NVIDIA GeForce GTX 560 vs NVIDIA Tesla C2070 Comparison

The NVIDIA Tesla C2070 and NVIDIA GeForce GTX 560 are both end-of-life Fermi-based graphics cards from 2011, but they target entirely different workloads. In the single head-to-head benchmark available, the Tesla C2070 wins decisively, but the data reveals a more nuanced story about compute throughput versus memory capacity versus raw texture performance. The C2070 is the compute-oriented card, while the GTX 560 is the consumer-oriented card, and their specifications confirm this split.

Where Each One Wins

The benchmark data gives the Tesla C2070 a clear victory in the only direct comparison available. In the Geekbench OpenCL test, the C2070 scores 9716 against the GTX 560's 9058, a delta of 7.3 percent. This is the sole head-to-head result, and the C2070 wins it. However, this does not mean the GTX 560 is without merit; it wins in other, non-benchmark categories that matter for specific use cases.

The Tesla C2070 wins where compute precision and memory capacity are paramount. Its 448 shading units, 6 GB of GDDR5 memory on a 384-bit bus, and 143.4 GB/s of bandwidth make it the superior choice for large datasets and compute-heavy tasks. The C2070's FP32 performance of 1,027.7 GFLOPS is substantial, and its 48 ROPs deliver a pixel rate of 16.07 GPixel/s, which is higher than the GTX 560's 11.34 GPixel/s. For scientific computing, simulation, or any workload that requires holding large working sets in memory, the C2070 is the data-driven winner.

The GTX 560 wins where texture throughput and power efficiency are more important. Its texture rate of 45.36 GTexel/s is significantly higher than the C2070's 32.14 GTexel/s, despite having the same 56 TMUs. This is because the GTX 560's memory clock is higher at 1000 MHz (4 Gbps effective) versus the C2070's 747 MHz (3 Gbps effective). The GTX 560 also has a lower TDP of 150 W compared to the C2070's 238 W, and it requires a less demanding 450 W suggested PSU versus 550 W. For gaming or real-time graphics where texture fill is the bottleneck, the GTX 560's architecture is better suited.

Architecture Differences

Both cards are built on NVIDIA's Fermi architecture, but they represent two distinct iterations. The Tesla C2070 uses the GF100 chip, which is the original Fermi design, while the GTX 560 uses the GF114 chip, which is the refined "Fermi 2.0" design. Both are manufactured by TSMC on a 40 nm process with a transistor density of 5.9M / mm², but the chip sizes differ dramatically. The GF100 die is 529 mm² and contains 3,100 million transistors, while the GF114 die is 332 mm² with 1,950 million transistors. This means the C2070 has a much larger, more complex chip that is inherently more expensive to produce.

The memory subsystems are also fundamentally different. The C2070 has 6 GB of GDDR5 on a 384-bit bus, yielding 143.4 GB/s of bandwidth. The GTX 560 has only 1024 MB (1 GB) of GDDR5 on a 256-bit bus, yielding 128.0 GB/s. While the bandwidth difference is modest (about 12 percent), the memory capacity difference is sixfold, which is the defining architectural gap for compute workloads. The C2070's memory clock is lower at 747 MHz versus the GTX 560's 1000 MHz, but the wider bus compensates.

The compute core counts differ as well. The C2070 has 448 shading units, 56 TMUs, and 48 ROPs. The GTX 560 has 336 shading units, 56 TMUs, and 32 ROPs. The TMU count is identical, but the C2070 has 33 percent more shading units and 50 percent more ROPs. This explains why the C2070's FP32 throughput is 1,027.7 GFLOPS despite a lower clock, while the GTX 560 achieves 1,088.6 GFLOPS with fewer cores but a higher clock. The GTX 560's texture rate is higher because its shading units operate faster.

Head-to-Head Benchmarks

The only benchmark in the dataset is Geekbench OpenCL, and the result is unambiguous. The Tesla C2070 scores 9716, while the GTX 560 scores 9058. The winner is the C2070 by a margin of 7.3 percent. This is a meaningful gap in a compute benchmark that exercises the GPU's general-purpose processing capabilities, including memory bandwidth and shader throughput.

The C2070's victory is consistent with its position in the nearest rivals list. Its average benchmark score of 9716 places it just 0.1 percent behind the NVIDIA Tesla M10 (9724) and 0.5 percent ahead of the NVIDIA Quadro P4000 (9665). It is also 0.7 percent ahead of the AMD Radeon Pro WX 2100 (9653) and 0.7 percent behind the NVIDIA GeForce GTX 1070 (9780). The GTX 560, with a score of 9058, sits at the 45th percentile of all GPUs, while the C2070 sits at the 47th percentile. The GTX 560's nearest rivals include the NVIDIA TITAN V CEO Edition (9037, 0.2 percent behind), the NVIDIA GeForce GTX 660 (9022, 0.4 percent behind), and the AMD Radeon 550X (8918, 1.6 percent behind).

The deltaPct values in the head-to-head show the C2070 is 7.3 percent faster than the GTX 560. This is a larger gap than any of the rival deltas for either card, indicating that the difference between these two specific cards is more pronounced than the difference between either card and its nearest competitors. The data shows that the C2070 is not just marginally better; it is substantially better in compute workloads.

The Verdict

The data points to a clear conclusion: the NVIDIA Tesla C2070 is the superior card for compute-intensive tasks, while the NVIDIA GeForce GTX 560 is the better choice for texture-bound workloads and lower power envelopes. If the primary use case is OpenCL compute, scientific simulation, or any task that benefits from 6 GB of memory and 448 shading units, the C2070 is the definitive pick. Its 7.3 percent benchmark lead over the GTX 560 is backed by a 50 percent larger ROP count and a wider memory bus.

For users who prioritize texture fill rate and power efficiency, the GTX 560 is the better option. Its 45.36 GTexel/s texture rate is 41 percent higher than the C2070's 32.14 GTexel/s, and its 150 W TDP is 37 percent lower than the C2070's 238 W. The GTX 560 also has a smaller physical footprint at 210 mm (8.3 inches) versus the C2070's 248 mm (9.8 inches), making it easier to fit in smaller cases. The GTX 560's launch MSRP was 199 USD, but no pricing data is available for the C2070.

Neither card is suitable for modern gaming at high settings due to their age and limited DirectX support (both support DirectX 12 with feature level 11_0). However, for legacy compute tasks or as a low-cost parallel processing unit, the C2070's memory capacity is a decisive advantage. The GTX 560's higher FP32 throughput (1,088.6 GFLOPS) is a minor point in its favor, but it does not overcome the C2070's memory and ROP advantages in the benchmark data.

FAQ

Q: Which card is faster in the Geekbench OpenCL benchmark?

A: The NVIDIA Tesla C2070 is faster, scoring 9716 versus the GTX 560's 9058, a 7.3 percent advantage.

Q: Do both cards use the same architecture?

A: Both use Fermi, but the C2070 uses the original GF100 chip (Fermi), while the GTX 560 uses the GF114 chip (Fermi 2.0).

Q: How much memory does each card have?

A: The Tesla C2070 has 6 GB of GDDR5 memory, while the GTX 560 has 1024 MB (1 GB) of GDDR5 memory.

Q: Which card has a higher texture fill rate?

A: The GTX 560 has a higher texture rate of 45.36 GTexel/s, compared to the C2070's 32.14 GTexel/s, despite both having 56 TMUs.

Q: What are the power requirements for each card?

A: The Tesla C2070 has a 238 W TDP and requires a 550 W suggested PSU, while the GTX 560 has a 150 W TDP and requires a 450 W suggested PSU.

Q: Which card has more shading units?

A: The Tesla C2070 has 448 shading units, while the GTX 560 has 336 shading units.

Specification Differences

| Specification | NVIDIA Tesla C2070 | NVIDIA GeForce GTX 560 |

|----------------|---------------------|-------------------------|

| Chip | GF100 | GF114 |

| Architecture | Fermi | Fermi 2.0 |

| Generation | Tesla Fermi (x20xx) | GeForce 500 |

| Transistors | 3,100 million | 1,950 million |

| Die Size | 529 mm² | 332 mm² |

| Memory Clock | 747 MHz (3 Gbps effective) | 1000 MHz (4 Gbps effective) |

| Memory Size | 6 GB | 1024 MB |

| Memory Bus Width | 384 bit | 256 bit |

| Memory Bandwidth | 143.4 GB/s | 128.0 GB/s |

| Shading Units | 448 | 336 |

| ROPs | 48 | 32 |

| Pixel Rate | 16.07 GPixel/s | 11.34 GPixel/s |

| Texture Rate | 32.14 GTexel/s | 45.36 GTexel/s |

| FP32 Performance | 1,027.7 GFLOPS | 1,088.6 GFLOPS |

| TDP | 238 W | 150 W |

| Power Connectors | 1x 6-pin + 1x 8-pin | 2x 6-pin |

| Suggested PSU | 550 W | 450 W |

| Display Outputs | 1x DVI | 2x DVI, 1x mini-HDMI 1.3a |

| Dimensions | 248 mm (9.8 inches) | 210 mm (8.3 inches) |

| Release Date | 2011-07-24 | 2011-05-16 |

| Launch MSRP | Not available | 199 USD |