Small Size, Giant Impact. That’s how all of AMD’s R9 Nano presentations begin and with good reason; this card is impressively compact but is supposed to offer up titanic performance numbers, particularly at 4K. While this may sound impossible, AMD has been able to achieve their goals by combining an efficient core design, HBM technology and no small amount of selective core binning.
AMD’s R9 Fury X and subsequent R9 Fury showed gamers what High Bandwidth Memory could accomplish, even when paired up with a slightly antiquated core architecture. Indeed, those Fiji-based cards ran neck and neck with their GeForce competitors and were actually able to pull slightly ahead at UHD resolutions. Now the R9 Nano is being introduced in an effort to once again leverage the strengths of HBM and a unified GPU design approach.
While we have already previewed the R9 Nano, a number of its key elements need to be highlighted once again since some of them simply defy belief. Instead of utilizing a cut-down version of the Fiji XT core like the vanilla R9 Fury, the Nano actually comes to the table packing a full 4096 Stream Processors, 64 ROPS and 256 Texture Units. There’s nothing particularly special here until you get to the TDP numbers AMD is claiming; 175W is a full 100W lower than the Fury X.
So how was AMD able to achieve this? By judiciously modulating the core’s voltage and frequency while also choosing cores with exceedingly low temperature / heat outputs. Core clocks should reach around 900MHz under normal operating conditions while there may be some instances where users will see speeds of up to 1GHz. Meanwhile, the HBM’s speed of 500MHz hasn’t been touched.
To make matters even more interesting AMD also claims this is one of the coolest, quietest running GPUs around. That almost sounds too good to be true since taming a 175W card with a relatively small cooing solution has proven to be quite challenging in the past.
In many ways the R9 Nano can be considered the flagship of AMD’s current lineup since it personifies the tantalizing possibilities offered by the use of HBM. That HBM which has contributed to cost overruns and production shortfalls but we are now seeing the first tantalizing possibilities of what it can bring to the table.
With that being said, AMD will likely face an uphill battle when trying to actually sell this card. It is small and seems to pack an almighty framerate wallop but you’ll need to pay a hefty price of $650 (nearly $1000CAD) for the luxury of owning the fastest compact graphics card on the market. However, the actual market for it may be a bit limited since most of today’s SFF chassis are designed in such a way that they easily accept longer, less expensive options. In addition, the Nano lacks features like HDMI 2.0 and HDCP 2.2 support, both of which would make it infinitely more appealing for the HTPC market. Naturally, the stringent binning process will likely lead to the Nano becoming something of a rarity on retailers’ shelves but for people who want high end ITX performance it may be a perfect fit.
Now the Nano’s initial previews weren’t without their fair share of controversy either. AMD claimed some exceedingly impressive performance numbers that showed the Nano beating NVIDIA’s GTX 970 by a solid 30%. When taken at face value that’s reason to be hopeful but it turns out those figures were achieved with a unique set of settings (disabling anisotropic filtering and anti-aliasing for example) which played to AMD’s architectural strengths while sacrificing image quality. Will AMD's internal numbers hold up in this review? I certainly hope so since the Nano really does have a great deal of intrinsic appeal.
Upon seeing the R9 Nano for the first time, my initial reaction was a straightforward “Awwww…..isn’t it CUTE!?”. Yes, it really is cute, and simple and wonderfully compact. At just 6” long the R9 Nano is deceptively sized given the performance it is meant to deliver.
The top shroud is an understated affair which is capped by a standard 80mm fan set up in a typical downdraft-style setup. Unfortunately, this design means that a good portion of the core’s heat will be dumped back into the chassis. According to AMD, much of the Nano’s relative strengths lie within its heatsink design. While the main fin array looks like many other cards we have reviewed in the past, there’s some substantial technological feats going on below the surface.
When placed next to a standard full length graphics card, the R9 Nano looks downright diminutive. This really puts into perspective how HBM has moved GPU design forward by leaps and bounds.
The next few sections have been taken from our initial preview article since nothing has changed with the Nano’s component layout and design.
Alongside the high static pressure but low RPM 80mm fan, the Nano incorporates a pure copper core contact plate which interfaces directly with a dual vapor chamber. This vapor chamber makes direct contact with the aforementioned aluminum fin array to insure quick transfer of heat and efficient cooling.
While we have seen this approach in the past, AMD’s temperature claims are impressive to say the least. Supposedly, this card will have a nominal operating temperature of just 75°C while still operating at just 42 dbA. That may seem like an impossibility but there’s nothing I can see to invalidate this claim.
Past the advanced heatsink design, AMD is also addressing the cooling for their VRM components since there really isn’t all that space to optimize airflow. In this case they have added a secondary copper heatpipe and heatsink which are supposed to remove any worry about components overheating.
Under the heatsink is a relatively anemic setup with a 4-phase GPU PWM along with a single phase for secondary functions. Additional phases really aren’t needed since overclocking isn’t one of this card’s focuses and the core itself is quite efficient.
Power input is handled by a single 8-pin input which makes installing this into certain ITX chassis a bit challenging. Some of the more compact cases on the market require SFF PSUs which are rare to begin with. The ones that have enough current to supply a 175W graphics card and incorporate a dual six-pin / single 8-pin PCI-E connector setup tend to be quite expensive and even rarer.
The Nano’s side holds a simple Radeon logo along with a small switch that can be used to change between BIOSes. According to AMD, this doesn’t engage an Uber Mode like on the previous generation cards but simply houses two identical BIOS files. Tweakers will likely use the secondary slot for modified profiles.
The card’s backside doesn’t house anything particularly interesting but AMD did add a few of their PWM components here in an effort to save space on the PCB’s topside. Unfortunately, these are left open to the elements and may heat up without adequate cooling.
The I/O plate is pretty much par for the course with a trio of DisplayPort outputs along with a single HDMI 1.4 connector. The lack of HDMI 2.0 certainly hurts this card’s possible use in HTPC scenarios since lower spec HDMI cables don’t have sufficient bandwidth for 4K60 transmission and most newer UHD TVs don’t include DisplayPort inputs. Supposedly there will be DP to HDMI 2.0 adapters available from some of AMD’s board partners but they will have to be purchased separately and won’t pass on a HDCP 2.2 signal so playing protected content will be impossible.
AMD’s R9 Fury X and subsequent R9 Fury showed gamers what High Bandwidth Memory could accomplish, even when paired up with a slightly antiquated core architecture. Indeed, those Fiji-based cards ran neck and neck with their GeForce competitors and were actually able to pull slightly ahead at UHD resolutions. Now the R9 Nano is being introduced in an effort to once again leverage the strengths of HBM and a unified GPU design approach.
While we have already previewed the R9 Nano, a number of its key elements need to be highlighted once again since some of them simply defy belief. Instead of utilizing a cut-down version of the Fiji XT core like the vanilla R9 Fury, the Nano actually comes to the table packing a full 4096 Stream Processors, 64 ROPS and 256 Texture Units. There’s nothing particularly special here until you get to the TDP numbers AMD is claiming; 175W is a full 100W lower than the Fury X.
So how was AMD able to achieve this? By judiciously modulating the core’s voltage and frequency while also choosing cores with exceedingly low temperature / heat outputs. Core clocks should reach around 900MHz under normal operating conditions while there may be some instances where users will see speeds of up to 1GHz. Meanwhile, the HBM’s speed of 500MHz hasn’t been touched.
To make matters even more interesting AMD also claims this is one of the coolest, quietest running GPUs around. That almost sounds too good to be true since taming a 175W card with a relatively small cooing solution has proven to be quite challenging in the past.
<iframe width="640" height="360" src="https://www.youtube.com/embed/NsjwQ1TwgIY?rel=0" frameborder="0" allowfullscreen></iframe>
In many ways the R9 Nano can be considered the flagship of AMD’s current lineup since it personifies the tantalizing possibilities offered by the use of HBM. That HBM which has contributed to cost overruns and production shortfalls but we are now seeing the first tantalizing possibilities of what it can bring to the table.
With that being said, AMD will likely face an uphill battle when trying to actually sell this card. It is small and seems to pack an almighty framerate wallop but you’ll need to pay a hefty price of $650 (nearly $1000CAD) for the luxury of owning the fastest compact graphics card on the market. However, the actual market for it may be a bit limited since most of today’s SFF chassis are designed in such a way that they easily accept longer, less expensive options. In addition, the Nano lacks features like HDMI 2.0 and HDCP 2.2 support, both of which would make it infinitely more appealing for the HTPC market. Naturally, the stringent binning process will likely lead to the Nano becoming something of a rarity on retailers’ shelves but for people who want high end ITX performance it may be a perfect fit.
Now the Nano’s initial previews weren’t without their fair share of controversy either. AMD claimed some exceedingly impressive performance numbers that showed the Nano beating NVIDIA’s GTX 970 by a solid 30%. When taken at face value that’s reason to be hopeful but it turns out those figures were achieved with a unique set of settings (disabling anisotropic filtering and anti-aliasing for example) which played to AMD’s architectural strengths while sacrificing image quality. Will AMD's internal numbers hold up in this review? I certainly hope so since the Nano really does have a great deal of intrinsic appeal.
Upon seeing the R9 Nano for the first time, my initial reaction was a straightforward “Awwww…..isn’t it CUTE!?”. Yes, it really is cute, and simple and wonderfully compact. At just 6” long the R9 Nano is deceptively sized given the performance it is meant to deliver.
The top shroud is an understated affair which is capped by a standard 80mm fan set up in a typical downdraft-style setup. Unfortunately, this design means that a good portion of the core’s heat will be dumped back into the chassis. According to AMD, much of the Nano’s relative strengths lie within its heatsink design. While the main fin array looks like many other cards we have reviewed in the past, there’s some substantial technological feats going on below the surface.
When placed next to a standard full length graphics card, the R9 Nano looks downright diminutive. This really puts into perspective how HBM has moved GPU design forward by leaps and bounds.
The next few sections have been taken from our initial preview article since nothing has changed with the Nano’s component layout and design.
Alongside the high static pressure but low RPM 80mm fan, the Nano incorporates a pure copper core contact plate which interfaces directly with a dual vapor chamber. This vapor chamber makes direct contact with the aforementioned aluminum fin array to insure quick transfer of heat and efficient cooling.
While we have seen this approach in the past, AMD’s temperature claims are impressive to say the least. Supposedly, this card will have a nominal operating temperature of just 75°C while still operating at just 42 dbA. That may seem like an impossibility but there’s nothing I can see to invalidate this claim.
Past the advanced heatsink design, AMD is also addressing the cooling for their VRM components since there really isn’t all that space to optimize airflow. In this case they have added a secondary copper heatpipe and heatsink which are supposed to remove any worry about components overheating.
Under the heatsink is a relatively anemic setup with a 4-phase GPU PWM along with a single phase for secondary functions. Additional phases really aren’t needed since overclocking isn’t one of this card’s focuses and the core itself is quite efficient.
Power input is handled by a single 8-pin input which makes installing this into certain ITX chassis a bit challenging. Some of the more compact cases on the market require SFF PSUs which are rare to begin with. The ones that have enough current to supply a 175W graphics card and incorporate a dual six-pin / single 8-pin PCI-E connector setup tend to be quite expensive and even rarer.
The Nano’s side holds a simple Radeon logo along with a small switch that can be used to change between BIOSes. According to AMD, this doesn’t engage an Uber Mode like on the previous generation cards but simply houses two identical BIOS files. Tweakers will likely use the secondary slot for modified profiles.
The card’s backside doesn’t house anything particularly interesting but AMD did add a few of their PWM components here in an effort to save space on the PCB’s topside. Unfortunately, these are left open to the elements and may heat up without adequate cooling.
The I/O plate is pretty much par for the course with a trio of DisplayPort outputs along with a single HDMI 1.4 connector. The lack of HDMI 2.0 certainly hurts this card’s possible use in HTPC scenarios since lower spec HDMI cables don’t have sufficient bandwidth for 4K60 transmission and most newer UHD TVs don’t include DisplayPort inputs. Supposedly there will be DP to HDMI 2.0 adapters available from some of AMD’s board partners but they will have to be purchased separately and won’t pass on a HDCP 2.2 signal so playing protected content will be impossible.
Last edited:
