Nvidia’s GeForce GTX 1080 set a new high-water mark for single-GPU graphics card performance in our review, and gamers will pay for the privilege of owning one. The card has been difficult to find in stock ever since its release, and prices on the models in stock at Newegg tend to start at the $699.99 suggested sticker for the Founders Edition card. The $599.99 suggested price for custom cards is nowhere to be seen—those models are selling for well over $800 when they’re available. Owning the fastest thing around doesn’t come cheap.
The second consumer Pascal card has a different mission. When it launched the GTX 1070 at DreamHack, Nvidia promised performance greater than a GeForce GTX Titan X for a $379.99 starting price (or $449.99 for the Founders Edition reference card). The Titan X sold for $1000 when it was still available, but a more relevant point of reference for most gamers is the other GM200-powered consumer graphics card: the GeForce GTX 980 Ti. That card listed for $650 when it first hit the market. No matter how we slice the numbers, the GTX 1070 should offer potent performance at a more accessible price point, much like the GTX 970 did when it launched alongside the GTX 980.
|GTX Titan X
To hit that more affordable sticker price, Nvidia lopped off a full graphics processing cluster from the GP104 GPU. That gives the GTX 1070 1920 stream processors and 120 texture units, compared to the GTX 1080’s complement of 2560 SPs and 160 texture units. The engineers left the chip’s ROP count and memory controllers alone, so the GTX 1070 ships with GP104’s 64-ROP complement and 256-bit path to memory intact. Here’s a picture of what that might look like as a block diagram:
Compared to the GeForce GTX 970 before it, the GTX 1070 has 15% more raw shading resources and 15% more texture units at its disposal. Pair those with the considerable clock boost that Pascal brings, though, and the card’s theoretical performance slightly eclipses both the GTX 980 Ti and the Titan X. Nvidia clocks the GeForce GTX 1070 Founders Edition card at 1506-MHz base and 1683-MHz boost speeds. Custom cards from Nvidia’s board partners can run even faster still. We’ll examine just what that means for the GTX 1070’s theoretical performance in a bit.
Nvidia also kept the GTX 1070’s price in check by relying on 8GB of good old GDDR5 RAM instead of the GDDR5X memory found on the GTX 1080. In its reference design, the GTX 1070 clocks that RAM at 8GT/s, up from the 7GT/s memory speeds we saw with the GTX 980 and GTX 970. Pair that with GP104’s 256-bit memory bus, and we get 256GB/s of potential bandwidth on tap.
Aside from those changes, the GTX 1070 offers the same improvements from the move to Pascal we detailed in our review of the GTX 1080. If you’re not already familiar with that card and the GP104 GPU, you should brush up on those changes before moving on here. Next, let’s take a look at the MSI GTX 1070 Gaming Z card we’ll be using to test out this configuration of the GP104 GPU.
The MSI GeForce GTX 1070 Gaming Z
We’re able to review the GTX 1070 today because MSI sent over its spiffy GTX 1070 Gaming Z card. Behold:
This GTX 1070 is the highest-end card in the company’s lineup. It uses a dual-fan “Twin Frozr VI” cooler to keep the GP104 chip in check, and it’s got LED accents scattered across its surface to please buyers’ inner 13-year-olds. In all seriousness, though, we think the card looks quite sporty. Those red lightning-bolt-ish things that surround the forward fan light up when the card is powered on.
The back side of the GTX 1070 Gaming Z is almost entirely covered with a metal backplate. The dragon crest embedded in the plate is backlit by RGB LEDs for easy color-coordination with other lighting in a build, and the MSI logo on the side of the card is similarly bedazzled. You can see the six-pin and eight-pin power connectors this card needs to do its thing from this angle, too.
MSI got in hot water a while back for sending reviewers graphics cards with a special BIOS that activates an “OC mode” clock profile by default. Our card is flashed with such a BIOS. Most custom graphics cards built these days feature several such clock speed profiles that can be selected in a proprietary companion utility, but reviewers don’t generally install that software. For what it’s worth, MSI’s rationale for sending reviewers cards configured this way is that it doesn’t want the press to miss out on the full performance potential of its products.
We’re mostly neutral about this practice, since the clock speed differences often amount to a few percent at best. Still, folks who go out and pick up an MSI GTX 1070 will want to install the MSI Gaming App and enable the card’s OC Mode if they’re looking for performance identical to what we measure in this review. We’ve detailed the changes that OC Mode produces in the following table. We’ve also tossed the GTX 1070 Founders Edition card in for comparison.
|MSI GeForce GTX 1070 Gaming Z (gaming mode)
|1x 6-pin, 1x 8-pin
|MSI GeForce GTX 1070 Gaming Z (OC mode)
|GeForce GTX 1070 Founders Edition
As we begin to field-strip the GTX 1070 Gaming Z, you can see the RGB LED backlight for the MSI crest and its associated connector. Otherwise, all that’s back here is a matte-black PCB. You can get some sense of how unusually wide the PCB on this card is by taking a look at how far it extends past the mounting bracket—we’re talking Asus Strix GTX 980 Ti proportions here.
Taking off the heatsink reveals five heatpipes running over a flat base plate. MSI also applies finned heatsinks to much of the card’s power circuitry. A flat metal plate transfers heat from the GDDR5 RAM ringing the GP104 GPU. You can also see the generous dollop of “premium thermal compound” MSI applies to the chip itself.
Removing these auxiliary heatsinks and cleaning off the thermal paste gives us a better look at the Gaming Z card’s 10-phase power-delivery subsystem, the GDDR5 memory, and the GP104 GPU itself. To send power to those VRMs, MSI uses a pair of PCIe power connectors: a six-pin and an eight-pin plug. The GTX 1070 Gaming Z also offers a DVI ouput, three DisplayPort 1.3 outputs, and a gold-plated HDMI 2.0 connector.
Overall, MSI makes the Gaming Z card feel worthy of its $469.99 price tag, and that’s no surprise. We’ve enjoyed the company’s custom GeForces for quite some time. MSI’s GeForce GTX 970 Gaming 4G card was a perennial feature of our System Guides when that card was current, and we think the GTX 1070 Gaming cards continue that tradition.
Along with the Gaming Z, MSI also offers a couple milder GTX 1070s. The GTX 1070 Gaming card has a simpler backplate, and its 1531-MHz base and 1721-MHz boost clocks are less aggressive than the top-end model’s. This card also doesn’t have the multiple clock profiles of MSI’s fancier 1070s. The Gaming X card keeps the simple backplate of its less-expensive cousin, but it pushes clocks to 1582-MHz base and 1771-MHz boost speeds in its “gaming mode.” Ticking the “OC mode” checkbox pushes those clocks to a 1607-MHz base and a 1797-MHz boost.
All of these cards appear to use similar PCB designs and coolers, so the question of which one to select comes down to your budget and desired clock speeds. The GTX 1070 Gaming card sells for $439.99, while the Gaming X card is $449.99 and the Gaming Z sells for the aforementioned $469.99. We suspect the Gaming card could probably be overclocked to match or beat its more expensive cousins, but folks who don’t want to mess with Afterburner could be forgiven for dropping the extra $10 on a Gaming X. The additional $20 for the Gaming Z card gets you extra bling and slightly higher clocks still. We think it’s hard to go wrong with any of these cards, so let your wallet decide.
Now that we’ve seen the GTX 1070 Gaming Z, let’s see what the cut-down GP104 GPU can do.
Our testing methods
As always, we did our best to deliver clean benchmarking results. Our test system was configured as follows:
|Asus X99 Deluxe
|16GB (4 DIMMs)
|Corsair Vengeance LPX
DDR4 SDRAM at 3200 MT/s
|Intel Management Engine 188.8.131.525
Intel Rapid Storage Technology V 184.108.40.2061
|Integrated X99/Realtek ALC1150
Realtek 220.127.116.1125 drivers
|Kingston HyperX 480GB SATA 6Gbps
|Fractal Design Integra 750W
|Windows 10 Pro
|Asus Strix Radeon R9 Fury
|Radeon Software 16.6.1
|Radeon R9 Fury X
|Radeon Software 16.6.1
|Gigabyte Windforce GeForce GTX 980
|MSI GeForce GTX 980 Ti Gaming 6G
|MSI GeForce GTX 1070 Gaming Z
|GeForce GTX 1080 Founders Edition
Our thanks to Intel, Corsair, Asus, Kingston, and Fractal Design for helping us to outfit our test rigs, and to MSI, Nvidia and AMD for providing the graphics cards for testing, as well.
For our “Inside the Second” benchmarking techniques, we use the Fraps software utility to collect frame-time information for each frame rendered during our benchmark runs. We sometimes use a more advanced tool called FCAT to capture exactly when frames arrive at the display, but our testing has shown that it’s not usually necessary to use this tool in order to generate good results for single-GPU setups. We filter our Fraps data using a three-frame moving average to account for the three-frame submission queue in Direct3D. If you see a frame-time spike in our results, it’s likely a delay that would affect when a frame reaches the display.
You’ll note that aside from the Radeon R9 Fury X and the GeForce GTX 1080, our test card stable is made up of non-reference designs with boosted clock speeds and beefy coolers. Many readers have called us out on this practice in the past for some reason, so we want to be upfront about it here. We bench non-reference cards because we feel they provide the best real-world representation of performance for the graphics card in question. They’re the type of cards we recommend in our System Guides, so we think they provide the most relatable performance numbers for our reader base.
To make things simple, when you see “GTX 1070,” “GTX 980,” or “GTX 980 Ti” in our results, just remember that we’re talking about custom cards, not reference designs. You can read more about the MSI GeForce GTX 980 Ti Gaming 6G in our roundup of those custom cards. We also reviewed the Gigabyte Windforce GeForce GTX 980 a while back, and the Asus Strix Radeon R9 Fury was central to our review of that GPU.
Each title we benched was run in its DirectX 11 mode. We understand that DirectX 12 performance is a major point of interest for many gamers right now, but the number of titles out there with stable DirectX 12 implementations is quite small. DX12 also poses challenges for data collection that we’re still working on. For a good gaming experience today, our money is still on DX11.
Finally, you’ll note that in the titles we benched at 4K, the Radeon R9 Fury is absent. That’s because our card wouldn’t play nicely with the 4K display we use on our test bench for some reason. It’s unclear why this issue arose, but in the interest of time, we decided to drop the card from our results. Going by our original Fury review, the GTX 980 is a decent proxy for the Fury’s performance, which is to say that it’s not usually up to the task of 4K gaming to begin with. You can peruse those numbers and make your own conclusions.
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:
|Radeon R9 290X
|Radeon R9 Fury
|Radeon R9 Fury X
|GeForce GTX 780 Ti
|Gigabyte GTX 980
|MSI GeForce GTX 980 Ti
|GeForce Titan X
|MSI GeForce GTX 1070
|GeForce GTX 1080
In the custom-card form we’re looking at today, the GTX 1070 delivers on its promise of beating out the Titan X in many regards, at least on paper. Its peak pixel fill rate even exceeds that of the GTX 1080 Founders Edition. Since it’s down a GPC—and therefore a raster unit—compared to the GTX 1080, however, the GTX 1070 can’t move as many triangles as its bigger brother, and it also falls short of the much larger GM200 chip. Let’s see how these numbers translate to performance in our directed Beyond3D benchmark suite.
Despite its peak numbers in the table above, the GTX 1070 falls slightly behind the GTX 980 Ti in our directed pixel-fill benchmark. This may be one of the areas where being down a GPC on the GTX 1080 hurts the 1070 a bit.
Since the GTX 980 Ti and the GTX 1070 both rely on GDDR5 memory, it’s easy to make cross-generational comparisons here. The GTX 1070’s higher-clocked GDDR5 and color-compression mojo can’t make up for the fact that its bus width is significantly narrower than that of the GM200 chip on board the 980 Ti.
The GTX 1070 ekes out a win over the GTX 980 Ti in our texturing tests, thanks to its prodigious clock speeds.
Despite its theoretical disadvantage in polygon throughput, the GTX 1070 can also move considerably more triangles than the GTX 980 Ti.
The GTX 1070 lives up to its number-crunching potential in our directed ALU tests, although the GTX 980 Ti isn’t far behind. It’s impressive that the GTX 1070 can deliver this kind of performance despite being down 896 stream processors on the GTX 980 Ti. Now that we’ve teased out the differences that disabling a GPC on GP104 makes, let’s run the GTX 1070 through some real-world tests.
Grand Theft Auto V
Grand Theft Auto V has a huge pile of image quality settings, so we apologize in advance for the wall of screenshots. We’ve re-used a set of settings for this game that we’ve established in previous reviews, which should allow for easy comparison to our past tests. GTA V isn’t the most demanding game on the block, so even at 4K you can expect to get decent frame times out of higher-end graphics cards.
Out of the gate, the GTX 1070’s average frame rates come out a nose ahead of the GTX 980 Ti’s. The card also delivers ever-so-slightly better frame times than the GM200 card in our 99th-percentile measure, and that result is backed up by our frame time graph. Neither card has any trouble delivering smooth gameplay at 4K with GTA V, although folks who want to break the 60-FPS barrier with such a high-res monitor will still need to step up to the GTX 1080.
These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid. The 50-ms threshold is the most notable one, since it corresponds to a 20-FPS average. We figure if you’re not rendering any faster than 20 FPS, even for a moment, then the user is likely to perceive a slowdown. 33 ms correlates to 30 FPS or a 30Hz refresh rate. Go beyond that with vsync on, and you’re into the bad voodoo of quantization slowdowns. And 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame.
The GTX 1070 spends less time beyond the critical 16.7-ms mark than the GeForce GTX 980 Ti, more or less halving that card’s time spent at frame rates lower than 60 FPS. Our older numbers with the Radeon R9 Fury and Fury X might look better with the latest 16.7.2 driver, which contained targeted updates designed to improve smoothness in GTA V with the Radeon RX 480. As things stand, the Fury and Fury X sandwich the GTX 980 in our results. All three of those cards deliver considerably less smooth experiences in GTA V than one might get with a GTX 980 Ti, a GTX 1070, or a GTX 1080.
Crysis 3 is an old standby in our benchmarks. Even though the game was released in 2013, it still puts the hurt on high-end graphics cards. Unfortunately, our AMD Radeon R9 Fury and the 4K display on our test bench have a disagreement of some sort, so the red team is only represented by the Fury X on this set of benches.
In Crysis 3, the GeForce GTX 980 Ti takes a razor-thin lead over the GTX 1070 in both our average FPS and 99th-percentile frame time measures. For most intents and purposes, the cards perform identically in this title, but our advanced frame-time metrics show where they differ.
The GTX 1070 spends barely any time past the 33.3-ms mark in our testing, but it spends slightly more time than the GTX 980 Ti churning on frames that drop the frame rate below 60 FPS. The GeForce GTX 1080 continues to lead the pack in both potential and delivered performance. The Radeon R9 Fury X has a slightly harder time of it still, while the GTX 980 is winded by Crysis 3 at 4K if we use 16.7 ms as our threshold.
Rise of the Tomb Raider
Rise of the Tomb Raider is the first brand-new game in our benchmarking suite. To test the game, we romped through one of the first locations in the game’s Geothermal Valley level, since it offers a diverse backdrop of snow, city, and forest environments. RotTR is a pretty demanding game, so we took this opportunity to dial the resolution back to 2560×1440. We also turned off the AMD PureHair feature to avoid predisposing the benchmark toward one card or another in this test, since Nvidia’s HairWorks has created significant performance deltas in past Tomb Raider games when we’ve had it on.
Noticing a pattern yet? The GeForce GTX 1070 takes back the average-FPS lead by a hair in Rise of the Tomb Raider, and it turns in a slightly better 99th-percentile frame time than the GTX 980 Ti, as well.
In our measures of “badness,” the GTX 1070 eclipses the GTX 980 Ti in the time spent working on frames past the critical 16.7-ms mark. The cards are still quite closely matched, though. Both Radeons deliver considerably less smooth experiences than any of the GeForce cards in our test suite—RoTR really seems to favor Nvidia cards.
The GeForce generations flip again in our Fallout 4 results. The gap is the widest it’s been in our results yet, though. The GTX 980 Ti is about 5% faster in our average FPS measure, and it delivers a better 99th-percentile frame time than the GTX 1070, too. Just goes to show the GTX 1070 can’t win ’em all, we suppose.
In our measures of “badness,” none of our cards spend any time past the 50-ms or 33.3-ms marks. The GTX 1070 spends about twice as much time past 16.7 ms working on challenging frames, but neither it nor the GTX 980 Ti spend more than a second doing so in absolute terms. The GTX 1080 is still the smoothness champion for Fallout 4 if you’re targeting 4K and 60 FPS.
The Witcher 3
The Witcher 3 is another benchmark where we’ve re-used the settings that we’ve settled on in past reviews. We also chose to test this title at 2560×1440 rather than 4K. We didn’t crank the resolution in part because we wanted to maintain consistency with the numbers we produced in our Radeon R9 Fury X review, but also because the game is demanding enough that playing the game at 4K with high settings wasn’t a great experience even on newer high-end cards.
The GTX 980 Ti turns in another razor-thin victory over the GTX 1070 in The Witcher 3. Let’s see if our frame-time metrics can tell us why.
As it happens, the GTX 1070 spends a bit more time working on frames past the 16.7-ms mark. Both cards still turn in admirable performance here, though. You’d have to be paying close attention to discern which card was painting Geralt on your screen with these numbers.
The 2016 version of Hitman closes out our test suite. We chose to bench this demanding title at 4K to really make our graphics cards sweat.
Hitman continues to be a major challenge for modern graphics cards to run at 4K. The GTX 1070 takes a tiny lead over the GTX 980 Ti here, but both cards are closer to the ragged edge of playability than we’d like.
In our measures of “badness,” we can see that the GTX 1070 delivers a slightly smoother experience overall than the GTX 980 Ti. Both cards spend considerable amounts of time past the 16.7-ms mark, though, so a smooth 60-FPS experience at 4K remains elusive.
Now that we’ve wrapped up our performance analysis, let’s see how the MSI GeForce GTX 1070 Gaming Z card performs on the power, noise, and temperature fronts.
Let’s take a look at how much juice the GTX 1070 needs to do its thing. Our “under load” tests aren’t conducted in an absolute peak scenario. Instead, we have the cards running a real game, Crysis 3, in order to show us power draw with a more typical workload.
At idle, our test system consumes a few more watts with the overclocked GTX 1070 inside than it does with the stock-clocked GTX 1080 Founders Edition. That shouldn’t come as any surprise.
Loading the GTX 1070 up with Crysis 3 brings our system power draw within 10 watts of the GTX 1080 FE. Impressively, we can see that the GTX 1070 delivers GTX 980 Ti-class performance while allowing our test system to draw 27% less power under load. Process shrinks are a wonderful thing.
At idle, all of our cards except the GTX 1080 stop their fans, so the noise floor of our lab is the limitation of our measurement. Under load, however, the GTX 1070 Gaming Z card produces noise levels bested only by the Radeon R9 Fury X’s closed-loop liquid cooler. MSI’s GeForce GTX 980 Ti was already among the quietest cards of that type we’ve tested, so it’s good to see the company continuing its commitment to quiet gaming.
Most importantly, the sound from MSI’s cooler is also a pleasant, broad-spectrum one. Even when we pushed them to full blast for the heck of it, the twin fans didn’t lose their polite character. Overall, MSI deserves high praise for the noise, vibration, and harshness characteristics of its latest cooler.
Of course, quiet computing is just one of the things a good graphics-card heatsink needs to deliver. The Radeon R9 Fury X and its liquid cooler aside, the GTX 1070 Gaming Z is only bested by the much louder and more aggressive cooler on the Gigabyte GTX 980 Windforce card. While it’s not an entirely fair comparison, it’s clear from these results just how much of a drop in temperatures one can get from a custom card compared to Nvidia’s reference design, too. The Founders Edition GTX 1080 runs a full 15° C hotter than the GTX 1070 Gaming Z, even if that card does have the benefit of exhausting hot air directly from the case. We may have to see how MSI’s cooler performs atop a fully-enabled GP104 chip at some point.
Let’s sum up our results with a couple of our famous scatter plots. The best values tend toward the upper-left corner of each plot, where performance is highest and prices are lowest. We’ve converted our 99th-percentile frame time results into FPS to make our higher-is-better system work.
To account for the actual conditions of the graphics card market right now, we’ve surveyed Newegg and averaged the price of all in-stock models of the GeForce GTX 980 Ti and the Radeon R9 Fury X. We’ve used the actual price of the Gigabyte Windforce GTX 980 card we employ in our testing, since it’s one of the few GTX 980s still available. We’ve also used the $469.99 retail price of the MSI GTX 1070 Gaming Z card itself in these graphs, since pricing is so wildly variable on GTX 1070s right now.
Unless you skipped straight to our conclusion, the plot of FPS per dollar above should come as no surprise. The GTX 1070 comes eerily close to duplicating the performance of the GTX 980 Ti for less money than that card demanded at launch. The GTX 1070’s dot would move even further to the left if we consider that some of those cards are selling for as little as $410 on Newegg right now. Products that deliver more bang for the buck are always easy to get excited about, and the GTX 1070 offers that improvement in spades. The $470 retail price of the fancy MSI card we tested skews the results a bit, though.
In our 99th-percentile FPS per dollar plot, the GTX 1070 just edges out the GTX 980 Ti. We’d be hard-pressed to call one card or the other better by this measure—they both deliver exceptionally smooth gameplay for the money. That said, we’d always pick the Pascal card over a deeply-discounted GTX 980 Ti for its higher power efficiency and new architectural features. Meanwhile, the Radeon R9 Fury X trails the GTX 1070 in both our average FPS-per-dollar and 99th-percentile FPS-per-dollar measures.
Since the GTX 1070 delivers about the same performance as hot-clocked GTX 980 Tis, folks with those cards have no reason to ditch them for this Pascal-powered product. Those gamers won’t get a performance boost without stepping up to a GTX 1080. Owners of GTX 770, GTX 780, and perhaps even GTX 780 Ti cards—and similar products from AMD—will likely find the GTX 1070 a compelling upgrade for the money, though. It’s also the only sensible step up from the just-released GTX 1060 for builders who want more performance from an Nvidia card without dropping $600 or more on a GTX 1080.
For better or for worse, the GTX 1070 also completes Nvidia’s stranglehold on the high-end graphics card market. Some Radeon R9 Fury X cards are now available for prices similar to some GTX 1070s, but the Fury X’s performance in our tests often trails the Pascal card, and it needs over 100W more power to hang in there under load. We’d love to see the same kind of vigorous competition at this price point that we’re seeing around $200 to $250 thanks to the Radeon RX 480 and the GTX 1060, but our best guess is that AMD’s next high-end graphics card is a while off yet.
It’s not all bittersweet news today, though. The GTX 1070 Gaming Z card that MSI sent us shows off what’s possible when Nvidia’s board partners work their magic with Pascal. It’s an excellent piece of hardware in every regard: quiet, cool, and fast. Thanks to the strong demand for Pascal cards, however, it’s selling for $90 more than Nvidia’s $380 suggested price for custom GTX 1070s. We still think it’s an OK value at that price compared to the Radeon R9 Fury X and the GTX 1080, but we think you’d really have to like the RGB LED lighting on the backplate to make it worth the $20 price jump over the nearly identical Gaming X variant.
Still, MSI has carried over the mojo that made its GTX 970 Gaming 4G card one of our favorites of its generation. We think the Gaming X and Gaming Z cards are well worth a look if you’re in the market for a GTX 1070.