Nvidia’s newest graphics card is — again — its most powerful, its most energy efficient, and its best for next-generation gaming in virtual reality and in Ultra HD resolutions. It’s also surprisingly cheap internationally, and unsurprisingly expensive in Australia. But price aside, if you do happen to pick up a new GTX 1080-based card either from Nvidia or any of its manufacturing partners, you’ll get yourself an extremely powerful and future-proofed card that also serves as a great bellweather for what will be a very important year in graphics technology.
What Is It?
- Base Clock: 1607MHz
- Boost Clock: 1733MHz
- Memory Clock: 10.0Gbps
- Memory: 8GB GDDR5X
- Power: 180W, 8pin
The Nvidia GeForce GTX 1080 is the first graphics card released by Big Green on its brand new Pascal microarchitecture. Pascal is the first 16-nanometre graphics card architecture out there, and that represents a significant leap forward in the efficiency of turning electricity into 1s and 0s versus the older 28-nanometre process used in Nvidia’s Kepler and Maxwell cards — you might know them better as the GeForce GTX 6xx, 7xx and 9xx series. The GeForce GTX 1080 is — for now — the flagship card in the line, with a slightly less powerful GTX 1070 also being released a couple of weeks after the 1080.
The GTX 1080 is significantly more efficient than the GTX 980 and GTX 980 Ti and GTX Titan X that it inherits Nvidia’s gaming performance crown from. The chipmaker says it trounces all three of these cards in outright gaming performance, but at the same time it does so while consuming less power. And you can tell that’s true purely through the card’s physical design — it uses a single 8-pin power connector and is rated for 180 Watts of energy usage, where previous models were in the 250 Watt range and used a combo of paired 8- and 6-pin or 8- and 8-pin cables.
[clear]
While the GTX 1080’s geometrically-interesting shroud looks new, its blower design and integrated vapor chamber are effectively the same as the one used on the GTX 980 Ti and Titan X. Third-party cards will have their own (usually superior) cooling solutions. The rear I/O design, too, is the same as 980 Ti with three full-size DisplayPort and one HDMI, as well as a legacy DVI-D connector. The difference is that these ports are all the latest standard — the HDMI, for example, is HDMI 2.0b and supports all the newest gizmos like High Dynamic Range video output and 4K at 60 frames per second.
The GTX 1080 is the first card in Nvidia’s new line-up to use GDDR5X memory, a faster and higher-bandwidth standard that’s a little more expensive than regular GDDR5 but not as expensive as High Bandwidth Memory (HBM), as used in AMD’s new Radeon cards. GDDR5X is anywhere up to two times as fast as GDDR5, and Nvidia has stacked the GTX 1080 with a full 8GB complement of it — this card is intended to be used for the high-res textures and assets of Ultra HD and VR games of this year and next.
What’s It Good At?
In our standardised 1080p benchmarks, the GeForce GTX 1080 monsters through every game thrown at it — it’s more than overkill if you’re looking for consistent 1080p performance above 60 frames per second. It’s a similarly powerful choice for 1440p; with the GTX 1080 being the first in a new generation of graphics cards and also the first of the year alongside AMD’s imminent Radeon refresh — code-named Polaris and Vega — we’re also moving to only running graphics tests both at 1440p and 2160p (4K Ultra HD, 3840×2160 resolution). We’re also going to abandon a few of these games and move to a more modern suite of Rise Of The Tomb Raider, Grand Theft Auto V, DOOM, Project Cars and Battlefield 4 for various DirectX 12, Vulkan and VR testing.
Nvidia GeForce GTX 1080: Average Frame Rates
Far Cry 4 (Ultra): 1080P: 132FPS 1440P: 96FPS 2160P: 49FPS
Battlefield 4 (Ultra): 1080P: 142FPS 1440P: 69FPS 2160P: 61FPS
Crysis 3 (Very High): 1080P: 141FPS 1440P: 69FPS 2160P: 55FPS
Metro: Last Light (Very High): 1080P: 85FPS 1440P: 60FPS 2160P: 46FPS
Tomb Raider (Ultimate): 1080P: 201FPS 1440P: 140FPS 2160P: 58FPS
Almost across the board, this represents a roughly 35 per cent performance improvement on Nvidia’s GeForce GTX 980, and around 20 per cent on the previous generation’s card to beat, the GTX 980 Ti, especially at 1080p and 1440p resolutions. The gap narrows somewhat up at the pointy end of 4K, but consistent average frame rates of 45-60FPS shows that this is just about the first single-card solution that makes 4K games playable at a consistently smooth frame rates. Anecdotally, I can say that it’s still a little short of impressive when it comes to demanding virtual reality resolution and frame rate, but I don’t have the numbers to back that up.
[clear]
Nvidia has sold its first-run Founder’s Edition cards to its potential buyers on the strength of its craftsmanship and design, and it absolutely is a beautiful piece of technology — as much as a graphics card can be. For something that sits hidden away inside most computer cases, this isn’t a big deal, but if you’re the kind of hardcore gamer that will buy the 1080 to show it off lit up inside your windowed PC case, then you’ll appreciate its chiseled good looks. It certainly feels and looks solid.
The GTX 1080 is partial to a bit of overclocking, too. I’m not a particularly avid overclocker myself — my experience ended with tracing pencil lines on old AMD chips — but the 1080 is the first Pascal card that supports a one-click, card-customised overclocking process through a third-party tool like EVGA’s Precision. The Pascal architecture lets users find the maximum safe clocks at each voltage increment — Nvidia calls this GPU Boost 3.0 — and it’s a very easy way to maximise performance. I hit a maximum full-voltage boost clock of 2.04GHz perfectly easily on my GTX 1080, and from everything I’ve read I’d expect that to be a relatively mainstream result.
What’s It Not Good At?
Nvidia’s claims about performance have, this particular card and this particular generation of cards, been significantly massaged to make it sound more powerful than it is — which is a pity because it’s already so powerful that it doesn’t need any superfluous marketing on top of that excellent hardware performance. Nvidia says you’ll get three times better performance than the previous generation of GeForce cards, but that 3x figure is only within one specific virtual reality workload — and relies on some new Nvidia tech inside the card that have significantly improved VR processing. So the claims are accurate, just very specific. For regular, everyday gaming, it’s about 30 per cent better than last generation’s GTX 980. Which is plenty.
The GTX 1080, as far as we can tell so far, is going to suffer from a serious dose of Australia Tax when it launches locally on May 27. At $US599 for ‘regular’ cards and $US699 for Founder’s Edition units, those prices translate to $850 and $960 respectively, but Kotaku is reporting local stockists charging up to $1500 for early adopters. This is seriously not on — considering I could almost certainly secure myself one shipping internationally for under $1000 — and it’s a good reason to wait a few weeks or months after launch to see prices settle significantly.
[clear]
And, on May 27, you’ll likely see a range of Founder’s Edition cards available from third-party vendors like MSI and Gigabyte and EVGA, with similar pricing to Nvidia’s regular faux-reference edition unit and with the same cooling solution and reference PCB design, but with third-party merchandise and software and enticing extras thrown in. It’s almost a certainty that these will be better value than the Nvidia version, which will still continue to be sold for the life of the card. Or you could also wait for a more basic third-party version, albeit sans out-of-the-box overclocking and a fancy dual- or triple-fan setup, for that $US599 price point or even lower.
Should You Buy It?
I’m really, genuinely impressed with the performance that I got out of the Nvidia GeForce GTX 1080. It’s a crazy powerful graphics card, and brings that 30 per cent boost in performance at only 70 per cent of the energy consumption of the previous generation of graphics card. Simply that fact, on paper and in abstract, should be enough to tempt you to purchase the GTX 1080. It’s an impressive feat for Nvidia to pull off.
But then there’s the price tag to consider. Taken in the context of the US$999 ($1600 locally) GTX Titan X that it beats soundly in outright flatscreen gaming performance, and the still-circa $1200 local price of GTX 980 Ti graphics cards available today through All Reputable PC Component Retailers, the GTX 1080 comes with a great price tag. But there’s a big Australia Tax, and a big early adopter tax on top of that.
[clear]
This is shaping up to be a very exciting year in PC graphics, and it’s guaranteed to be one of the best years to upgrade your gaming rig — especially if you’re using Nvidia’s 700-series or AMD’s 7000-series GPUs in your PC already, or an even older card. A card like the GTX 1080 will likely be more expensive than its sticker price if you take into account the fact that you should probably get a new (1440p or 4K) monitor, and you don’t want your old CPU to bottleneck things, and you could really do with a SSD as well…
I’m very interested to see what AMD brings to the table in the next couple of weeks and months. With that said, it’s almost a certainty that Nvidia’s GeForce GTX 1080 will hold the single-card gaming performance crown until the 1080 Ti or the next Titan is announced, and if flagship levels of graphical grunt are what you’re looking for then you won’t be disappointed with the GTX 1080 that you buy.
Our testbed for all our 2016 graphics card benchmarks is a pretty powerful rig — it’s the kind you’d buy if you were a serious gamer, but not necessarily into extreme overclocking and getting the best possible synthetic test scores. It uses a sixth-gen Intel Core i7-6700K processor running at stock 4.0GHz clocks on an Asus Maximus VIII Gene, has 32GB (4x8GB) of Corsair Dominator Platinum DDR4 RAM, and uses a 512GB Samsung 950 Pro M.2 SSD for Windows and game files alike. Cooling is courtesy of a Noctua NH-U12S and everything is housed in a Fractal Design Define Mini.
This story originally appeared on Gizmodo Australia
Comments
26 responses to “Here’s The Australian Review Of NVIDIA’s GeForce GTX 1080”
Thank you for the aussie review.
Thats a serious mark up for founders editions..
I am waiting for the ASUS DirectCU IV model, they are usually quite beastly when it comes to overclocking.
Being fair, once you add tax + cost of shipping to Aus, these cards would retail for about $1100-$1200.
If your running a 9xx series at the moment, the biggest advice I think is wait. Mostly to see what pressure AMD can put on Nvidia pricing. Even if you won’t buy AMD regardless, these launch prices are crazy for what is essentially a midend card with IPC regression.
Again with the ‘midend’ nonsense. The benchmarks and in-depth analysis by respected hardware sites tell a very different story to what you’re saying.
Using computerbase.de, mostly because they did the most consistent testing of OC’d performance in DX11 @ 1440p the difference between an OC’d 980 Ti and a 1080 is 13.74%. We saw a larger jump than that from 980 to 980 Ti for petes sake.
On top, we all know that GP102 is coming, and that we have two generations of history of nvidia launch x70\x80 cards as a small die chip, followed by a large die Ti\Titan. And each time, they have increased the MSRP of x70\x80 cards on launch.
Give me A: Any reason why this will change (and that 1080 is actually high end Pascal), and B: that clock to clock, Maxwell has a higher fps per mhz.
You might be confusing ‘high end’ with ‘the best single card in the series’. The entire x80 series throughout Nvidia’s product range are high end cards. Nobody’s saying the 1080 is the best Pascal card, just that it’s very much in the high end of Nvidia’s range, as are the Titan and x80 Ti models.
FPS per MHz is a meaningless statistic on its own. Performance is a composite metric, it can be raised by improving any or all of its contributing factors. It’s why synthetic benchmarks have declined so much in favour of application benchmarks which give a more accurate picture of real world performance.
Eh, unless your someone who looks favourably on P4 Netbursts, IPC is king, and raising clocks to cover it is just lazy.
I would argue (since any card below these are just rebrands). 950\960 (Low End), 970\980 (Mid), 980 Ti\Titan X (High End). I know Nvidia\AMD technically call Ti\X enthusiast level (and 970\980 high), but if your pricing these cards as high as enthusiast level, I’m still calling bullshit.
I’d argue that IPC is the least independent metric. It has no meaning outside of its relationship with the clock speed, so the two are only relevant as a pair. As long as the multiplication of the two nets a higher result you’ll get higher performance.
I agree that IPC is the better metric to improve from a tech standpoint because then you can slap high clock speed on top to get even more output, but that’s really an academic distinction. MIPS is what matters for end use.
Ever since the 580 series Nvidia just moved the bar. What used to just be the high end card is now the “enthusiast card” and their mid tier are now high end. I generally go by the chip codes GP100, GP104 etc when determining where the card should be as opposed to the model numbers. For example the GTX560 was GF114, the GTX580 was GF110 and the GTX1080 is GP104. The 680/780 generation should have been combined with the 680 being mid range. I can’t help but feel ripped off these days. This is just my paranoid opinion.
I don’t use their definitions of high/enthusiast/whatever, I usually compare by price bracket and relative performance.
The problem is Nvidia moved that price bracket to give us all a jolly good rogering. I’ve been saving up for years and I’ve got about $1100 burning a hole in my pocket but I can’t bring myself to buy this card. I have to wait for the TI due to my own sense of value.
I’m just trying to explain the thought process for those of us who consider it midend.
Almost every reputable hardware review site would speak contrary to that, but okay…
I have not had good experiences with AMD in the past, and have few reasons to doubt nVidia at this point. The positive reviews for the 1080 along with my pleasant experiences with previous DirectCU overclocks (anything ROG to be honest) make this a very appealing upgrade, especially as I have a 780 at the moment.
I was remarking at the few retailers asking 1500 for the card. Regardless I tend to wait a few months with new hardware to allow the proces to settle a bit. It is “fair enough” only because we must swallow the crazy hardware tax in order to obtain goods which are quite affordable otherwise.
If you honestly think a card that can hit 60fps on Metro: Last Light at 2560×1440 is “mid end” then I want whatever you’re smoking.
EDIT: I do agree that upgrading from a 900-series like for like is probably unnecessary. But for anyone on a 700 series or older this is a ridiculous performance jump.
Even so, 35% odd increase over a 980Ti is nothing to sneeze at.
If you use last gen cards as an example, yes, this is mid end. 980 vs 980 Ti. Ti\Titan dies are both much larger and much faster than 980’s.
Considering we know GP102 is coming, this is indeed a midrange card. And its a midrange card being sold for what was high end pricing.
680, 780, 980 were not mid-end cards. Just like the 290X wasn’t a mid-end card for AMD. Just like the 1080 isn’t a mid-end card. That’s what the 770, 970, 1070 are for NVIDIA and the 380 is for AMD.
Just because they aren’t the top card for their series doesn’t make them mid-end. It just makes the top-end cards like the 980Ti, Titan X, Fury X etc the highest of high end.
It’s like saying a Lamborghini Huracan is a mid-end car. Sure, it’s not the most expensive car they make, but it’s not mid-end.
I would argue (since any card below these are just rebrands). 950\960 (Low End), 970\980 (Mid), 980 Ti\Titan X (High End). I know Nvidia\AMD technically call Ti\X enthusiast level (and 970\980 high), but if your pricing these cards as high as enthusiast level, I’m still calling bullshit.
I agree with you Camm. This card is replacing the 980. I wouldn’t pay more than $AU700.
A really great card but given I couldn’t help myself and bought 2 Titan X’s less than a year ago I’ll leave it for 2 or 3 more generations. What I will be looking for is a card that can handle max settings without the need to SLI which is a bit of a joke given how many current Triple A titles can really use it.
Max with 1080p or 4K? The card can max for sure, but obviously starts to struggle with intense games at 4k. Doesnt do too badly, mind you, we’re just not quite there yet. Decent for the starter of the generation, hooking up in SLI would be if you want seamless max 4k.
Sorry, I meant 4K as obviously current some mid range cards can handle 1080p. I’m just not on board with SLI as much as I hoped I would feel that the future will be single GPU beast cards like GTX 1080 etc.
At least its only the beginning of the series. 4K is taking some time to kick off, even on the TV side of things. For your purpose I would wait a bit, maybe snag something in 2017 when 4K is a bit more common practise than now where its still a bit of a niche.
I would be happy with 1440p, but thats because I havent upgraded in some time and am used to 1080p.
I don’t doubt that the card will be pricey, but all this discussion of local costs being ~$1500 etc is a moot point when the Australian prices haven’t even been confirmed yet.
Will it be expensive? Undoubtedly. Will it be $1500? I wouldn’t bet on it. I’m betting on it being a similar price to 980Tis when they launched. So around $1200.
Funny how Metro still makes it work hard. That game is a GPU hungry bitch
Really? I’m sure an article suggesting the exact opposite of that assumption a few weeks ago…
Shhh don’t say it allowed. It’s a touchy subject.
I’ve already been personally attacked by the person in question multiple times.
Of course I’m in the wrong for doubting him for being such an expert in the field. Also it’s definitely the retailers gouging us, Nvidia definitely hasn’t put up a huge MSRP despite the fact it’s been NDA’d and nobody is telling us what it actually is…