AMD officially announces the non X model of the Radeon R9 Fury Graphics Card. The card has a Fiji Pro GPU. Today is the announcement of the 2nd fastest Fiji GPU . Only two partners will make the card, Sapphire and ASUS.
Radeon R9 Fury uses the Fiji Pro GPU, and it is literally a beast as it is still based on a 28nm fabrication node, that means a chip just over the size of 5 x 5cm. There's lots of good stuff going on inside that chip as, the memory you guys all know as GDDR5 typically has been seated on the graphics card PCB. Well, with the Radeon R9 Fury that has changed. AMD has made a bold move to HBM memory (we'll talk about it over the next pages), the 4 GB of memory now is seated onto the actual GPU (chip).
Radeon R9 Fury X | Radeon R9 Fury | Radeon R9 390X | |
Fabrication Process | 28nm | 28nm | 28nm |
GPU | Fiji | Fiji | Hawaii / Grenada |
Streaming Processors | 4096 | 3584 | 2816 |
Graphics memory | 4 GB HBM | 4 GB HBM | 8 GB GDDR5 |
Memory Clock | up-to 500 MHz / 1.0 Gbps | up-to 500 MHz / 1.0 Gbps | 6.0 Gbps |
Core Clock | 1050 MHz | 1000 MHz | 1050 MHz |
Memory Bandwidth | up-to 512 GB/s | up-to 512 GB/s | 384 GB/s |
Power Connectors | 2 x 8-pin | 2 x 8-pin | 1 x 6-pin - 1 x 8-pin |
Form Factor | Full Height, Dual slot | Full Height, Dual slot | Full Height, Dual slot |
Freesync | Yes | Yes | Yes |
DirectX 12 Support | Yes | Yes | Yes |
So, try to comprehend this, the Fiji Xt GPU has 8.9 Billion transistors, and that is EXCLUDING the HBM memory chips, I know..! crazy figures right? To be able to fit all that on 28nm, well it's impressive to say the least. The GPU itself (and we'll talk in detail about it on the following pages) is based on GCN 1.2 architecture and then scaled upwards, the Fiji pro products will have 3584 shader processors whereas that is 4096 shader processors for the X model. The end result is a product with 7.2 TFLOPS of compute performance offering nice game performance in the more difficult and complex to render situations like Ultra HD gaming. As stated the product comes with a Fiji pro GPU that is a cut-down Fiji XT chip with 56 compute units and thus 3584 stream processors, 64 Raster operation units and 224 texture mapping units. The base clock for the reference AMD Radeon R9 Fury is 1000 MHz. The card chunks out 7.2 TFlops of compute performance. The card obviously has 4GB of HBM memory left in-tact at 500 MHz / 1.0 Gbps (effective data rate).
You will notice that the ASUS card has a DVI connector, if you look closer you'll see it is a single link one, so basically that is just a converted HDMI -> DVI with a max res of 1920x1080. The cards are expected to cost 549 USD.
We will have a review for you guys next week as our sample has not arrived just yet.