AMD demonstrates its 12 TFLOP dual-GPU monster, the Fury X2

For all of the acclaim the unofficially named R9 Fury X2 has garnered, AMD hasn’t done much more than wave a dual-chip PCB in the year, proclaiming it the best GPU for virtual reality and calling it a day.

That’s been the case for very nearly every event the Fury X2 is brought up in conversation, followed by AMD’s assurance that the card is kind of… sort of ready, and that it has just been delayed to coincide with the release of VR headsets.

We’ve been talking about AMD’s dual-Fiji GPU, codenamed ‘Gemini’, for some time now.

A lot of that discussion has centred on the use of two Fiji XT chips, the same chip that powers the beastly Radeon R9 Fury X, on a single PCB.

The result is 8192 stream processors, 8GB of HBM (first generation) memory and more than a terabyte a second of bandwidth.

Even the GeForce GTX Titan X would shiver at the site of this card.

Well, it seems like a report by Benchlife.info last year that the card would receive a soft launch during February or March was right on the money.

In fact, during VRLA 2016 Winter Expo in Los Angeles, which just recently took place, AMD’s Roy Taylor not only revealed the compute performance the card is capable of, but Falcon Northwest demonstrated their latest ultra-compact Tiki PC, which just happens to use a Fury X2.

Big thanks to @AMDRadeon engineers for making this possible! Dual Fiji's in our 4-inch wide PC is insane! https://t.co/YT272DCKJB

— Falcon Northwest (@FalconNW) January 23, 2016

The Tiki was used to power an HTC Vive presentation, which it undoubtedly had no problem with.

Virtual reality may require a lot of graphic performance to do correctly, but with its huge computational grunt – 12 TFLOPs to be exact – and massive bandwidth, that’s exactly what the Fury X2 does best.

Prototype Tiki from @FalconNW powering #htcvive with dual Fiji @AMDRadeon at the #vrla pic.twitter.com/2gCxgzucB5

— Antal Tungler (@coloredrocks) January 23, 2016

That’s right, the Fury X2 can manage 12 TFLOPs of compute performance. It may only be 0.5 TFLOPs more than the R9 295X2, but it does so with significantly less power draw. And it’s still 5 TFLOPs more than NVIDIA’s Titan X can manage.

It’s also more compute than every component in this R100,000 PC can manage.

“Last time I was here I also promised you that we would make the world’s most powerful small computer for developers. We promised you we would take two of our highest end GPUs and put it inside that tiny box and if you go downstairs we actually have a demonstration of a dual GPU, 12 TeraFlops, fastest GPU solution in the world, inside of Tiki. It’s a feat of engineering we are delighted with,” said Taylor.

We were hoping for a little more compute performance from the Fury X2, but its performance at its reported TDP is pretty significant. Plus, its bandwidth will still give it the oomph necessary to run circles around the R9 295X2 at 4K resolution.

Category	AMD Radeon R9 Fury X2	AMD Radeon R9 295X2
GPU	Fiji XT x 2	Hawaii XT x 2
Stream Processors	8192	5632
Render Output Units	128	128
Texture Mapping Units	512	352
GPU Frequency	TBC	1018 Mhz
Memory	8GB HBM (4 GB Per GPU)	8GB GDDR5 (4GB Per GPU)
Memory Interface	8192bit (4096 Per GPU)	1024bit (512 Per GPU)
Memory Bandwidth	1 Terabyte/s	640 GBps
Cooling	TBC	Liquid
Performance (FP32)	12 TFLOPS	11.5 TFLOPS
TDP	~375W	~500W
GFLOPS/Watt	32	23
Launch Price	TBC	$1499
Launch Date	TBD-2016	April 2014

*Table courtesy of WCCFTech.

Weirdly, we’re more excited to see how Northwest Falcon’s latest Tiki performs.

It might have something to do with the compute of the Fury X2 being lower than expected, but the last Tiki we covered was an absurdly overspecced 18-core (36-thread) ultra-compact PC the size of an Xbox 360 (smaller in fact), sporting a GTX Titan X and priced roughly R110,000 at the time of that piece.

And we can’t wait to see AMD’s own Fury X2 project, the Quantum, realised – perhaps the most amazing ultra-compact, bespoke PC we’ve ever laid our eyes on.

Perhaps Zen is what’s necessary to make that happen.