The block diagram for AMD’s next generation R9 290 series “Hawaii” chip has been leaked by EasyCom in their review of the R9 280X. Recently we covered the R9 Series in detail and talk about the specifications, but until now very little was known about the Hawaii chip itself. Now we will get to see the architectural changes the Hawaii chip has made over its predecessor Tahiti.
The Hawaii chip itself is 438mm2 which is 24% larger than the Tahiti chip, which was 389mm2. The block diagram you see above shows us that the core consists of 8 Asynchronous Compute Engines (ACE) modules. This is twice as many as were on the Tahiti chip. In the middle of these modules is the main graphics command processor that allows global data share. There are 4 Shader Engines with their own parallel linked geometry processors and rasterizers. Each Shader Engine will hold 44 Compute Units with 64 Stream Processors. The full blown Hawaii GPU will come with 44 Compute Units, 2816 stream processors, 176 texture mapping units and 64 raster operators.
All of the shader engines share the same pool of 16 64 KB L2 cache with up to 1 TB/s L2/L1 bandwidth. The chip also has 8 64-bit memory controllers which result in the 512-bit bus interface with the 4 GB of GDDR5 memory. Multimedia processors are featured along side the main GEs that include VCE, UVD, DSPs for TrueAudio, XDMA, Eyefinity Controllers and the DMA engine. A small PCI-e 3.0 Bus interface is located which enables support with the latest PCI-e 3.0 enabled motherboards.
Looking at the GCN 2.0 architecture efficiency slide we can see the main improvements between Tahiti and Hawaii. Geometry processing has been increased by 1.9x, compute by 1.3x, texture fill rate by 1.3x, pixel fill rate by 1.9x, peak bandwidth by 1.2x, die area by 1.24x and peak GLOPS by 1.05x.
The Hawaii R9 290 Series will be launching this upcoming week so say tuned!