AMD launches the compact Radeon R9 Nano with high-end performance

[Scorpus]

AMD has officially unveiled the Radeon R9 Nano today, the company's new performance graphics card that's just six inches long. Featuring a fully-unlocked Fiji GPU, the R9 Nano might just be the perfect card for small form factor and mini-ITX gaming PCs.

Lets get the specifications of the R9 Nano out of the way first. The card features a 28nm Fiji GPU with 4,096 stream processors, the same number as the company's flagship Fury X, alongside 256 texture units and 64 ROPs. There's also 4 GB of HBM at 500 MHz, providing 512 GB/s of memory bandwidth.

The R9 Nano's core clock speed is rated as "up to 1000 MHz", with a rated TDP of 175 W, meaning the card only requires a single 8-pin PCIe power connector. In total, AMD calculates that the R9 Nano can deliver 8.19 TFLOPs of compute performance.

In AMD's press materials for the R9 Nano, the company extensively compares the new card to the Hawaii-based Radeon R9 290X. According to their data, the R9 Nano is "up to" 30% faster than the R9 290X while being 40% shorter, 20°C cooler with an operating temperature target of 75°C, 30% less power hungry, and 16 dBA quieter.

Putting all of these figures together, and AMD says the R9 Nano has 2X the performance per watt and 2X the performance density of the R9 290X. These are pretty impressive figures, but there are some critically important things to note about this card.

For starters, the R9 Nano's specs are nearly identical to that of the R9 Fury X. Same core configuration, same frame buffer size and bandwidth, and a rated maximum clock speed just 50 MHz less than the Fury X. While AMD doesn't directly compare the Nano to the Fury X in their release notes, they do list the Nano's peak performance as just slightly below that of a Fury X.

However there are some key differences between the Nano and the Fury X that make this card rather unusual. The Nano is a lot smaller than the Fury X, and doesn't feature the closed-loop water-cooler, with AMD opting instead for a dual vapor chamber and thermal piped heatsink with a single fan. This cooling solution is apparently quite effective, despite rated performance close to that of the Fury X.

Critically, the R9 Nano's TDP is 36% lower than the Fury X, yet performance is supposed to be quite close between the two cards. Unless AMD has magically been able to shave 100W off the operating power of their largest Fiji GPU, something else has changed in the Nano's design.

AMD hasn't stated specifically what they are doing in the Nano to keep power consumption so low, but there's a fairly simple explanation: the R9 Nano has been capped to run under 175W through AMD's PowerTune technology. This means that while the Fury X will run at a full 1,050 MHz core clock during workloads, the Nano will mostly sit at a much lower clock speed and core voltages.

It's not clear what the Radeon R9 Nano's clock speeds will be during most gaming workloads, but there's a possibility that we'll see clock speed spikes as high as 1,000 MHz with the card's typical frequency sitting several hundred megahertz lower.

Scott Wasson from The Tech Report also pointed out that AMD's game benchmarks using the R9 Nano are specifically geared towards loading the shader cores, rather than simulating settings gamers actually use. Every benchmark is run with no anisotropic filtering (0xAF) and minimal anti-aliasing, instead focusing on high resolutions and high shader effects.

These tests are designed to favor the huge shader core array of the R9 Nano when placed up against competitors such as the Nvidia GeForce GTX 970 and even AMD's own Radeon R9 290X. It will be interesting to see how the card performs in real-world benchmarks when we get our hands on a review sample.

The R9 Nano will go on sale on September 10th, with a retail price of $649. This places the card at the same price point as the Fury X and the GeForce GTX 980 Ti, although we expect performance to be lower than both these cards due to the power and thermal constraints of the Nano's design.

This graphics card launch is a particularly interesting one, so stay tuned for a full review of the R9 Nano in the coming weeks.

Permalink to story.

https://www.techspot.com/news/61907-amd-launches-compact-radeon-r9-nano-high-end.html

 

[deemon]

Sorry AMD! I really really really wanted to buy this card for my new Skylake ITX build. But your performance/price is not nearly good enough. I would have bought this card for +20% (maybe even up to +30% if your hyped performance over 970 were been actually true) of nvidia GTX 970 price, but this +100% now... is just absurd. Sorry.

 

[madboyv1]

Yeah, $650 is really hard to swallow (almost choked on my morning coffee) for what this card is... at least on paper. I still await with abated breath for a 3rd party review as I never trust the manufacturer's numbers, but on price alone I don't think I would be able to justify it. Many performance mini-ITX cases have the space for a full size card or can be modded to fit one, so at that price you ALMOST might as well get a 980ti.

 

[VitalyT]

This new spec doesn't make any sense. According to official information, the only spec change from the Fury X is reducing the GPU clock from 1050Mhz to 1000Mhz, while keeping everything else exactly the same.

In the meantime, they say the power consumption drops from 275W to 175W and performance lowered to 70% of FuryX. Those two facts together just don't add up.

 

[Mieksr]

This will be amazing in a Micro ATX System

 

[SuperVeloce]

This new spec doesn't make any sense. According to official information, the only spec change from the Fury X is reducing the GPU clock from 1050Mhz to 1000Mhz, while keeping everything else exactly the same.

In the meantime, they say the power consumption drops from 275W to 175W and performance lowered to 70% of FuryX. Those two facts together just don't add up.

Because it is "up to 1000", look at the slides. That's how the turbo boost works. Because TDP is lower and there is no waterloop to help, this will usually run at much lower clocks (and voltage)

 

[CaptainTom]

Yeah, $650 is really hard to swallow (almost choked on my morning coffee) for what this card is... at least on paper. I still await with abated breath for a 3rd party review as I never trust the manufacturer's numbers, but on price alone I don't think I would be able to justify it. Many performance mini-ITX cases have the space for a full size card or can be modded to fit one, so at that price you ALMOST might as well get a 980ti.

Why was the Titan 3x as much as the 7970 for 30% more performance? Because they had no competition in that space for 6 months.

 

[mosu]

Yeah, $650 is really hard to swallow (almost choked on my morning coffee) for what this card is... at least on paper. I still await with abated breath for a 3rd party review as I never trust the manufacturer's numbers, but on price alone I don't think I would be able to justify it. Many performance mini-ITX cases have the space for a full size card or can be modded to fit one, so at that price you ALMOST might as well get a 980ti.

Yes, but you won't have the same DirectX 12 performance. It looks like every tech site stopped testing DirectX 12 performance waiting for Nvidia to improve whatever they can.No comment on that.

 

[madboyv1]

@mosu : DX12 is still in it's infancy, both in driver support and game support so to comment on DX12 performance would premature at this time. Out of box you're right, that AMD seems to have an upper hand in DX12 while Nvidia is still playing catch-up. The only issue is that many games that are relevant to people are still running DX11. As Windows 10 continues to roll out I hope to see more and more development for DX12 at a faster pace than DX10 or 11.

Why was the Titan 3x as much as the 7970 for 30% more performance? Because they had no competition in that space for 6 months.

The Titan was also touted (and proven) as the most powerful single GPU solution in the consumer market and was also designed to push the limits of the architecture. The retail price never went down; they simply replaced it with the Titan X at the same price point. There was also nothing like it on the market at the time either, as you said. It's hard (read: impossible) to say the same for the R9 Nano. There's no immediate wow factor for this card, besides its new architecture, that makes the R9 Nano a must have even in the itx build space where it is most relevant. Since many enthusiast ITX cases can handle full size cards anyways, this immediately makes the 980ti and the FuryX more palatable from a performance and economic standpoint.

Let’s go back to a form factor apples to apples. The closest thing from ANY camp form factor wise would be the GTX 970, which a number of third parties have made short versions of (Asus and Gigabyte immediately come to mind). On average the R9 Nano is around 2x more expensive, rated at approx. 20% more power, and using AMD's marketing benchmarks, should be up to 30% faster than the GTX 970. But what concerns me is how AMD got to that number. To wit:

Scott Wasson from The Tech Report also pointed out that AMD's game benchmarks using the R9 Nano are specifically geared towards loading the shader cores, rather than simulating settings gamers actually use. Every benchmark is run with no anisotropic filtering (0xAF) and minimal anti-aliasing, instead focusing on high resolutions and high shader effects.

This by itself invalidates these benchmarks as real world results, and there are plenty of unknowns at the moment on how this card actually does what it does considering R9 Fury’s wattage and configuration. If it has to aggressively throttle down to preserve that 175w, I would think that it will hurt the R9 Nano’s real world performance.

Still, I'll be waiting until third party reviews show up in the coming weeks. If at least for the sake of keeping competition alive I want this card to succeed at what it sets out to do, but that price is prohibitive considering what it promises, even moreso than the Titan/TitanX's pricing.

 

[misor]

Oh, dr. phil, please don't tell me that 'having nano equipment' means faster performance...

The R9 Nano will go on sale on September 10th, with a retail price of $649

on second note, dr. phil, I'll take my meds now and just wait for next gen NVidia gpu (~300$)...

 

[Evernessince]

Sorry AMD! I really really really wanted to buy this card for my new Skylake ITX build. But your performance/price is not nearly good enough. I would have bought this card for +20% (maybe even up to +30% if your hyped performance over 970 were been actually true) of nvidia GTX 970 price, but this +100% now... is just absurd. Sorry.

It's always funny to read people make posts like this without even seeing actual benchmarks. This card isn't meant for full desktops, hence why it's priced akin to the fury. It's like saying "I won't buy this new sweater because it isn't amazing in winter".

Yeah, $650 is really hard to swallow (almost choked on my morning coffee) for what this card is... at least on paper. I still await with abated breath for a 3rd party review as I never trust the manufacturer's numbers, but on price alone I don't think I would be able to justify it. Many performance mini-ITX cases have the space for a full size card or can be modded to fit one, so at that price you ALMOST might as well get a 980ti.

Sure they are SOME that can fit a 980Ti but your going to make serious sacrifice in terms of hard drives, disk drives, and other components like fans. You act like it's a trivial matter to find an ITX case that can fit a full sized graphics card but in reality they are few and far between. Those that do have to make sacrifices with what little space they have.

 

[hahahanoobs]

From what I've read you're more likely to see a frequency closer to 900MHz than 1000MHz while gaming. Even lower when bench marking. Now imagine how it will perform in a tight space such as a mini-ITX case or a hot room.

The card costs as much as it does because the premium materials used, HBM, and its performance/size ratio. If Fury X needed water and Fury Pro needed 2 or more fans, there is no way in heck Nano will come close to those in performance... unless it does, which would tell me the water cooling on the Fury X that did next to nothing for overclocking, was unnecessary. You can boast about ~50C load temps all you want, but what good are such low temps when the overclocking is pants? Rememeber, consumers are still buying air coolers over AIO water cooling. As a result, the only benefit to using water would be if you were running it in a hot environment such as a SFF chassis, and that is not always possible due to the thick rad and room needed for the tubing.

The $649 price is the nail in the coffin. AMD has to recoup their high costs to give us this card, and they thought the size, performance and HBM would be enough for you to part with your cash. They were dead wrong as the reactions have shown. AMD can lower the price, which would have to instantly come down by AT LEAST $100 resulting in AMD losing money they desperately need. Money AMD can't afford to lose. AMD stock was as high as $4.19 a year ago, and has fallen as low as $1.66, currently sitting at $1.79.

As people have mentioned, the short versions of the GTX 970 (I'm adding the GTX 950) are far better options. They are WAY more energy efficient and much more reasonably priced for the performance you get. That is all.

 

[madboyv1]

Sure they are SOME that can fit a 980Ti but your going to make serious sacrifice in terms of hard drives, disk drives, and other components like fans. You act like it's a trivial matter to find an ITX case that can fit a full sized graphics card but in reality they are few and far between. Those that do have to make sacrifices with what little space they have.

Just to name a few out in retail that should support a GTX980ti and being lazy and pulling from Newegg/memory:

AZZA CSAZ-103, BitFenix Phenom, BitFenix Prodigy, COOLER MASTER Elite 110, COOLER MASTER Elite 120, COOLER MASTER Elite 130, Corsair Obsidian Series 250D, Corsair Graphite 380T, COUGAR QBX, EVGA Hadron, Fractal Design Node, IN WIN 901, LIAN LI PC-Q08, LIAN LI PC-Q28, Lian LI PC-O5S, Phanteks Enthoo Evolv, RAIDMAX Atomic (tight), Rosewill Neutron, SilverStone RVZ01, SilverStone ML07, Thermaltake Core V1 (very tight), Xigmatek Aquila

That list of 22 cases includes both cube cases and SFF towers, and a lot fewer than expected require the removal of a drive tray as well. I don't think any of them require removing cooling fans. As far as prices are concerned, they cover the full spectrum of dirt cheap to high premium. Of what was available on newegg, there were not that many double pci slot mini-ITX cases that did not also support a double width card of 10.5" or greater. Most of the ones that did not support this configuration had one PCI slot or none.

ITX gaming builds have seen a hefty boon in the last couple years with manufacturers paying attention with making motherboards, graphics cards, and cases specifically with gaming in mind so it's MUCH easier to do it now than when I first did it in 2013.

 

[mrjgriffin]

Idk why people are bashing the price......the furyx basically equals a 980ti and this nano fury card is barely any slower than the fury x and considering the very nice materials used on this card and how little the power consumption is I'd say 650 is fair. I paid 650 for my gtx 980 a little bit back and if I bought this nano card I would technically have a better card in my system. price sounds fair to me. amd can have my money with this one. now they need to get their butts in gear and make a new processor that actually has motherboards that support pcie 3.0. there is actually ONE am3+ motherboard out there by asus that supports pcie 3.0 im pretty sure its a discontinued sabertooth board, but there is one. its hard to find though

 

[cliffordcooley]

It's always funny to read people make posts like this without even seeing actual benchmarks. This card isn't meant for full desktops, hence why it's priced akin to the fury. It's like saying "I won't buy this new sweater because it isn't amazing in winter".

Forgive me but the picture I'm looking at makes this a full height card. It is only shorter. I'm not sure I would have made that statement for a card that wasn't a half height card.

 

[Evernessince]

Just to name a few out in retail that should support a GTX980ti and being lazy and pulling from Newegg/memory:

AZZA CSAZ-103, BitFenix Phenom, BitFenix Prodigy, COOLER MASTER Elite 110, COOLER MASTER Elite 120, COOLER MASTER Elite 130, Corsair Obsidian Series 250D, Corsair Graphite 380T, COUGAR QBX, EVGA Hadron, Fractal Design Node, IN WIN 901, LIAN LI PC-Q08, LIAN LI PC-Q28, Lian LI PC-O5S, Phanteks Enthoo Evolv, RAIDMAX Atomic (tight), Rosewill Neutron, SilverStone RVZ01, SilverStone ML07, Thermaltake Core V1 (very tight), Xigmatek Aquila

That list of 22 cases includes both cube cases and SFF towers, and a lot fewer than expected require the removal of a drive tray as well. I don't think any of them require removing cooling fans. As far as prices are concerned, they cover the full spectrum of dirt cheap to high premium. Of what was available on newegg, there were not that many double pci slot mini-ITX cases that did not also support a double width card of 10.5" or greater. Most of the ones that did not support this configuration had one PCI slot or none.

ITX gaming builds have seen a hefty boon in the last couple years with manufacturers paying attention with making motherboards, graphics cards, and cases specifically with gaming in mind so it's MUCH easier to do it now than when I first did it in 2013.

Your list is worth squat if you don't even take the time to check them over to make sure they can actually run a 980 Ti.

The cooler master elite 110 doesn't not support the 980 ti (10.5 inches), only up to 8.3
http://www.coolermaster.com/case/mini-itx-elite-series/elite110/

The corsair 250D will just barley squeeze it in, assuming your willing to stick to just an SSD and that the PCIe power is side mounted.

http://www.corsair.com/en-us/obsidian-series-250d-mini-itx-pc-case

same thing for the Graphite
http://www.corsair.com/en-us/graphite-series-380t-portable-mini-itx-case

The EVGA case supports 10.5 inches, the exact same size as the REFERENCE 980 Ti. This means that a vast majority of cards most likely won't fit.
http://www.evga.com/Products/Product.aspx?pn=110-MA-1001-K1

Fractal Design Node is a series, not a specific case. You really don't check what your saying before it comes out. Same thing for Phanteks Enthoo.

I'm going to point out this case, even though it does support a 980 Ti
http://www.lian-li.com/en/dt_portfolio/pc-q08/
Note the words "300mm (when HDD cage is removed)"

Last one I'm going to do
ThermalTake Core V1 only supports 10.1 inches max
http://www.thermaltakeusa.com/products-model_Specification.aspx?id=C_00002373

It's obvious at this point that there are few mini-ITX cases that support a 980 Ti, and even less that do without compromise. I could go through the rest but chances are they would be no good, I've better things to do when you aren't even going to put minimal effort, instead shoveling out blatant lies.

 

[Evernessince]

Forgive me but the picture I'm looking at makes this a full height card. It is only shorter. I'm not sure I would have made that statement for a card that wasn't a half height card.

You don't see many half height cards anymore because case makers have caught on. If you can give me a data to support that there are more half height requirement cases out there than full ones, I would like to see it.

 

[cliffordcooley]

If you can give me a data to support that there are more half height requirement cases out there than full ones, I would like to see it.

That is irrelevant coming from your perspective that this is a unique card. A unique card would suggest it is for a unique case. And I'm also pointing out that this card could be 3 inches longer and still fit in 90% of the cases available, making the short length a moot point. The only way for this card to be acceptable at it's pricing, would be if it could fit the other 10% not the majority.

 

[madboyv1]

Stuff

@ Elite 110: I apologize, as that was an error on my part. I did make an attempt to make a cursory glance at GPU clearances and clearly misread that one.
@both Corsair cases: "If it fits it sits." You plan for the build you're making so where the power connectors are located is practically a non issue. Both cases also have all four of their drives installed below the motherboard so they do not interfere with GPU installation. This makes your assumption that you can only install 1 SSD false.
@EVGA case: see above. The two available 3.5" drive bays do not interfere with GPU installation.
@Fractal case: I was referring to the 304, as I did not know of the existence of the 202 (which can also do it). Yes you lose fan capacity in the 202 (but not 304), but even if you put in just the R9 Nano you'd still be losing fans because of the case layout.
@Phanteks case: There's only one relevant case, why would I put a mini-ITX build in a full ATX case? If it makes you feel better it's the PH-ES215 (330mm clearance).
@PC-Q08 case: and in the following paragraph I said "and a lot fewer than expected require the removal of a drive tray as well." My list was not making the attempt to exclude such cases. That being said the drive cage being removed only hold two 3.5" drives. The other FOUR 3.5" drives in the middle cage remain, as does the 5.25" bay. not that it matters but this was the first ITX case I did a gaming build in.
@Thermaltake case: Funny, this case fits the 10.5" long PNY XLR8 GeForce GTX 770 just fine, though as I said, tight fit. (source)

I'll go through the rest for you (since I have time):

AZZA CSAZ-103: "Accommodate VGA card up to 11", No loss of drive capacity.
BitFenix Prodigy: Around 12.5" (Remove middle drive cage). Remaining drive capacity: one 5.25" bay, two 3.5" drives (Three 2.5" drives if you remove this cage as well), and two 2.5" drives.
BitFenix Phenom: Practically the same internally as the Prodigy, but the 5.25" bay is internal only.
COUGAR QBX: 350mm, no loss of drive capacity. Also has the highest number of fan mounts I've seen in a chassis this small... I kind of want to play with one of these...
IN WIN 901: 300mm, no loss of drive capacity.
LIAN LI PC-Q28: 290mm with removal of case floor drive cage, remaining drive capacity: one 5.25" bay, one 2.5" drive, and four 3.5" drives.
Lian LI PC-O5S: 310mm(Remove HDD rack) and cannot use PCI drive bracket, remaining drive capacity: two 2.5" drives and slim ODD can be swapped for a 3.5" drive.
RAIDMAX Atomic: 265mm, VERY tight fit like the Thermaltake Core V1
Rosewill Neutron: 340mm, no loss of drive capacity.
SilverStone RVZ01 & SilverStone ML07: "Supports graphics card up to 13 inches" and no loss to drive capacity
Xigmatek Aquila: I apologize, on second look this is technically a mATX case, for some reason I thought it was ITX only.

So, all of the cases you did not point out can support a GTX 980, though one does not count (xigmatek) and another would be overspec tight (Raidmax). Those that require the removal of a drive cage do so without removing drive capacity as severely as you portray.

It might also behoove you to be a little more amiable as you come across as very cross, defensive, and accusing.

 

[GeforcerFX]

The Pricing makes no sense to me, if I was AMD I would drop the R9 Fury to $525, drop the Nano to $550 and put the R9 Fury X at a even $600. But like I said on the other AMD graphics thread unless they are making massive profit on these high end cards they need to focus some fiji love into the $200-$300 price range.

 

[Darren delaMare]

Sorry chaps, perhaps I am missing the point in all of this. Even at the smallest build level, HTPC, why would anyone install the R9 Fury Nano over the R9 Fury X? What benefits does the end user receive from a Nano setup that cannot be achieved from the Fury X setup? The ONLY thing I can see would be the price difference, but difference would need to be fairly substantial considering the Nano's drawbacks.

The PCB is exactly the same size on both cards, or am I missing something?, so I'm struggling to understand why you would opt for a Nano over a Fury X? Unless you want a card that dumps lots of hot air into the case, throttles after 2 minutes of game play and you consider water-cooling some form of voodoo, then fine!? I guess...

No matter what anyone states, the Nano is not a card designed for HTPC Gaming or Rendering. It will simply cook every other component alive inside your shoe-box size case until something gives up and pops.

If you look at the previous generations of GPU's, find an example of the most power hungry variant then on the same family find the least power hungry. There is usually always a massive difference in spec between the cards so that the lesser variant can run at max 190W TDP vs the 275W TDP we are all accustomed to.

If Fury x and Nano are IDENTICAL cards, cooling aside, I recon that the throttled frequency for the Nano is going to be around the 700-800Mhz range and not the marketing crap of 'Up to 1000mhz'. You heard it here first folks!

Also, I believe that the operational speed and noise generated from that fan is going to sound exactly like my Sapphire R9 280X Vapor-X 3GB once it hits 50 Degrees. WWWHHHHHIIIIIIIIIIINNNNNNNNEEEEEEEEEEEEEE

Feel free to correct me, I'm still reading up on these cards.

 

[Darren delaMare]

R9 Nano Sorry! not Fury X Nano Although thats essentially what it is or is it? I'm all confused by the twins...

 

[Evernessince]

That is irrelevant coming from your perspective that this is a unique card. A unique card would suggest it is for a unique case. And I'm also pointing out that this card could be 3 inches longer and still fit in 90% of the cases available, making the short length a moot point. The only way for this card to be acceptable at it's pricing, would be if it could fit the other 10% not the majority.

It is for unique cases, Mini ITX. This is the most powerful card that will fit ANY mini ITX case. Name another card of this level that can claim that.

@ Elite 110: I apologize, as that was an error on my part. I did make an attempt to make a cursory glance at GPU clearances and clearly misread that one ...................

With all that said, can we agree that sacrifices almost always have to be made in most mini-ITX cases in order to install a 980 Ti. Depending on the performance numbers of the Fury Nano, it's targeting this kind of niche market.

 

[madboyv1]

With all that said, can we agree that sacrifices almost always have to be made in most mini-ITX cases in order to install a 980 Ti. Depending on the performance numbers of the Fury Nano, it's targeting this kind of niche market.

We can agree to disagree about the cases, but it is AMD intent for sure that the R9 Nano is targeting the SFF gaming market. I've been building in this space for some time now so any new hardware for SFF gets my jimmies rustled. I can't wait to see the reviews because I really want to see how the lower wattage effects the performance of this chip.

 

[Evernessince]

We can agree to disagree about the cases, but it is AMD intent for sure that the R9 Nano is targeting the SFF gaming market. I've been building in this space for some time now so any new hardware for SFF gets my jimmies rustled. I can't wait to see the reviews because I really want to see how the lower wattage effects the performance of this chip.

Same here. It would be even more interesting if someone could get the chip to near fury levels with a good enough cooling solution. Chances are the Fury Nano's are going to be lower binned Fury cards though.