AMD Unveils Radeon R9 Nano, Targets Mini ITX Gaming Systems With a New Fury

MojoKid writes: AMD today added a third card to its new Fury line that's arguably the most intriguing of the bunch, the Radeon R9 Nano. True to its name, the Nano is a very compact card, though don't be fooled by its diminutive stature. Lurking inside this 6-inch graphics card is a Fiji GPU core built on a 28nm manufacturing process paired with 4GB of High Bandwidth Memory (HBM). It's a full 1.5 inches shorter than the standard Fury X, and unlike its liquid cooled sibling, there's no radiator and fan assembly to mount. The Fury Nano sports 64 compute units with 64 stream processors each for a total of 4,096 stream processors, just like Fury X. It also has an engine clock of up to 1,000MHz and pushes 8.19 TFLOPs of compute performance. That's within striking distance of the Fury X, which features a 1,050MHz engine clock at 8.6 TFLOPs. Ars Technica, too, takes a look at the new Nano.

Comments

as an engineer I couldnt be happier

[nimbius]

I purchased an ITX system with this radeon as part of a project ive been working on. The system runs a complex thermodynamics application, Crysis 3, in order to physically model stresses on nuclear containment vessels during a meltdown.

Re:Well, I read *that* headline wrong

[UnknownSoldier]

That way --> to the Geek Hierarchy chart

Bringing this back on topic: Disappointed with new tech? Welcome to the club. Hardware has become so stagnant in the last 5 years. 28nm. *yawn*. Yet-another-Megaherz or "core". /sarcasm Yay.

When are the GPU OEM's going to move to 22 nm?

When the hell is Knights Corner going to be ready for the masses?

Business as usual. Smaller, Faster, Cheaper.

When is the next (tech) revolution going to happen?

Fifteen years.

[jeffb (2.718)]

In 2000, the fastest supercomputer in the world was IBM's ASCI White, with a peak performance of 7.226 TFLOPS. Its theoretical maximum performance was 12.3 TFLOPS. It weighed over 100 tons, and drew 3MW of power, plus another 3MW for cooling.

One. Six. Inch. Card.

Re:Fifteen years.

[NJRoadfan]

Somehow this reminded me of 3dfx's old ad campaign.... all that power.... to play games: https://www.youtube.com/watch?... https://www.youtube.com/watch?...

Re:Fifteen years.

[Anonymous Coward]

Wrong. That's 7.226 TFlops of DOUBLE PRECISION performance on the supercomputers.
Even the full-bore Furry-X gets about 0.55 TFlops of double precision and let's optimistically assume that these cut-down parts can still pull 0.5 TFlops. That's 14.452 (round up to 15) cards.

Now, Knight's Landing? That's about 3 TFlops of double-precision in a single chip, so three CHIPS to beat that supercomputer.. not too shabby.

Still pretty impressive, but that's just proof that despite the naysayers, Moore's law is alive and well.

Re:Fifteen years.

[tgetzoya]

I loved these....haven't forgotten about them or 3dfx. Too bad they couldn't get their tech together.

Re:Fifteen years.

[Tough Love]

Your cell phone packs considerably more compute power than a Cray 1.

175W

[apcullen]

Do you REALLY want a 175W card in a mini itx system? How are you going to keep it cool?

Re:175W

[U2xhc2hkb3QgU3Vja3M]

Still too big for a mini-ITX case. Why are they still making them so deep? Mini-ITX is square, why are the damn cases still rectangle? How about a tiny tower the same size as the motherboard? Vertical space doesn't cost anything on my desk.

Re:175W

[slaker]

And as a related issue, who is making a true SFF power supply big enough to handle that card plus a gaming-class CPU? A lot of "ITX" rigs are built using configured mini-towers (e.g. the Bitfenix Prodigy), but if I wanted to throw one of those in a vanilla case like an Apex MI-08 or Antec ISK-150, their PSUs would die approximately 10 seconds after I fired up Crysis or whatever it is that kids are playing these days.

Re:175W

[sexconker]

All you need is a bit of clearance between the fan and the desk. Put a small bit of wood under each corner.

Sorry, no

[Yunzil]

At $450 it would have been intriguing. At $650 it's pointless.

Re:Well, I read *that* headline wrong

[Kjella]

Bringing this back on topic: Disappointed with new tech? Welcome to the club. Hardware has become so stagnant in the last 5 years. 28nm. *yawn*. Yet-another-Megaherz or "core". /sarcasm Yay. (...) When is the next (tech) revolution going to happen?

Actually I feel we've had several since the PC revolution. There was the network revolution with the Internet. The mobile revolution that lets you use it anywhere, any time. And with fiber rolling out I'd say we're in the middle of a bandwidth revolution. Even if you extrapolate like crazy going from 8GB to 16GB RAM isn't going to feel like going from 8MB to 16MB. The changes were huge because there was so much you couldn't do with 8MB, there's not so much you can't do with 8GB. Welcome back to the real world, where cars and planes don't go twice as fast with double the capacity and mileage three years later. Has it actually bugged you that you don't have terahertz processors or terabytes of RAM or petabytes of storage lately? I can't really say that I have, I often wish shit would work better but it's not because they lack hardware resources. There was a time when the really hardware wasn't capable even if you wrote optimized assembler, today it's 99.99% the software that's not capable.

Re:no fiji under $500?

[cb88]

Specifically it means it leaks less current.. especially when ran at a lower voltage.

The Fury X and Pro chips may or may not run at the voltage these chips do and they'll probably leak more current even when they do.

In short... this is a higher efficiency chip. Most likely it would be able to clock higher than Fury X or Pro chips due to less leakage as well given appropriate cooling.