It started with the arrival of Nvidia's GTX Titan X earlier this year - a new GPU with an unprecedented amount of graphics power, backed up by a gargantuan 12GB of GDDR5 dedicated video memory. Meanwhile, behind the scenes, AMD was working on its own top-tier technology - the Fiji processor. It's the largest piece of graphics silicon the company has ever made, and it's backed by the inclusion of HBM - next-gen graphics memory that sees memory modules stacked one on top of the other, and arranged around the processor, reducing latency and increasing bandwidth.
The expected arrival of the product that would eventually be called the Radeon R9 Fury X spurred Nvidia into action, pre-emptively launching the GTX 980 Ti - a beautiful example of why competition is so important in this market. The newer Nvidia card featured fewer CUDA processing cores and slashed memory allocation down to a 'mere' 6GB, but remarkably, it appeared to lose next to nothing of its overall performance compared to the Titan X. But the price was certainly more amenable, dropping from $999 down to $650.
When the liquid-cooled R9 Fury X eventually arrived, AMD called it the most powerful GPU on the market - but alas, our benchmarks (and indeed everyone else's) painted a different picture. Both Titan X and the GTX 980 Ti were more powerful, and the inclusion of HBM appeared to yield no knockout blow compared to the Nvidia offerings. However, the closer integration of memory and processor produced one major difference - a much smaller board. This paved the way for the arrival of the R9 Nano, the least powerful GPU in the grouping we have in this extreme price-range, but a unique product in its own right. It's absolutely tiny, cooler and quieter than any of the other air-cooled cards in this category, making it highly tempting for those running smaller form-factor PCs.
The category is completed with an air-cooled version of the Fury X, a more traditionally sized product - and also the cheapest. It offers slightly higher levels of performance than the Nano, but with its $550 price-point, it's actually aimed at competing with the $500 GTX 980. So how do all these expensive cards shake out once we break out the Digital Foundry benchmarks and gameplay tests? Let's take a look, but before we do, some quick tips on gaming with these powerful cards:
- Choosing between GTX 980 Ti and Titan X is actually no choice at all - the two products are essentially interchangeable, but one of them costs a third less. Buy the GTX 980 Ti if you absolutely must have an ultra-powerful GPU right now.
- AMD's Fury line is highly competitive with Nvidia at 4K, but the green team's cards are more flexible overall, with better performance at lower resolutions.
- At 4K, anything more than post-process anti-aliasing really is a waste of time - and GPU resources.
- We recommend pairing this level of graphics hardware with an Intel Core i7 processor, and we strongly recommend that you overclock it.
- That said, if you are targeting 4K resolution, frame-rates will automatically be lower, meaning less stress for the CPU. In this scenario, a Core i5 should do fine.
|Crysis 3 1440p60/V-Sync Gameplay||GTX 980 Ti||GTX Titan X||R9 Fury X||R9 Fury||R9 Nano|
|Dropped Frames (from 18650 total)||624 (3.35%)||477 (2.56%)||1141 (6.12%)||2084 (11.17%)||2680 (14.37%)|
Alternative video analyses:
- Crysis 3 1440p gameplay - R9 Fury vs Fury X vs GTX 980
- Crysis 3 1440p gameplay - R9 Nano vs Fury X vs Fury
Buy the graphics cards tested on this page from Amazon with free shipping:
- Nvidia GeForce GTX 980 Ti 6GB [?]
- Nvidia Titan X 12GB [?]
- AMD Radeon R9 Fury X 4GB [?]
- AMD Radeon R9 Fury 4GB [?]
- AMD Radeon R9 Nano 4GB [?]
The Crysis 3 challenge caps top-end performance at 60fps at 1440p via v-sync, and can be particularly taxing in terms of momentary lapses in performance, signified by the lowest frame-rate metric. What's clear here is that the 'ancient' 2013 CryEngine release is still an extremely tough work-out, even for the latest graphics hardware. We should still expect to see frame-rate drops from the tough CPU workout (typically this can account for a 0.5 per cent dip in performance assuming no GPU limits) so it's clear that while both AMD and Nvidia tout these cards for their 4K performance, 1440p can still cause issues even for these immensely powerful cards. The challenge also reveals that the GTX 980 Ti and the Titan X are the more consistent, stable performers.
So how does that change when we look at more games running completely unlocked with v-sync disabled? This allows us to test top-end performance and throws up some interesting - and some might say confusing - results. What becomes clear is that the GTX 980 Ti and the Titan X generally only show a 25 per cent improvement over the less capable GTX 980 at 1080p, but the situation is much more of a concern on the Fiji products, where the 980 can actually run a few of our games more quickly than the AMD flagship. Suffice to say that while we have trouble recommending any of these cards for those with full HD displays, if you're going to do it, make sure you're using an Nvidia card.
The R9 Nano's performance here is interesting though, bearing in mind that its small form-factor status makes it entirely unique among the cards tested. It offers virtually all of the air-cooled Fury's performance, despite running at lower clock-speeds. Our theory here is that the vast Fiji silicon is under-utilised by the modest resolution, so more of its power budget can be used to push the card faster.
|1920x1080 (1080p)||Titan X||GTX 980 Ti||R9 Fury X||R9 Fury||R9 Nano|
|The Witcher 3, Ultra, HairWorks Off, Custom AA||84.4||82.6||67.1||62.8||61.7|
|COD Advanced Warfare, Console Settings, FXAA||159.9||156.8||116.8||112.0||113.7|
|Battlefield 4, Ultra, 4x MSAA||112.4||109.9||86.9||81.3||81.5|
|Crysis 3, Very High, SMAA||105.2||104.0||94.3||88.5||84.3|
|Assassin's Creed Unity, Ultra High, FXAA||74.7||74.4||62.8||57.3||56.8|
|Far Cry 4, Ultra, SMAA||101.4||101.2||82.0||81.5||79.4|
|Ryse: Son of Rome, High, SMAA||99.2||97.8||86.7||84.1||80.3|
|Shadow of Mordor, Ultra, High Textures, FXAA||119.0||118.5||110.2||102.7||101.2|
|Tomb Raider, Ultimate, FXAA||150.3||150.1||127.4||114.1||113.8|
Based on the page-topping Crysis 3 gameplay work-out, there's a compelling argument that 2560x1440 resolution - not 4K - is the more natural home for this new breed of graphics hardware, and the reasoning here is pretty straightforward: if we get close to a locked 60fps on top-end presets, we are achieving a strong balance between resolution and image quality. We may fall a little short of the complete 60Hz lock, but we have plenty of leeway for settings tweaks to get frame-rates where we want them without too many compromises to the visuals.
However, the Fury cards still fall short of the Nvidia offerings, even as we scale up in terms of resolution. That said, what's clear is that at least the gap is narrowing. Indeed, the Fury X even scores a couple of wins over the GTX 980 Ti - there's a decent 4.6 per cent increase on Far Cry 4, but Ryse is only 0.3 per cent faster - basically, the in-game experience is identical. Shadow of Mordor, Assassin's Creed Unity and Crysis 3 show the Fury X dropping back, with GTX 980 Ti around four to six points clear over the AMD challenger. And the situation gets worse with games that tend to favour Nvidia's cards, as you might expect: The Witcher 3, Battlefield 4 and Advanced Warfare are all at least 17 per cent faster on the GTX 980 Ti.
Again, the R9 Nano is at the bottom of the table, but we should stress once more that it is the most power-efficient card of the lot, and based on our more in-depth review, it's still faster than the GTX 980 and the R9 390X. That's not bad considering its ultra-mini chassis configuration.
|2560x1440 (1440p)||Titan X||GTX 980 Ti||R9 Fury X||R9 Fury||R9 Nano|
|The Witcher 3, Ultra, HairWorks Off, Custom AA||63.3||61.7||52.4||48.8||46.6|
|COD Advanced Warfare, Console Settings, FXAA||123.2||121.3||103.0||99.6||98.5|
|Battlefield 4, Ultra, 4x MSAA||76.1||75.0||62.2||58.2||57.4|
|Crysis 3, Very High, SMAA||68.0||66.2||63.4||58.0||54.9|
|Assassin's Creed Unity, Ultra High, FXAA||49.6||48.3||45.8||40.1||39.4|
|Far Cry 4, Ultra, SMAA||77.0||75.4||78.9||73.4||69.3|
|Ryse: Son of Rome, High, SMAA||72.8||71.2||71.4||64.8||62.4|
|Shadow of Mordor, Ultra, High Textures, FXAA||87.2||87.2||82.5||77.7||75.8|
|Tomb Raider, Ultimate, FXAA||101.9||99.2||91.6||81.8||79.7|
Before we go on to discuss 4K performance, we feel that prospective Fury X or GTX 980 Ti owners should check out our piece on ultra-wide 21:9 aspect ratio gaming, where we benchmark both cards at 3440x1440 resolution. The plain and simple fact of the matter is that driving a 4K screen is equivalent to running four 1080p displays simultaneously, and the impact to performance can be drastic at extreme settings - which is why we drop things down a notch here, which helps to varying degrees.
If you are running a 4K screen, Titan X, GTX 980 Ti and Fury X offer very similar performance across the board. Battlefield 4 favours Nvidia while Far Cry 4 is very AMD-friendly, but elsewhere the experience offered by all three cards is very similar. What's interesting here is that the three Fiji-based products really are pushed to their limits, and we see the biggest differential between the cards here. That said, there's still not much in it - the air-cooled Fury offers around 93 per cent of the Fury X's power and can be overclocked to match, while the tiny Nano drops down again by a similar margin.
It's great that this new tier of GPU is able to make 4K gaming a viable experience on a great many modern games, but some might find that the drop in quality settings when you're spending so much money on a GPU is a little too much. And certainly, from our perspective, the 27-inch and 32-inch 4K displays we've tested are immensely pixel-dense, to the point where you can't help but wonder how much GPU power is being expended on detail we can't fully appreciate.
|3840x2160 (4K)||Titan X||GTX 980 Ti||R9 Fury X||R9 Fury||R9 Nano|
|The Witcher 3, High, HairWorks Off, Custom AA||37.5||36.9||36.2||33.1||30.9|
|COD Advanced Warfare, Console Settings, FXAA||90.8||86.9||85.3||80.8||75.9|
|Battlefield 4, High, Post-AA||61.3||61.0||51.0||47.7||46.0|
|Crysis 3, High, SMAA||52.4||52.5||49.2||45.6||42.2|
|Assassin's Creed Unity, Very High, FXAA||27.4||26.5||25.3||23.0||22.1|
|Far Cry 4, Very High, SMAA||46.7||47.1||50.5||47.3||43.4|
|Ryse: Son of Rome, Normal, SMAA||42.2||41.7||44.0||40.9||38.2|
|Shadow of Mordor, High, High Textures, FXAA||54.8||54.8||55.5||51.6||48.3|
|Tomb Raider, Ultra, FXAA||64.6||61.3||63.9||55.5||53.2|
There's no prizes for guessing which of the cards tested is the most power-efficient. The R9 Nano is designed for small form-factor PCs, meaning that it has an operating power-draw of around 175W. However, we suspect that this may well be higher when the system is really under stress, as it is when we run our peak power-draw and overclock stability tests. Total system draw from the wall came in at 303W - impressive compared to the competition, with the Fury cards around 100W more intensive. Nvidia's top-tier cards both have a small but tangible power efficiency advantage over the higher-end Fiji cards, but generally speaking, the green team's advantage in this area is significantly lower than it is elsewhere within the market.
One area where Nvidia hardware remains unassailable is in terms of overclocking. Both the GTX 980 Ti and the Titan X can easily be pushed to 1400MHz on the core clock, with some higher-end custom designs even reaching 1500MHz. Meanwhile, the two higher-end Furies will struggle to reach stability at 1130-1150MHz. You'll be lucky to get a seven to eight per cent increase in performance on AMD, while that can be doubled on the Nvidia side of the equation. Overclocking isn't entirely key to the top-tier experience, but it's certainly a function that enthusiasts want when they're spending upwards of $650 on a graphics card.
We actually managed to get an additional 11 per cent of raw performance out of the small form factor R9 Nano, but it loses its power efficiency big-time, the fan starts to spin up noticeably - and there's a strong argument that in real-life conditions, you will be thermally limited in a standard small form-factor case, limiting overclocking potential in the first place.
|R9 Nano||R9 Fury||R9 Fury X||GTX 980 Ti||GTX Titan X|
|Peak System Power Draw||303W||392W||407W||375W||379W|
The best graphics card over £400/$500: GeForce GTX 980 Ti
Honourable mention: Radeon R9 Nano
The GTX 980 Ti may not be the fastest card in the collection, but what it loses in terms of performance to the top-end Titan X is easily forgivable bearing in mind that it is $350 cheaper, and can easily be overclocked to move way beyond stock performance on Nvidia's thousand dollar offering. AMD's Fury X is a worthy contender - but only really at 4K resolutions, and even there, any performance advantage is lost once the GTX 980 Ti's overclocking capabilities are factored in. In short, if you want the best balance of price vs performance in this category, the GTX 980 Ti is unmatched.
But with that said, the R9 Nano is worthy of consideration if you are building a small form-factor PC - if you can stomach the colossal $650 price-tag. Nvidia simply doesn't have a card offering the same kind of performance in such a tiny chassis - but we offer this recommendation with a couple of caveats. Firstly, its 1080p turn-out isn't great, to the point where even the GTX 970 at half the price can be competitive on some games. And secondly, many mini-ITX SFF cases can hold a GTX 980 Ti or a Fury X. Nvidia's card certainly won't be as quiet in this scenario, but it'll definitely be faster. We suspect that the Nano is the first in a new breed of GPU, and where that product leads, smaller gaming cases will follow.
Will you support the Digital Foundry team?
Digital Foundry specialises in technical analysis of gaming hardware and software, using state-of-the-art capture systems and bespoke software to show you how well games and hardware run, visualising precisely what they're capable of. In order to show you what 4K gaming actually looks like we needed to build our own platform to supply high quality 4K video for offline viewing. So we did.
Our videos are multi-gigabyte files and we've chosen a high quality provider to ensure fast downloads. However, that bandwidth isn't free and so we charge a small monthly subscription fee of €5. We think it's a small price to pay for unlimited access to top-tier quality encodes of our content. Thank you.Support Digital Foundry