performance left on the table in our Founders Edition card.
To start off, we ran the Unigine Heaven benchmark for 10 minutes to allow the card to get nice and toasty. After that period, the card settled on a 1759 MHz boost clock, and temperatures hovered around 84° C. As is usual for Pascal cards, Nvidia is quite modest about the boost clock range on offer with the GTX 1080 Ti.
With that baseline established, we maxed out the card’s power and temperature limits and began adding a boost clock offset to the card’s stock figure until the Heaven benchmark crashed or we observed other instability. At the end of that process, we achieved a stable 1974 MHz boost clock, and GPU temperatures hovered around 85° C (albeit with much higher fan speeds and more noise than at stock clocks). We didn’t observe any throttling or other limits kicking in at that speed.
Once we had a core clock dialed in, we began increasing memory speeds while playing Doom with its highest-quality textures enabled and its virtual texturing page file at its maximum size. We progressively increased memory clocks and looked for artifacts or other signs of instability. After many iterations of this process, we settled on a 6048 MHz Afterburner memory clock for an effective transfer rate of 12.1 GT/s. While the card was stable at that eye-popping speed, we saw clocks begin dipping into the 1898 MHz-1936 MHz range under load. GPU-Z indicated that the card was hitting a power limit when this happened.
A 12% increase in core clocks and a roughly 10% boost in memory clock speeds is an impressive result, although we would have preferred to have some extra voltage or board power to work with so that the card could sustain both its full memory and boost clock speeds at once. We expect Nvidia’s board partners will take care of this balancing act with their custom card designs.
For an idea of what that performance boost bought us, we ran the built-in benchmark for Rise of the Tomb Raider at 4K using the same test settings we used earlier in the review. At stock clocks, the GTX 1080 Ti ran the benchmark at 67 FPS on average, while our overclock pushed that figure to 73 FPS. 9% more performance at the cost of higher power draw and fan noise is nothing to sniff at.
Be ready for that extra power draw, though, because a pushed GP102 chip sucks down a lot of extra watts. We observed a total system power draw of 414W with our overclocked card running all-out, compared to 365W for stock settings. That’s nothing new with overclocking, though.
We’ll kick off our conclusions with our famous value scatter plots. The best values cluster toward the upper left of the chart, where performance is highest and prices are lowest. We’ve presented this data two ways: one as average FPS per dollar, and the other as 99th-percentile FPS per dollar (converted from frame times so that our higher-is-better approach makes sense).
So there you have it. Whether in performance potential (as measured by average FPS per dollar) or delivered smoothness (as measured by 99th-percentile FPS per dollar), the GeForce GTX 1080 Ti is the finest single-GPU graphics card that we’ve ever tested by a wide margin. It makes smooth, fluid gaming at 4K with ultra settings an accessible prospect, and it’ll scream at similar settings and lower resolutions. In an age where incremental advances are becoming the norm, the GTX 1080 Ti sets a new bar for what’s possible from a gaming PC. Don’t play games on a GTX 1080 Ti unless you want to go home with one. It’s that good.
Though $700 is a lot to pay for a graphics card, I think the GTX 1080 Ti is fairly priced, too. Its 99th-percentile-FPS-per-dollar result is 33% or so better than the GTX 1080’s. With Nvidia’s new $500 suggested price tag for those cards, the 1080 Ti is about 40% more expensive in its Founders Edition form than the former king. Considering that we’re getting similar performance to what the $1300 Titan X Pascal offered before this point, the GTX 1080 Ti seems like a great value. Of course, a $1300 graphics card makes almost anything seem like a great value.
The GTX 1080 Ti Founders Edition card is a fine piece of hardware in its stock form, allowing the GP102 GPU to boost well above Nvidia’s modest 1582 MHz boost clock for sustained periods with low enough noise levels. Overclockers attempting to wring the most from GP102 may want to wait for custom cards with more board power to spare, though, because overclocking both the memory and the GPU core clock on our example required the GPU to run well below its peak overclocked boost speed from time to time thanks to power limits.
Even so, we got our 1080 Ti’s GDDR5X memory to a jaw-dropping 12.1 GT/s stable speed and its GPU to a 1974 MHz boost clock without much drama. Those tweaks were good for an extra 10% or so of performance in our brief tests, and that could make the difference between near-60-FPS and 60 FPS on average in some of the titles we tested at 4K. It’s definitely worth spending some time in MSI Afterburner with the GTX 1080 Ti if you don’t mind more noise.
Whither AMD in these results? We decided not to include the R9 Fury X in our final tally, since not all of our games ran acceptably on it and retail stock of that card seems to be drying up ahead of the Radeon RX Vega launch. Still, some back-of-the-napkin math suggests Nvidia is extracting about twice the performance that Fiji offered from the same power budget. That’s an impressive achievement.
AMD may have a high-end fighter soon in its upcoming Radeon RX Vega, but the GTX 1080 Ti’s arrival now likely isn’t happenstance. Our seat time with pre-production Vega hardware suggests that at least one of those cards will provide GTX 1080-class performance rather than GTX 1080 Ti-slaying power. If so, cutting GTX 1080 prices to $499 is a rather aggressive play by the green team, since it could set a ceiling on what AMD can charge for some members of its latest and greatest. Still, we expect that RX Vega will be competitive with higher-end Pascal for the first time in a while, and if there’s one thing we welcome in the PC hardware market, it’s more competition.
Back in the present, we have no real complaints about the GTX 1080 Ti. It’s as close to perfect as a high-end graphics card can be right now, and its killer combo of delivered performance, high efficiency, and a reasonable price tag make it a shoo-in for a TR Editor’s Choice award.