GPU Wiki Architecture
GCN 1.0
GCN 1.0 (Graphics Core Next, first generation) launched in 2011 with the Radeon HD 7000 series, replacing AMD's VLIW-based TeraScale architecture with a scalar shader design. The shift to GCN marked AMD's move to a compute-oriented GPU architecture that would define the company's discrete lineup for nearly a decade.
- Launch era
- 2011 to 2013
- Process nodes
- 28nm (TSMC)
- Predecessor
- TeraScale 3 (VLIW4)
- Successor
- GCN 2.0
- Source review
- Partial
Related GPUs
Radeon R5 330 OEM
Radeon HD 8730 OEM
Radeon R7 350
Radeon HD 7730M
Radeon R9 255 OEM
Radeon R7 265
Radeon R7 430 OEM
Radeon R7 450 OEM
Radeon R7 240
Architecture History
Tahiti (Radeon HD 7970 / HD 7950) was the flagship, introducing AMD's first 28nm desktop GPU and setting single-precision compute throughput records at launch.
More architecture facts
- Architecture name
- GCN 1.0
- Important chips
- Tahiti (HD 7970 / 7950), Pitcairn (HD 7870 / 7850), Cape Verde (HD 7770 / 7750), Oland (HD 8000M)
- Memory technologies
- GDDR5
- Architecture feature generation
- GCN 1.0 Stream Processors; DirectX 11.1; OpenCL 1.2; first generation AMD GCN compute pipeline
- Consumer series
- Radeon HD 7900 series, HD 7800 series, HD 7700 series
- Workstation / professional series
- FirePro W9000, W8000, W7000 series
- Data center series
- FirePro S9000, S7000 compute cards
What this architecture changed
Shift from VLIW to scalar shader design: each GCN Compute Unit contains 64 stream processors executing scalar instructions rather than the wide VLIW5 bundles of TeraScale.
Asynchronous Compute Engines (ACEs): hardware support for concurrent graphics and compute workloads.
DirectX 11.1 and OpenCL 1.2 compliance.
PCI Express 3.0 support on high-end SKUs.
GDDR5 memory interface across the lineup.
Why it mattered
GCN 1.0 established the shader architecture that AMD used across nearly a decade of consumer and professional GPUs. The scalar design proved highly amenable to general-purpose GPU compute, making GCN-generation hardware a popular choice for compute workloads and contributing to AMD's relevance in the early deep learning GPU era. Driver support for GCN 1.0 hardware remained in the standard Adrenalin stack longer than most architectures of the same age.
