NVIDIA GeForce GTX 580 vs NVIDIA GeForce GTX 690 Comparison

FAQ

Q: How much faster is the GTX 690 than the GTX 580 in the recorded OpenCL benchmark?

A: The GTX 690 scores 17,399, while the GTX 580 scores 15,283. That gives the GTX 690 a 13.8% lead in the Geekbench OpenCL test.

Q: Which card has the higher average benchmark score?

A: The GTX 690 has an average benchmark score of 17,037. The GTX 580 has an average benchmark score of 15,283. The GTX 690 sits at the 60th percentile of all GPUs, while the GTX 580 sits at the 58th percentile.

Q: How do the memory systems compare between the two cards?

A: The GTX 690 has 2 GB of GDDR5 memory on a 256-bit bus with 192.3 GB/s bandwidth. The GTX 580 has 1536 MB of GDDR5 memory on a 384-bit bus with 192.4 GB/s bandwidth. The bandwidth figures are nearly identical, but the GTX 690 has a larger total memory pool.

Q: What are the power requirements for each card?

A: The GTX 690 has a TDP of 300 W and requires a 700 W power supply, using two 8-pin connectors. The GTX 580 has a TDP of 244 W and requires a 550 W power supply, using one 6-pin and one 8-pin connector.

Q: Does the GTX 580 support Vulkan?

A: No. The GTX 690 supports Vulkan 1.2.175, but the GTX 580 has no Vulkan support listed. Both cards support DirectX 12 (11_0) and OpenGL 4.6.

Q: Which card has more shading units and texture mapping units?

A: The GTX 690 has 1,536 shading units and 128 TMUs. The GTX 580 has 512 shading units and 64 TMUs. The GTX 690 has triple the shading units and double the TMUs.

Where Each One Wins

The GTX 690 wins the only head-to-head benchmark available in the database: Geekbench OpenCL, with a 13.8% advantage over the GTX 580. The recorded data shows a single win for the GTX 690 and zero wins for the GTX 580, so the performance split is clear.

However, the GTX 580 holds its own in specific architectural areas. It has more ROPs (48 versus 32) and a wider memory bus (384-bit versus 256-bit). These characteristics favor fill-rate-bound workloads and high-resolution rendering with heavy overdraw. The GTX 580 also consumes significantly less power (244 W versus 300 W), making it a more power-efficient choice for systems with smaller power supplies.

The GTX 690 is the clear winner for compute-heavy tasks. Its FP32 throughput is 3.130 TFLOPS versus 1.581 TFLOPS for the GTX 580, nearly double. Its texture rate is 130.4 GTexel/s versus 49.41 GTexel/s, and its pixel rate is 32.61 GPixel/s versus 24.70 GPixel/s. For general compute, physics simulation, and texture-heavy rendering, the GTX 690 dominates.

The GTX 580 also has a smaller physical footprint: 267 mm in length versus 279 mm for the GTX 690. Both cards are dual-slot and have the same height of 111 mm.

Architecture Differences

The GTX 690 is built on the Kepler architecture, specifically using the GK104 chip. The GTX 580 uses the Fermi 2.0 architecture with the GF110 chip. This architectural gap is substantial.

The GTX 690 is fabricated on a 28 nm process at TSMC, while the GTX 580 uses a 40 nm process at the same foundry. The smaller node allows the GTX 690 to pack 3,540 million transistors into a 294 mm² die, yielding a transistor density of 12.0M per mm². The GTX 580 has 3,000 million transistors on a much larger 520 mm² die, giving it a density of 5.8M per mm².

Kepler introduces a different shader organization compared to Fermi 2.0. The GTX 690 has 1,536 shading units, while the GTX 580 has 512. This is a direct consequence of the architectural shift. The GTX 690 also has 128 TMUs versus 64 TMUs on the GTX 580. However, the GTX 580 has more ROPs: 48 versus 32. This means the GTX 580 is better at final pixel output per clock, while the GTX 690 is better at geometry and texture processing.

The memory architecture also differs. The GTX 690 uses a 256-bit bus, while the GTX 580 uses a 384-bit bus. The memory clock differs as well: the GTX 690 runs at 6 Gbps effective, while the GTX 580 runs at 4 Gbps effective. The net effect is nearly identical bandwidth: 192.3 GB/s versus 192.4 GB/s.

The GTX 690 supports PCIe 3.0 x16, while the GTX 580 supports PCIe 2.0 x16. The GTX 690 also supports Vulkan 1.2.175, while the GTX 580 has no Vulkan support.

Specification Differences

The two cards differ across nearly every specification field:

• Process node: 28 nm (GTX 690) versus 40 nm (GTX 580)
• Transistors: 3,540 million versus 3,000 million
• Die size: 294 mm² versus 520 mm²
• Transistor density: 12.0M / mm² versus 5.8M / mm²
• Base clock: 915 MHz (GTX 690) versus no base clock listed (GTX 580)
• Boost clock: 1019 MHz (GTX 690) versus no boost clock listed (GTX 580)
• Memory clock: 1502 MHz / 6 Gbps effective versus 1002 MHz / 4 Gbps effective
• Memory size: 2 GB versus 1536 MB
• Memory bus: 256 bit versus 384 bit
• Shading units: 1536 versus 512
• TMUs: 128 versus 64
• ROPs: 32 versus 48
• Pixel rate: 32.61 GPixel/s versus 24.70 GPixel/s
• Texture rate: 130.4 GTexel/s versus 49.41 GTexel/s
• FP32: 3.130 TFLOPS versus 1.581 TFLOPS
• TDP: 300 W versus 244 W
• Power connectors: 2x 8-pin versus 1x 6-pin + 1x 8-pin
• Suggested PSU: 700 W versus 550 W
• Bus interface: PCIe 3.0 x16 versus PCIe 2.0 x16
• Display outputs: 3x DVI + 1x mini-DisplayPort 1.2 versus 2x DVI + 1x mini-HDMI 1.3a
• Vulkan: 1.2.175 versus none
• Length: 279 mm versus 267 mm
• Width: 38 mm versus not specified

Both cards share the same height (111 mm), slot width (Dual-slot), DirectX version (12 with 11_0 feature level), and OpenGL version (4.6). Both are end-of-life products from NVIDIA.

Head-to-Head Benchmarks

The database contains a single head-to-head benchmark between these two cards: Geekbench OpenCL. The GTX 690 scores 17,399, and the GTX 580 scores 15,283. This gives the GTX 690 a 13.8% advantage.

This margin reflects the architectural gap. The GTX 690 has nearly double the FP32 throughput (3.130 TFLOPS versus 1.581 TFLOPS), which is a major factor in OpenCL compute workloads. Its texture rate is also more than double (130.4 GTexel/s versus 49.41 GTexel/s), which helps in workloads that sample textures heavily.

On the other side, the GTX 580 has a higher ROP count (48 versus 32), which historically benefits pixel-heavy workloads. Its memory bandwidth is essentially identical (192.4 GB/s versus 192.3 GB/s), so the GTX 580 could close the gap in scenarios that are bandwidth-bound rather than compute-bound. The GTX 580 also has a lower TDP of 244 W versus 300 W, which may allow it to sustain boost behavior under thermal constraints in some systems.

However, the recorded data shows only one benchmark, and the GTX 690 won it. The delta of 13.8% is significant but not overwhelming. It suggests that while the GTX 690 is the faster card overall, the GTX 580 can still be competitive in specific workloads that favor its ROP count and memory bus width.

The average benchmark scores confirm the hierarchy: the GTX 690 averages 17,037, while the GTX 580 averages 15,283. The GTX 690 sits at the 60th percentile of all GPUs, while the GTX 580 sits at the 58th percentile. In the nearest rival context, the GTX 690 is within 1% of the RTX 3070 (which averages 17,208) and within 0.3% of the RX 7600 XT (which averages 17,083). The GTX 580 is within 0.1% of the RTX 2060 (which averages 15,290) and within 0.7% of the RX 7600 (which averages 15,171).

The data indicates that the GTX 690 is the stronger performer in compute applications, while the GTX 580 retains some unique strengths in pixel throughput and power efficiency. For buyers looking at legacy hardware, the GTX 690 offers the higher ceiling, but the GTX 580 remains a viable alternative in scenarios where ROP count and power draw matter more.