On March 24, 2011, having refined the Fermi architecture with the mature GF110 silicon, NVIDIA unleashed its ultimate enthusiast statement: the GeForce GTX 590 dual-GPU flagship graphics card. Concurrently, NVIDIA rolled out CUDA 4.0, establishing unified memory addressing and peer-to-peer multi-GPU communication that redefined the developer landscape.
Engineering the Beast: GeForce GTX 590 Specs
The GeForce GTX 590 fused two fully enabled GF110 GPUs onto a single ultra-rigid 12-layer PCB:
- 1024 CUDA Cores (512 cores per GPU), setting an unprecedented parallel density benchmark for consumer hardware.
- Dual 384-bit Memory Interfaces with 3GB GDDR5 VRAM, achieving an aggregated bandwidth of 327.7 GB/s.
- Custom Dual Vapor Chamber Acoustic Design: Dual copper vapor chambers cooled by a centrally positioned low-noise fan engineered to maintain quiet acoustics under heavy load.
- Single-Card NVIDIA 3D Vision Surround: For the first time, a single NVIDIA card could power three high-refresh stereoscopic 3D displays simultaneously with a fourth accessory monitor.
CUDA 4.0: Breakthroughs in Parallel Programming
The release of CUDA 4.0 was equally transformative for high-performance computing (HPC):
- Unified Virtual Addressing (UVA): Unified system memory and multi-GPU memory into a single continuous address space, freeing software engineers from manual pointer bookkeeping.
- GPUDirect 2.0 Peer-to-Peer Transfers: Enabled direct data transfers between GPUs over PCIe without passing through system host memory, drastically slashing communication latencies.
- Thrust Parallel Algorithms Library and Enhanced C++ Support: Lowered the barrier for software engineers to harness parallel acceleration without mastering low-level assembly.
Historical Perspective: Monument of Multi-GPU Scalability
The GeForce GTX 590 remains an enduring icon of the dual-GPU golden age. Simultaneously, CUDA 4.0’s peer-to-peer breakthroughs set the foundational communication primitives that would eventually evolve into modern NVLink clusters powering global AI data centers.
Support Independent Perspectives & In-Depth Insights
Every thoughtful analysis and candid critique comes from our dedication to truth and quality. We choose not to follow sensational algorithms or clickbait headlines.
Sustaining independent research requires reader support. Make a one-time or monthly contribution, securely processed by Google.
Payments secured by Google · Manage or cancel anytime in your Google Account
Comments