GPU Wiki Architecture

GCN 2.0

GCN 2.0 (second-generation Graphics Core Next) arrived in 2013 with the R9 200 series. Hawaii (R9 290X / R9 290) was the flagship die, a large 28nm chip delivering competitive performance against NVIDIA's Kepler at high resolutions.

Launch era
2013 to 2015
Process nodes
28nm (TSMC)
Predecessor
GCN 1.0
Successor
GCN 3.0
Source review
Partial

Related GPUs

25 entries
View
Reset filters
GPU Architecture Launch Date Shader Processors Memory Bus Width TDP/TBP Segment
Radeon R9 360 OEM GCN 2.0 2015-05-05 768 2 GB GDDR5 128-bit 75 W Gaming
Radeon R7 260 GCN 2.0 2013-12-17 768 2 GB GDDR5 128-bit 95 W Gaming
Radeon HD 8770 OEM GCN 2.0 2013-09-02 896 1 GB GDDR5 128-bit 85 W Gaming
Radeon HD 7790 GCN 2.0 2013-03-22 896 1 GB GDDR5 128-bit 85 W Gaming
FirePro W5100 GCN 2.0 2014-03-31 768 4 GB GDDR5 128-bit 50 W Workstation
FirePro W9100 GCN 2.0 2014-03-26 2,816 16 GB GDDR5 512-bit 275 W Workstation
FirePro S9170 GCN 2.0 2015-07-08 2,816 32 GB GDDR5 512-bit 275 W Data Center
FirePro S9100 GCN 2.0 2014-10-02 2,560 12 GB GDDR5 512-bit 225 W Data Center
FirePro S9150 GCN 2.0 2014-08-07 2,816 16 GB GDDR5 512-bit 235 W Data Center
FirePro W4300 GCN 2.0 2015-12-01 768 4 GB GDDR5 128-bit 50 W Workstation

Architecture History

GCN 2.0 refined the original GCN design with improved compute throughput, doubled FP64 throughput on professional variants, and added TrueAudio DSP hardware on certain SKUs.

More architecture facts

Architecture name
GCN 2.0
Important chips
Hawaii (R9 290X / R9 290), Tonga (R9 285 / 380), Bonaire (R7 260X), Hainan
Memory technologies
GDDR5
Architecture feature generation
GCN 2.0 Stream Processors; DirectX 11.2; OpenCL 2.0 (Tonga); TrueAudio DSP on select SKUs
Consumer series
Radeon R9 290X, R9 290, R9 285, R7 260X, R7 260
Workstation / professional series
FirePro W9100 (Hawaii-based, 16 GB HBM — actually GCN 3?), FirePro W8100
Data center series
FirePro S9150 (Hawaii compute)

What this architecture changed

Hawaii die: largest GCN die to date at launch with 2816 Stream Processors and 512-bit GDDR5 memory bus.
Improved geometry throughput and tessellation performance over GCN 1.0.
Doubled FP64 compute rate on Hawaii versus GCN 1.0 at the same clock (relevant for workstation/HPC variants).
TrueAudio: dedicated audio DSP on Tonga/Bonaire for hardware-accelerated 3D audio effects.
OpenCL 2.0 support on Tonga (first GCN chip with OpenCL 2.0 compliance).

Why it mattered

GCN 2.0's Hawaii die was a direct, competitive response to NVIDIA's GK110 (GTX780 Ti), trading performance points across the benchmark stack at similar prices. The generation reinforced GCN's viability as a compute platform and kept AMD relevant at the high end of the discrete GPU market through the mid-2010s. Hawaii variants remained in AMD's professional lineup for several years after consumer retirement.