AMD says that with the R9 295X2 it was "proud to bring the first closed loop liquid cooling solution to a reference graphics card." That might be true but it also had little choice in the matter, as air-cooling simply wouldn't suffice, at least not at a tolerable operating volume. Not so long ago liquid-cooling was a niche industry, but with so many enthusiasts now opting to use closed-loop liquid cooling for their CPU, it is almost becoming the norm. To maintain its 7.5" PCB the Fury X is equipped with a water-block on the GPU alongside an external 120mm radiator. Dissipating at least 275 watts of power needs a serious cooler and instead of equipping a 13" long heatsink and fan combo, AMD looked to the R9 295X2 for inspiration. Of course while the PCB is compact, the cooling isn't.
Never before have we seen an ultra-high-end graphics card that measures less than 10" in length and the Fury X is well under this at just 7.5". This then brings us to the Fury X board design which, like HBM, is unique in itself. For example 1GB of GDDR5 memory (comprised of four 256MB chips) requires roughly 672mm2 of space and using HBM technology that same 1GB needs only about 35mm2.Īccording to AMD, 9900mm2 of PCB is required for the 290X GPU and its associated memory whereas less than 4900mm2 is used for the Fury X. We tend to agree.Īpart from the obvious performance advantages and power savings, another key advantage to HBM technology is saving space as the DRAM dies are stacked on top of each other. However, AMD says those situations won't matter much because the games most likely won't be playable at those settings anyway. This means HBM comes at a cost of a little over 35GB/s per watt while GDDR5 offers just 10.6GB/s per watt.įor now the key downside is the 4GB capacity limitation, which could potentially become a problem at ultra-high resolutions. This means even at much lower clock speeds the wider bus allows HBM to deliver over 100GB/s per stack compared to 28GB/s per GDDR5 chip. However keep in mind that HBM memory features a bus width of 1024-bits per stack versus 32-bits of a GDDR5 memory chip. The core is clocked at up to 1050MHz, which is the same frequency as the 390X.Īs noted earlier, the HBM is clocked at just 500MHz (1.0Gbps) which is obviously much lower than the GDDR5 memory of the 390X. In total, the Fury X core features 4096 SPUs, 256 TAUs and 64 ROPs, 45% more SPU and TAUs than the 390X with the same amount of ROPs. All told, the Fury X has 44% more transistors than the 390X while the die is 36% larger. That said, if you include the 4GB of HBM then the die area measures 1011mm2. The R9 Fury X core is comprised of an incredible 8.9 billion transistors and because the 28nm process is still being used, the die measures a massive 596mm2. For the Fury X, AMD has matched the GTX 980 Ti's $650 MSRP out of the gate and that price includes a liquid cooler. AMD wasn't particularly aggressive with pricing on the 300 series, charging $100 more for the 390X than the 290X, which can be overclocked to rival the 390X's performance. We're also interested to see how the card's pricing checks out. We expect the Fury X to put up more of a fight against Nvidia's flagship.
HBM is a superior technology to GDDR5 by almost every memory metric, offering more bandwidth while using less power and 19x less surface area than GDDR5.Īlthough the Radeon 300 series brought a performance boost and helped strengthen AMD's position in the high-end market, the R9 390X couldn't top the GTX 980 much less the 980 Ti or Titan X. The three-dimensional die stacking and tiny interconnects allows for superior bandwidth at lesser power consumption than both GDDR5 and DDR4.
HBM is an exciting new memory technology that will allow AMD to develop some truly innovative GPUs, and the Fury X should be only the beginning.Įnabling this massive memory bandwidth is High Bandwidth Memory (HBM) memory, a new high-performance memory standard that uses vertically stacked DRAM dies and fast microscopic interconnects called through-silicon vias (TSVs). That's a huge leap on what was already a fast memory subsystem using high-speed GDDR5 6Gbps memory. Whereas the recent R9 390X/390 were architecturally identical to the R9 290X/290 in every conceivable way, the R9 Fury X (codenamed 'Fiji XT') employs the GCN 1.2 architecture as the R9 380 but doubles the SPU count, which also amounts to 45% more SPUs than you'll find in the R9 390X.Īlong with packing an impressive core specification, the Fury X touts an incredible memory bandwidth of 512GB/s, which is 33% faster than that of the R9 390X and 60% higher than the R9 290X. On the heels of delivering its latest round of rebadges with the Radeon 300 series, AMD has launched what may be its most hyped product of 2015.