Sapphire Launches A New Flagship: Nitro R9 Fury

by Kevin Carbotte January 12, 2016 at 6:55 AM - Source: Sapphire

Sapphire has announced a new flagship graphics card based on AMD’s Fury Pro specifications. The Nitro R9 Fury features a custom PCB with reworked circuitry and a higher overclock than the Tri-X OC Fury we tested last year.

Sapphire’s R9 Fury Tri-X was one of the first overclocked Fury Pro cards to hit the market, and now the company has come back with even more aggressive clock speeds. The Sapphire Nitro R9 Fury has just been announced, and it comes in two flavors. The Nitro R9 Fury OC has a modest 20 MHz overclock, setting the Fiji GPU at 1020 MHz. The Nitro R9 Fury OC+ comes with a more aggressive 1050 MHz overclock, which represents a 50 MHz boost from stock and 10 MHz over Sapphire’s own Tri-X OC edition.

Sapphire’s Nitro R9 Fury shares many of the same features as the Tri-X R9 Fury. Both cards use the same robust Tri-X cooling system, with central 10 mm heat pipe and triple 90 mm fans, and they both include a thick aluminum back plate to aid cooling the backside of the PCB. The main difference between the Tri-X R9 Fury and the newly launched Nitro R9 Fury is the PCB design.

Sapphire said the Nitro R9 Fury’s PCB includes 6-phase circuitry on an 8-layer PCB, which features 2 Oz of copper, enabling Sapphire to push 360 Amps through the GPU. The company said this is to help deliver reliable, stable power while overclocking. Sapphire claimed tha t this results in 20% lower VRM temperatures and higher stable clock speeds.

Sapphire’s Nitro R9 Fury has also been equipped with a dual-BIOS switch. The default setting sets the GPU to 1000 MHz and limits the power draw to 260 W and the GPU temperature to 75 degrees C. The second BIOS is set much more aggressively. The power limit is increased to 300 W, and the GPU temperature target raised to 80 degrees C. With this BIOS, the GPU’s factory setting is 1020 MHz for the OC version and 1050 MHz for the OC+ variant.

Sapphire noted that the Nitro R9 Fury uses Black Diamond chokes, and the design has been optimized to minimize coil noise, which is something we noted to be prominent on the overclocked Tri-X R9 Fury.

Model	SAPPHIRENITRO R9 FURYOC+	SAPPHIRE NITRO R9 FURY OC
Graphicsprocessor	Fiji Pro	Fiji Pro
Manufacturing process	28 nm	28 nm
Memory	4 GB HBM 4096-bit	4 GB HBM 4096-bit
Stream processors	3584	3584
Texture units	224	224
ROPs	64	64
Engine clock	1050 MHz (OC+)	1020 MHz (OC)
Memory clock	500 MHz	500 MHz
Cooling system	Tri-X triple-fancooling with dual-ball bearings, 90mm fans;	Tri-X triple-fancooling with dual-ball bearings, 90mm fans;
	5 copper heatpipes (1x 10mm, 2x8mm,2x 6mm),	5 coppe rheatpipes (1x 10mm, 2x8mm,2x 6mm),
	AMD Advanced Fan Control and SAPPHIRE Intelligent Fan Control II	AMD Advanced Fan Control and SAPPHIRE Intelligent Fan Control II
Outputs	3x DisplayPort,HDMI 1.4a,DVI-D	3x DisplayPort,HDMI 1.4a,DVI-D
Interface	PCI-Express 3.0	PCI-Express 3.0
Special features	Dual Firmware (Default &Performance)	Dual Firmware (Default &Performance)
Special features	Black Diamond Choke	Black Diamond Choke
Technologies	DirectX 12,Vulkan, Mantle, OpenGL 4.5, OpenCL 2.0, AMDFreeSync, AMDCrossFireXDMA (bridgeless), AMD Eyefinity, AMDLiquidVR	DirectX 12,Vulkan, Mantle, OpenGL 4.5, OpenCL 2.0, AMDFreeSync, AMDCrossFireXDMA (bridgeless), AMD Eyefinity, AMDLiquidVR

The Sapphire Nitro R9 Fury is available now through various retailers and online stores.

Follow Kevin Carbotte @pumcypuhoy. Follow us on Facebook, Google+, RSS, Twitter and YouTube.

About the author

Kevin Carbotte @pumcypuhoy

Kevin Carbotte is a Contributing Writer for Tom's Hardware US. He writes news and reviews of graphics cards and virtual reality hardware.

This thread is closed for comments

27 comments

Comment from the forums

Your comment

2Be_or_Not2Be

"Sapphire said the Nitro R9 Fury’s PCB includes 6-phase circuitry on an 8-layer PCB, which features 2 Oz of copper, enabling Sapphire to push 360 Amps through the GPU. "

Well, 360 Amps in a single GPU is absolutely incredible! Perhaps a little too incredible...

Maybe you meant 360 watts.
Epsilon_0EVP

From what I remember, the power delivery circuitry in the card reduces voltage to about 1V by the time it gets to the GPU. So 360A is exactly equivalent to 360W; they're just quoting the figure in amps.
Moeseph de tyre

360 amps would be correct. P=VI, 1.2 volts times 360 amps equals 432 watts. So there is the headroom for a hungry Fiji when over clocked.
kcarbotte

1396902 said:

"Sapphire said the Nitro R9 Fury’s PCB includes 6-phase circuitry on an 8-layer PCB, which features 2 Oz of copper, enabling Sapphire to push 360 Amps through the GPU. " Well, 360 Amps in a single GPU is absolutely incredible! Perhaps a little too incredible... Maybe you meant 360 watts.

it's possible that Sapphire meant watts, but the release clearly states 360 amps.
Just because the PCB can handle that much power, doesn't mean it will ever happen in the real world. I believe this is just Sapphire over building the PCB, but you bring up a good point.

I will reach out to Sapphire for clarification of this.
2Be_or_Not2Be

Quote:

From what I remember, the power delivery circuitry in the card reduces voltage to about 1V by the time it gets to the GPU. So 360A is exactly equivalent to 360W; they're just quoting the figure in amps.

True, but I don't think they are reducing the voltage and greatly increasing the amperage. That would seem like a bit of unnecessary circuitry.
ko888

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit.

Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw:
440 Watts ÷ 1.22 Volts = 361 Amps.
Epsilon_0EVP

217811 said:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

That's not Ohm's Law, but the calculations themselves are correct.
ko888

781954 said:

217811 said:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

That's not Ohm's Law, but the calculations themselves are correct.

When I took Electrical Engineering in university it sure was.
toddybody

How are non-reference Fiji clocks now? Last I saw (launch articles), they were struggling to take Fury X's over 75Mhz...which blew my mind.
Epsilon_0EVP

217811 said:

When I took Electrical Engineering in university it sure was.

From the ECE/Physics courses I took, what is now known as Ohm's Law is only the equation V = IR (and any directly related equation). The formula P = VI is the electric power formula. Perhaps it was taught differently where/when you were taught.

The diagram you provide only uses the two laws in conjunction to write any one of the four variables in terms of the two others. Note that no single expression in the diagram uses more than two variables, indicating two degrees of freedom consistent with two equations.
Wisecracker

12-volt line from the PSU ---- 360w (not Amps) = 30 Amps
Epsilon_0EVP

117113 said:

12-volt line from the PSU ---- 360w (not Amps) = 30 Amps

As explained already, the card's power delivery transforms the 12V input into around 1.22V by the time it gets to the GPU. 360A at that voltage is perfectly reasonable.
Wisecracker

Historically, that ain't how specs on GPUs -or CPUs for that matter- are reported. Presumably at 360w you may compute the volt delivery available from the system required for the card:

by the PCIe x16 slot: 75w
. . . . . by a 6+2pin: 150w

Max system power deliver = 375w
Card power requirements = 360w

See how that worked out?
dwhapham

Quote:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

Quote:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

Guys - 361 Amps in impossible.. Think about it, Most of us have our PCs plugged into an 110v, 15 or 30amp outlet. So tell me how your video card is going to pull 361 amps from that same outlet? (Spoiler: It's not).
ko888

151502 said:

Quote:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

Quote:

That card's GPU operates at 1.22 Volts when running a 3D load. The input voltage of 12 Volts needs to be dropped down to 1.22 Volts by the VRM circuit. Since that card can draw over 440 Watts, when running Furmark, then Ohm's Law tells us that the GPU will draw: 440 Watts ÷ 1.22 Volts = 361 Amps.

Guys - 361 Amps in impossible.. Think about it, Most of us have our PCs plugged into an 110v, 15 or 30amp outlet. So tell me how your video card is going to pull 361 amps from that same outlet? (Spoiler: It's not).

No one said it would be pulling 361 Amps from the wall outlet. The GPU chip may draw up to 361 Amps from the graphics card's VRM circuit at 1.22 Volts DC.

The graphics card, itself, will draw less than 37 Amps from the PSU's +12V rail and less than 5 Amps AC from a 115VAC outlet and that's using a PSU conversion efficiency of 88% for example.
ko888

117113 said:

Historically, that ain't how specs on GPUs -or CPUs for that matter- are reported. Presumably at 360w you may compute the volt delivery available from the system required for the card: by the PCIe x16 slot: 75w . . . . . by a 6+2pin: 150w Max system power deliver = 375w Card power requirements = 360w See how that worked out?

If you're basing your example on history then why didn't you cite the AMD Radeon R9 295X2 that only has two 8-pin PCIe power connectors and can draw up to 646 Watts?

AMD has been known to violate the PCIe specs. For the Radeon R9 295X2 to function they even state that the PSU needs to be able to provide at least 28 Amps (i.e. 336 Watts) from the +12V rail(s) to each 8-pin PCIe connector.
turkey3_scratch

That's why thick wiring comes in handy at times. I forget what PSU it was that is coming out that has 1.5x thicker wiring.
farlandprince

still waiting for the dual GPU AMD ...
Wisecracker

Quote:

... If you're basing your example on history then why didn't you cite the AMD Radeon R9 295X2 that only has two 8-pin PCIe power connectors and can draw up to 646 Watts?

You're conflating total system power load, Skippy, with the individual power consumption of a 2-GPU video card.

But please feel free to ramble on . . .
TJ Hooker

117113 said:

Historically, that ain't how specs on GPUs -or CPUs for that matter- are reported. Presumably at 360w you may compute the volt delivery available from the system required for the card: by the PCIe x16 slot: 75w . . . . . by a 6+2pin: 150w Max system power deliver = 375w Card power requirements = 360w See how that worked out?

You're right, GPU specs typically mention power, not current. However, in this case, they're specifically talking about the PCB and VRMs, in which case discussing current makes complete sense.

Here's another example: http://www.tomshardware.com/news/amd-fury-x-fiji-preview,29400.html
"The board features a six-phase power design that was engineered to handle up to 400 A of current."

117113 said:

Quote:

... If you're basing your example on history then why didn't you cite the AMD Radeon R9 295X2 that only has two 8-pin PCIe power connectors and can draw up to 646 Watts?

You're conflating total system power load, Skippy, with the individual power consumption of a 2-GPU video card. [...] But please feel free to ramble on . . .

It's still drawing nearly 450 watts, significantly more than 375W that PCIe + 2x8 pin connectors are rated to supply. So ko888's point still stands.
turkey3_scratch

@Wisecracker: Though, that is gaming power consumption. Stress test will be quite higher.
ko888

117113 said:

Quote:

... If you're basing your example on history then why didn't you cite the AMD Radeon R9 295X2 that only has two 8-pin PCIe power connectors and can draw up to 646 Watts?

You're conflating total system power load, Skippy, with the individual power consumption of a 2-GPU video card. ... But please feel free to ramble on . . .

The Tom's Hardware graph, that you've linked to, shows the graphics card's power consumption during gaming not the system's total power consumption. Gaming does not show the card's maximum power draw.

The only one confused is you.

Here's another review of that card where they actually measure the graphics card's maximum worse case power consumption when running Furmark. It doesn't measure the useless system power consumption because they're not doing a system review.

This German review that measures only the graphics card's power consumption also backs up what I said. When they ran Furmark the card drew 630 Watts.
vaughn2k

Here we go..
ichigokunbaka

so how many 230w psu's daisy chained gives best idle/full load performance and how do i deploy this idea ?