AMD Radeon Pro 455 vs NVIDIA Tesla M2090 Comparison

The NVIDIA Tesla M2090 and AMD Radeon Pro 455 are two very different GPUs from different eras, aimed at different segments of the market. The Tesla M2090 is a dual-slot compute accelerator from NVIDIA’s Fermi generation, built for high-throughput server workloads, while the Radeon Pro 455 is a low-power MXM module from AMD’s GCN 4.0 family, designed for mobile workstations. The benchmark data shows a clear head-to-head result in OpenCL, but the broader picture is one of contrasting strengths: the Tesla leads in raw compute throughput, while the Radeon Pro 455 counters with modern API support and dramatically lower power draw. This analysis breaks down the numbers, the architecture, and what each card does best.

Head-to-Head Benchmarks

The only direct benchmark comparison available is Geekbench OpenCL, and here the NVIDIA Tesla M2090 comes out ahead. The Tesla scores 13075, while the AMD Radeon Pro 455 scores 10336. That is a delta of 26.5% in favor of the NVIDIA card. This is not a small gap; it represents a substantial lead in general-purpose compute performance as measured by OpenCL. For tasks like physics simulation, data processing, or any workload that leverages OpenCL, the Tesla M2090 is clearly the stronger performer.

Looking at the Tesla’s position among its nearest rivals reinforces this result. Its average benchmark score of 13075 places it just 0.7% above the NVIDIA GeForce GTX 1660 SUPER (12986) and 1% above the NVIDIA GeForce RTX 3050 Ti Mobile (12940). It also edges out the AMD Radeon RX 580 (12928) by 1.1%. The only rival it trails is the NVIDIA GeForce GTX 950 (13189), by 0.9%. This puts the Tesla M2090 in a tight cluster of mid-range cards from much later generations, which is remarkable for a 2011 compute card. The 26.5% advantage over the Radeon Pro 455 in OpenCL is thus not just a win; it places the Tesla in a different performance tier altogether.

The AMD Radeon Pro 455, by contrast, has an average benchmark score of 12831, which sits at the 52nd percentile of all GPUs. Its nearest rivals show a very tight grouping: the NVIDIA GeForce GTX 590 scores 12830 (0% delta), the AMD FirePro W5100 scores 12847 (-0.1%), the AMD Radeon 740M scores 12870 (-0.3%), and the NVIDIA GeForce GTX 670 scores 12773 (0.5%). This indicates that the Radeon Pro 455 is right in the middle of the pack for its performance class. However, its OpenCL score of 10336 is notably lower than its average, suggesting that its best results come from other APIs. Indeed, the Radeon Pro 455 achieves a Geekbench Metal score of 15916 and a Geekbench Vulkan score of 12240, both of which are higher than its OpenCL result. This shows that the card’s performance is API-dependent, with Metal being its strongest environment.

In the head-to-head, the Tesla wins the only shared test. But the data also reveals that the Radeon Pro 455 has strengths outside of OpenCL. The Tesla’s single benchmark score is its only data point, while the Radeon Pro 455 has three, covering Metal and Vulkan as well. That diversity matters for real-world use, as different applications leverage different APIs.

Where Each One Wins

The NVIDIA Tesla M2090 wins in raw OpenCL compute. With a score of 13075, it is 26.5% faster than the Radeon Pro 455’s OpenCL result of 10336. This makes it the better choice for any workload that is heavily dependent on OpenCL performance, such as scientific computing, financial modeling, or certain rendering tasks that use OpenCL for acceleration. Its 6 GB of GDDR5 memory on a 384-bit bus provides 177.4 GB/s of bandwidth, which is more than double the Radeon Pro 455’s 81.28 GB/s. That memory bandwidth is critical for large datasets, and the Tesla’s 1,332.2 GFLOPS of FP32 performance is slightly higher than the Radeon Pro 455’s 1,313.3 GFLOPS. The Tesla also has a higher pixel rate (20.83 GPixel/s vs 13.68 GPixel/s) and a higher texture rate (41.66 GTexel/s vs 41.04 GTexel/s), though the texture rate difference is marginal.

The AMD Radeon Pro 455 wins in API versatility and power efficiency. It supports DirectX 12 (12_0), Vulkan 1.3, and OpenGL 4.6, while the Tesla M2090 only supports DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan support. The Radeon Pro 455’s Metal score of 15916 is its highest benchmark result, indicating strong performance in Apple’s ecosystem, which is relevant given its “Radeon Pro Mac” generation. Its Vulkan score of 12240 also shows capability in modern cross-platform APIs. The Radeon Pro 455 is also a mobile part: it is an MXM module with a 35 W TDP and no power connectors, making it suitable for laptops. The Tesla M2090, by contrast, is a dual-slot card with a 250 W TDP, requiring a 1x 6-pin + 1x 8-pin power connector and a 600 W suggested PSU. It has no display outputs, so it cannot drive a monitor at all. The Radeon Pro 455’s display outputs are portable device dependent, meaning it can support displays in a laptop context.

The Tesla wins on memory capacity and bandwidth, which matters for large-scale compute. The Radeon Pro 455 wins on power consumption and API support, which matters for mobile workstations and modern software. The Tesla is a server accelerator; the Radeon Pro 455 is a mobile workstation GPU.

Architecture Differences

The two cards are built on fundamentally different architectures and process nodes. The NVIDIA Tesla M2090 uses the GF110 chip, based on the Fermi 2.0 architecture, manufactured on a 40 nm process at TSMC. The die size is 520 mm², with 3,000 million transistors, yielding a transistor density of 5.8M per mm². The AMD Radeon Pro 455 uses the Baffin chip, based on GCN 4.0, manufactured on a 14 nm process at GlobalFoundries. Its die is much smaller at 123 mm², but it also has 3,000 million transistors, giving a much higher transistor density of 24.4M per mm². The process node difference is stark: 40 nm vs 14 nm, which explains the massive disparity in die size for the same transistor count.

The compute units also differ. The Tesla M2090 has 512 shading units, 64 TMUs, and 48 ROPs. The Radeon Pro 455 has 768 shading units, 48 TMUs, and only 16 ROPs. Despite having more shading units, the Radeon Pro 455’s FP32 performance (1,313.3 GFLOPS) is slightly lower than the Tesla’s (1,332.2 GFLOPS), likely due to lower clock speeds and different architecture efficiency. The Radeon Pro 455 also supports FP16 at a 1:1 ratio (1,313.3 GFLOPS), which the Tesla does not list. The Tesla’s memory is 6 GB of GDDR5 on a 384-bit bus with a memory clock of 924 MHz (3.7 Gbps effective), resulting in 177.4 GB/s bandwidth. The Radeon Pro 455 has 2 GB of GDDR5 on a 128-bit bus with a memory clock of 1270 MHz (5.1 Gbps effective), yielding 81.28 GB/s bandwidth. The Radeon Pro 455 has a higher memory clock but a much narrower bus, so its bandwidth is less than half.

The Tesla’s bus interface is PCIe 2.0 x16, while the Radeon Pro 455 uses PCIe 3.0 x8. The Radeon Pro 455’s PCIe 3.0 interface is newer and offers higher per-lane bandwidth, though the x8 width halves the total lanes. The Tesla has no display outputs, while the Radeon Pro 455’s outputs are portable device dependent. In terms of API support, the Radeon Pro 455 is clearly ahead: it supports Vulkan 1.3 and DirectX 12_0, while the Tesla only supports DirectX 11_0 and no Vulkan. Both support OpenGL 4.6. The Radeon Pro 455’s GCN 4.0 architecture is designed for modern graphics APIs, while the Tesla’s Fermi 2.0 is older and compute-oriented.

The release dates also differ significantly: the Tesla M2090 was released on 2011-07-24, while the Radeon Pro 455 was released on 2016-10-29. This five-year gap explains the architectural and process improvements in the AMD part.

FAQ

Q: Which GPU has a higher OpenCL benchmark score?

A: The NVIDIA Tesla M2090 scores 13075 in Geekbench OpenCL, which is 26.5% higher than the AMD Radeon Pro 455’s score of 10336.

Q: Does the AMD Radeon Pro 455 support Vulkan?

A: Yes, the Radeon Pro 455 supports Vulkan 1.3, along with DirectX 12 (12_0) and OpenGL 4.6. The Tesla M2090 does not list Vulkan support.

Q: What is the memory bandwidth difference between the two cards?

A: The Tesla M2090 has 177.4 GB/s of bandwidth from its 6 GB GDDR5 memory on a 384-bit bus. The Radeon Pro 455 has 81.28 GB/s from 2 GB GDDR5 on a 128-bit bus.

Q: Which card consumes more power?

A: The Tesla M2090 has a 250 W TDP and requires a 1x 6-pin + 1x 8-pin power connector, with a suggested PSU of 600 W. The Radeon Pro 455 has a 35 W TDP and uses no power connectors.

Q: Can the Tesla M2090 output video to a display?

A: No, the Tesla M2090 has no display outputs. The Radeon Pro 455’s display outputs are portable device dependent.

Q: What is the transistor density of each chip?

A: The Tesla M2090’s GF110 chip has 3,000 million transistors on a 520 mm² die, giving a density of 5.8M per mm². The Radeon Pro 455’s Baffin chip has 3,000 million transistors on a 123 mm² die, giving a density of 24.4M per mm².

Specification Differences

The table below highlights the key specification differences between the two GPUs.

| Specification | NVIDIA Tesla M2090 | AMD Radeon Pro 455 |

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

| Architecture | Fermi 2.0 | GCN 4.0 |

| Process Node | 40 nm | 14 nm |

| Foundry | TSMC | GlobalFoundries |

| Die Size | 520 mm² | 123 mm² |

| Transistor Density | 5.8M / mm² | 24.4M / mm² |

| Memory Size | 6 GB | 2 GB |

| Memory Bus Width | 384 bit | 128 bit |

| Memory Clock | 924 MHz (3.7 Gbps effective) | 1270 MHz (5.1 Gbps effective) |

| Memory Bandwidth | 177.4 GB/s | 81.28 GB/s |

| Shading Units | 512 | 768 |

| TMUs | 64 | 48 |

| ROPs | 48 | 16 |

| Pixel Rate | 20.83 GPixel/s | 13.68 GPixel/s |

| Texture Rate | 41.66 GTexel/s | 41.04 GTexel/s |

| FP32 Performance | 1,332.2 GFLOPS | 1,313.3 GFLOPS |

| FP16 Performance | Not listed | 1,313.3 GFLOPS (1:1) |

| TDP | 250 W | 35 W |

| Slot Width | Dual-slot | MXM Module |

| Power Connectors | 1x 6-pin + 1x 8-pin | None |

| Suggested PSU | 600 W | Not listed |

| Bus Interface | PCIe 2.0 x16 | PCIe 3.0 x8 |

| Display Outputs | No outputs | Portable Device Dependent |

| DirectX Support | 12 (11_0) | 12 (12_0) |

| Vulkan Support | Not listed | 1.3 |

| Release Date | 2011-07-24 | 2016-10-29 |

| Dimensions (Length) | 248 mm (9.8 inches) | Not listed |