AMD Radeon R9 Nano vs. NVIDIA GeForce GTX 970: Which is better for your mini-ITX system?

Which card for your next small form factor system?

We’re now at that point of limbo in the graphics product cycle where consumers are already looking ahead to the next generation of cards from both red and green camps. There’s a lot to be excited about for the upcoming batch of GPUs, and the news pipeline has been abuzz with talk about second-generation High Bandwidth Memory (HBM2), GDDR5X memory, and the promise of even lower power consumption. Of course, this also means that we have more questions than answers, given that concrete details are still few and far between.

So instead of just sitting around twiddling our thumbs and looking up at the sky for Polaris, we’d like to turn our attention to the current slate of cards. The AMD Radeon R9 Nano was first released in late August, about two months after it was teased at AMD’s E3 launch event.

The compact card stood out for a lot of reasons, not least because it promised a fully-enabled Fiji GPU in a card that measured just six inches long. In comparison, the far larger Radeon R9 Fury featured a cutdown Fiji chip with 12.5% fewer stream processors and texture units.

Furthermore, this is a time when small and compact form factors are becoming increasingly attractive. From diminutive powerhouses like the MSI Vortex to innovative ways to combine mobility and performance with external graphics enclosures and technologies, companies appear to be tapping into a very real demand for space-efficient products that pack a punch.

If you’re looking to build a powerful mini-ITX system, you’d definitely have considered the Radeon R9 Nano. After all, it is so compact that it will fit onto a mini-ITX motherboard with absolutely zero overhang.

At US$499 (after a US$150 price drop in January), the Nano isn’t cheap, but that’s the price you pay for its sleek design. Its price – close to a thousand dollars in local stores – would also appear to set it up to go head-to-head with the NVIDIA GeForce GTX 980, which currently retails for around the same, but we’re actually more interested in how it compares with a card one rung below, the GeForce GTX 970.

That’s because we consider the US$299 GeForce GTX 970 more likely to end up in a small form factor system than its bigger brother. While there are definitely mini-ITX cases with room for the GeForce GTX 980, the 970’s smaller footprint is less restrictive and gives you a wider range of chassis options. It is also a powerful card in its own right, and NVIDIA’s add-in card partners have even released custom-designed mini-ITX versions of it, like the ASUS GeForce GTX 970 DirectCU Mini.

Before we proceed, here's a snapshot of the reference specifications of the Radeon R9 Nano and GeForce GTX 970:

What’s under the hood?

Because we haven’t had the opportunity to do a full review of the Radeon R9 Nano yet, we’ll also be taking the chance to briefly detail the Fiji chip on the Nano itself, in addition to the design decisions that went into creating a card for mini-ITX systems.

As we mentioned earlier, the Nano’s size actually belies the fact that it features a fully-enabled Fiji GPU. This means that all 64 Compute Units (CUs) are enabled, so the card has the same 4,096 stream processors and 256 texture units as the Radeon R9 Fury X. In fact, the only real differences between the two cards, other than the size of course, are the targeted power envelope and clock speeds.

While the Radeon R9 Fury X is officially rated at 275 watts, AMD has set the Nano’s power limit on paper to a mere 175 watts. To do this, AMD has relied on a process of chip binning and a redesigned power circuit to both reduce power consumption and size. Ironically, the Fiji chips going into the Nano are the best and most power-efficient, and AMD has essentially created a non-flagship product using the best Fiji GPUs, from a power standpoint at least. As a result, the Nano requires just a single 8-pin PCIe connector to power it.

The Nano has a rated clock speed of up to 1,000MHz, compared to the Fury X’s 1,050MHz. But even then, it’s unlikely that the card will even be able to maintain that high a speed because of power throttling to keep within the specified power envelope, which is why you won’t see Fury X levels of performance. In addition, the card’s size means that there have had to be some compromises in the power delivery circuit. For instance, AMD has gotten rid of some VRM power phases – it has four versus the six on the Radeon R9 Fury X –  to accommodate the card’s intended size. You can find more basic details of the Radeon R9 Nano here.

But all things considered, the Radeon R9 Nano amounts to quite a luxurious take on a rather niche demand. There are mini-ITX versions of the NVIDIA GeForce GTX 970 available, and it also costs less and consumes less power. However, the Nano was created expressly for compact systems, and that is one advantage it holds over a generic GeForce GTX 970.

In the next section, we’ll take a closer look at the performance benchmarks and see how both cards perform against each other.

Test Setup & Performance Benchmarks

Test Setup

We tested each card on the following graphics testbed, which has now been updated to use Windows 10:

• Intel Core i7-5930K
• ASUS X99-Pro (Intel X99 chipset) motherboard
• 4 x 4GB Corsair Vengeance LPX DDR4-2133 (Auto timings: CAS 15-15-15-36)
• Samsung SSD 840 Pro 256GB SATA 6Gbps solid state drive (OS)
• Western Digital Caviar Black 1TB SATA 6Gbps hard drive (benchmark + games)
• Windows 10 Professional 64-bit
• Intel INF 10.1.1.14

In addition to the NVIDIA GeForce GTX 970 and AMD Radeon R9 Nano, we also tested the Radeon R9 Fury and R9 390X to better flesh out the comparisons. Here’s a list of the cards we used, along with their driver versions (all custom cards have been taken down to reference speeds):

• Sapphire Radeon R9 Nano (Radeon Software Crimson Edition 16.3.1)
• Palit GeForce GTX 970 JetStream (ForceWare 364.51)
• ASUS Strix Radeon R9 Fury (Radeon Software Crimson Edition 16.3.1)
• ASUS Strix Radeon R9 390X (Radeon Software Crimson Edition 16.3.1)

The following tables outline the specifications of the above cards, along with the reference speeds:

[hwzcompare]

[products=547151,514922,515698,547195]

[width=200]

[caption=Tested card specifications]

[showprices=1]

[/hwzcompare]

As much as we wanted to get hold of truly compact GeForce GTX 970 graphics cards like the ASUS GTX970-DCMOC-4GD5 and the Gigabyte GV-N970IXOC-4GD, we waited patiently for weeks, but allocations were just not available for testing purposes at this point of time. In place of these cards, we've decided to test a standard sized GeForce GTX 970 in the form of the Palit GeForce GTX 970 JetStream 4GB GDDR5 to depict the average performance you can expect out this GPU class.

Since the Sapphire Radeon R9 Nano was basically a reference card with reference clock speeds, we've also dialled down the comparison cards to operate at their respective GPU reference clock speeds to present a fairer comparison.

Benchmarks

We ran a full suite of performance benchmarks, including a handful of the latest titles. Some of these also take advantage of DirectX 12, so we were also able to get a look at how the cards play with Microsoft’s latest API.

Here’s a full list of the benchmarks we used:

• Futuremark 3DMark (2013)
• Crysis 3
• Middle-earth: Shadow of Mordor
• Ashes of the Singularity
• Hitman
• Tom Clancy’s The Division

We used the Fire Strike Extreme test in 3DMark (2013) for our power and temperature tests.

Futuremark 3DMark (2013)

The Fiji-based Radeon R9 Nano did quite well against the GeForce GTX 970 in all three 3DMark benchmarks. In 3DMark Fire Strike for instance, the Nano pulled ahead of its NVIDIA counterpart by just over 20%. This was a lead it went on to widen in the more demanding Fire Strike Extreme and Ultra benchmarks, where it held a 23% and 50% advantage over the GeForce GTX 970 respectively.

The performance difference here is actually to be expected, given the large bandwidth advantage that the Nano’s 4GB of HBM gives it over the GeForce GTX 970 4GB of GDDR5 memory. Like the Radeon R9 Fury X, the Nano boasts a whopping 512GB/s of memory bandwidth, compared to the GeForce GTX 970’s 224GB/s.

Furthermore, even though the Nano has been tuned for power efficiency with a TDP of just 175 watts, that is still ahead of the GeForce GTX 970’s 145 watts, which is also the cheaper card here at just US$299. These are still cards that are in relatively different price brackets, and as we’ll see in the following benchmarks, the combination of differences in memory bandwidth, power draw, and shader-bound performance continue to give the Nano the edge over the GeForce GTX 970.

Crysis 3

In Crysis 3, the performance gap between the GeForce GTX 970 and the Radeon R9 Nano widened as the display resolution and anti-aliasing settings became more demanding. At the most trying settings – 2,560 x 1,600 pixels and 8x AA – the Nano was almost 25% faster than the NVIDIA card.

The Nano also continued to turn out very similar performance to the Radeon R9 Fury, which costs the same at US$499. This puts the Nano in quite a favorable light as the Fury is considerably larger and definitely would not be as compelling a choice for a mini-ITX system. Between the two – one has a cutdown Fiji GPU while the other has been throttled for power – AMD has taken two very different routes to achieve close to the same level of performance, while also managing to create two cards that will appeal to different groups of users at the same price.

Middle-earth: Shadow of Mordor

Memory bandwidth plays a big factor in our Middle-earth: Shadow of Mordor benchmark, where we deliberately scale up the resolution to ultra-high settings to test the cards’ ability to deal with such scenarios. At a resolution of 2,560 x 1,600 pixels, which is also the resolution closest to what most gamers are likely to play at, the Radeon R9 Nano proved to be around 31% faster than the GeForce GTX 970. It even managed 30.46fps at a resolution of 3,840 x 2,400 pixels, while the GeForce GTX 970 pushed closer to 20fps, arguably the bare minimum required to have a playable experience of any sort.

Clearly, if you want to game upward of 1440p, the Radeon R9 Nano is your go-to card over the GeForce GTX 970.

Ashes of the Singularity

Ashes of the Singularity is a new addition to our stable of benchmark titles, and we ran the same settings in both DirectX 11 and 12 modes to assess the difference between the AMD and NVIDIA cards. DirectX 12 is a low-overhead API intended to confer various benefits like better utilization of multi-threaded CPUs, improved management of GPU memory resources, and support for asynchronous computing. In fact, Ashes is the first PC game to make extensive use of asynchronous compute, which boosts performance by allowing the GPU to perform both graphics and compute workloads simultaneously.

Asynchronous Compute Engines (ACE) are built into AMD’s Graphics Core Next (GCN) architecture, which is why the Fiji-based Radeon R9 Nano, Fury, and even the Hawaii-based R9 390X exhibit improved performance when moving from DirectX 11 to 12. For the Nano, this was as much as a 21% increase at High settings and a 1600p resolution. On the other hand, the GeForce GTX 970 went the other way, and we observed a minuscule drop in performance in DirectX 12 compared to DirectX 11. NVIDIA’s Maxwell architecture doesn’t support asynchronous compute at the hardware level to the extent that AMD’s GCN does, which is why it failed to benefit from the new low-level API.

And considering that the Nitrous Engine that Ashes is based on reportedly uses asynchronous compute capabilities for up to 30% of a given frame’s workload, that’s quite a significant performance boost that NVIDIA is failing to tap.

Hitman

Note 1: We were unable to get the GeForce GTX 970 to work with Hitman on DirectX 12 (the benchmark would simply freeze mid-way), which is why the values for the NVIDIA card are nil in the relevant graph.

Note 2: One other issue we noted was that AMD’s Radeon Software Crimson Edition 16.3 drivers exhibited an odd problem where the frame rate would be capped to the 60Hz refresh rate in DirectX 12 for certain cards and test runs. The problem was fixed when we updated to the 16.3.1 beta drivers.

The Radeon R9 Nano once again appeared to benefit from the move to DirectX 12. However, the performance improvement in this case was smaller than that observed in Ashes of the Singularity. The most significant leap was when we ran the benchmark at a 1080p resolution and High detail settings, where the Nano managed a 13% improvement.

In DirectX 11, the Nano also demonstrated quite a large lead over the GeForce GTX 970, as much as a 33% advantage at 1080p and Ultra settings. This bodes well for AMD’s mini-ITX card, seeing as Hitman is one of the latest PC titles to hit the market.

Unfortunately, we were unable to get figures for the GeForce GTX 970, so we can’t comment on if or how the card would respond to DirectX 12. But given the issues we experienced with the AMD cards that were fixed with a driver update, it’s possible that both camps will benefit from future driver upgrades further down the line.

Tom Clancy’s The Division

The results we observed in The Division reminded us of what we saw in Crysis 3, where the performance differential between the Radeon R9 Nano and GeForce GTX 970 widened as the benchmark became more graphically demanding. The GeForce GTX 970 performed decently against the Nano at 1080p resolutions, but once we moved up to 1600p, the gap grew to a hefty 24%.

The GeForce GTX 970 is definitely still capable of giving you a playable experience in the latest titles, but the Radeon R9 Nano has a not-so-trivial performance advantage here.

Temperature and Power Consumption

In the temperature department, both the Radeon R9 Nano and GeForce GTX 970 returned very similar results, with the Hawaii-based Radeon R9 390X serving to highlight how cool these cards really are. Heat is a primary concern in mini-ITX systems where cooling options and airflow are more limited, but you shouldn’t have an issue with either of these cards.

When it came to power consumption, the two cards also displayed a relatively similar performance profile. As it turns out, the Radeon R9 Nano actually consumed slightly more power at the wall outlet, which is probably to be expected given that it is still rated to draw more power than the NVIDIA card (175 watts vs. 145 watts). And compared to the other two AMD cards we tested here, the Nano and GeForce GTX 970 are practically sipping power.

Overclocking

We didn’t expect the Radeon R9 Nano to overclock that well because of its power limitations, and that was borne out in our results. We managed to achieve a top speed of 1,075MHz, a modest 75MHz overclock over its 1,000MHz stock speed. On the other hand, the GeForce GTX 970 did quite a bit better, and we could take it up to 1,263MHz, which translated into a boost clock of 1,414MHz. That’s actually a 200MHz overclock over its 1,050MHz stock speeds, or around a 110MHz boost over the 1,153MHz factory overclock.

In 3DMark Fire Strike, the respective overclocks resulted in a small 3% gain for the Nano, but nearly a 10% boost for the GeForce GTX 970. But unfortunately for the NVIDIA card, mini-ITX system builders aren’t likely to be looking to squeeze out extreme overclocks from their system, so a win here doesn’t do much for it in our overall considerations.

The mini-ITX card wins the day

Ultimately, it’s pretty clear that the Radeon R9 Nano trumps the GeForce GTX 970 in almost every area when it comes to putting together a mini-ITX system. The only thing the GeForce GTX 970 has going for it is its price – it is considerably cheaper than the Nano, which might appeal to casual gamers who want to build a more cost-effective small form factor PC. Because the GeForce GTX 970 performs more favorably relative to the Nano at the less graphically demanding settings (typically 1080p resolution), it also looks like the attractiveness of the NVIDIA card depends on the resolution and settings you intend to game at, and by implication, the resolution of your monitor.

Here’s a table with the prices of the Radeon R9 Nano and a competing mini-ITX version of the GeForce GTX 970 to give you a better picture:

There is also a mini-ITX version of the card from ASUS, the GeForce GTX 970 DirectCU Mini. We’ve yet to get word on local pricing and availability of the card, but we’d expect it to cost around the same as the Gigabyte card, if not slightly more.

As it turns out, the Radeon R9 Nano has around a S$300 premium over the GeForce GTX 970. Is its better performance worth the extra dollars? If the GeForce GTX 970 were just released last month, we’d be more inclined to argue that you could get away with saving the money and building a system that would still deliver playable frame rates in most games. However, the GeForce GTX 970 first debuted in 2014, and we are now on the cusp of the release of Pascal, NVIDIA’s next-generation GPU architecture. If that wasn’t enough to convince you of the GeForce GTX 970’s diminishing value moving forward, also remember that so far, its performance in DirectX 12 isn’t encouraging.

On the other hand, the Radeon R9 Nano just became available in late August, and its low power consumption, compact size, and stellar performance make it a very attractive option for a powerful mini-ITX system. Polaris may be releasing in the middle of the year, but that will be a mobile-focused GPU, and there’s no word when AMD will be releasing a follow-up to the Nano (we expect that to happen sometime in Q3 this year). As far as graphics cards go, the Nano is in a class of its own, having been expressly designed for compact systems and low power consumption.

Then there is the huge tilt in AMD’s favor when it comes to DirectX 12 performance, or even getting the game to work with the new API at all. While it might be premature to conclude based on just a single game (Ashes of the Singularity) that the Radeon R9 Nano (and other AMD cards with ACEs by extension) benefits more than the GeForce GTX 970 from the shift to DirectX 12, it is certainly looking that way at this point. Of course, we’ll still need to run further comparisons on more games as they become available to be really sure.

If you’re looking to build a small and powerful PC today, the Radeon R9 Nano is probably your best bet. It also certainly doesn’t hurt that the card is rather easy on the eyes with its sleek design. If you have a smaller budget and you don't intend to upgrade your 1080p resolution monitor anytime soon, a GeForce GTX 970 is still a good option, just that it's not class leading.

Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.