NVIDIA A10G vs NVIDIA GeForce RTX 5090 D Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA A10G has an average benchmark score of 151,963, while the NVIDIA GeForce RTX 5090 D has an average score of 77,712. The difference is significant, but it is important to note that the averages are drawn from different benchmark suites.

Q: How does the A10G compare to its nearest rivals in the database?

A: The A10G sits at the 97th percentile among all GPUs. It is 1.1% ahead of the NVIDIA Tesla V100 PCIe 32 GB, 9.3% ahead of the AMD Instinct MI100, and 5.4% behind the AMD Radeon Pro W6800X. It also trails the NVIDIA A100 PCIe 40 GB by 6.5%.

Q: What is the performance gap in OpenCL and Vulkan between the two cards?

A: In Geekbench OpenCL, the RTX 5090 D scores 310,674 versus 158,063 for the A10G, a 49.1% lead. In Geekbench Vulkan, the RTX 5090 D scores 376,915 versus 145,863, a 61.3% lead. The RTX 5090 D wins both head-to-head tests.

Q: Which card has more memory and bandwidth?

A: The RTX 5090 D has 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The A10G has 24 GB of GDDR6 on a 384-bit bus, delivering 600.2 GB/s. The RTX 5090 D offers roughly three times the bandwidth.

Q: Are both cards built on the same manufacturing process?

A: No. The A10G uses an 8 nm process from Samsung, while the RTX 5090 D uses a 5 nm process from TSMC. The RTX 5090 D also packs more transistors: 92,200 million versus 28,300 million.

Q: What is the power requirement for each card?

A: The A10G has a TDP of 150 W and a suggested PSU of 450 W. The RTX 5090 D has a TDP of 575 W and a suggested PSU of 950 W. The RTX 5090 D is a much more power-hungry card.

Architecture Differences

The two GPUs come from entirely different architectural generations. The NVIDIA A10G is built on the Ampere architecture, using the GA102 chip, and belongs to the Server Ampere generation. The NVIDIA GeForce RTX 5090 D uses the Blackwell 2.0 architecture with the GB202 chip, part of the GeForce 50-series.

Manufacturing technology separates them further. The A10G is fabricated on an 8 nm process at Samsung, with 28,300 million transistors on a 628 mm² die. The RTX 5090 D is built on a 5 nm process at TSMC, with 92,200 million transistors on a 750 mm² die. The transistor density tells the story: the RTX 5090 D packs 122.9 million transistors per mm², compared to 45.1 million per mm² for the A10G.

Compute resources differ dramatically. The A10G has 9,216 shading units, 288 TMUs, and 96 ROPs. The RTX 5090 D has 21,760 shading units, 680 TMUs, and 176 ROPs. Ray tracing hardware also scales: the A10G carries 72 RT cores and 288 tensor cores, while the RTX 5090 D carries 170 RT cores and 680 tensor cores.

Clock speeds show the generational jump. The A10G runs at a base of 1320 MHz and boosts to 1710 MHz. The RTX 5090 D starts at 2017 MHz and boosts to 2407 MHz. Memory clocks differ too: the A10G uses 1563 MHz with 12.5 Gbps effective, while the RTX 5090 D runs at 1750 MHz with 28 Gbps effective.

Both cards support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. However, the RTX 5090 D adds display outputs (1x HDMI 2.1b and 3x DisplayPort 2.1b), while the A10G has no display outputs at all. The A10G is a compute-focused server card, whereas the RTX 5090 D is a dual-slot desktop card with a 16-pin power connector.

Where Each One Wins

The A10G wins in the server and datacenter context. It is a single-slot card with an 8-pin EPS power connector, a 150 W TDP, and a suggested PSU of only 450 W. Its end-of-life production status and 2021 release date place it in the earlier Ampere generation, but its 97th percentile ranking among all GPUs shows it remains a strong compute performer. The A10G is clearly designed for dense rack deployments where power and space are constrained.

The RTX 5090 D wins in raw performance and modern feature support. It holds the 92nd percentile among all GPUs, but its average benchmark score is dragged down by the inclusion of many different tests, including older DirectX 9 and DirectX 10 PassMark results. In the head-to-head tests, the RTX 5090 D leads by 49.1% in OpenCL and 61.3% in Vulkan. It also has more than double the shading units, nearly double the texture units, and a much larger memory bandwidth.

For gaming and workstation use, the RTX 5090 D is the obvious pick. It has display outputs, a PCIe 5.0 interface, and a much higher pixel rate (423.6 GPixel/s versus 164.2 GPixel/s). The A10G, with no display outputs and a PCIe 4.0 interface, is not suited for interactive workloads.

For AI inference and training in a power-limited environment, the A10G has merit. Its 150 W TDP is a fraction of the 575 W TDP of the RTX 5090 D. The A10G also has 24 GB of memory, which is still substantial, though the RTX 5090 D doubles that to 32 GB.

Specification Differences

| Specification | NVIDIA A10G | NVIDIA GeForce RTX 5090 D |

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

| Architecture | Ampere | Blackwell 2.0 |

| Process Node | 8 nm | 5 nm |

| Foundry | Samsung | TSMC |

| Transistors | 28,300 million | 92,200 million |

| Die Size | 628 mm² | 750 mm² |

| Transistor Density | 45.1M / mm² | 122.9M / mm² |

| Base Clock | 1320 MHz | 2017 MHz |

| Boost Clock | 1710 MHz | 2407 MHz |

| Memory Clock | 1563 MHz, 12.5 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 24 GB | 32 GB |

| Memory Type | GDDR6 | GDDR7 |

| Memory Bus | 384 bit | 512 bit |

| Memory Bandwidth | 600.2 GB/s | 1.79 TB/s |

| Shading Units | 9216 | 21760 |

| TMUs | 288 | 680 |

| ROPs | 96 | 176 |

| RT Cores | 72 | 170 |

| Tensor Cores | 288 | 680 |

| Pixel Rate | 164.2 GPixel/s | 423.6 GPixel/s |

| Texture Rate | 492.5 GTexel/s | 1,636.8 GTexel/s |

| FP32 | 31.52 TFLOPS | 104.8 TFLOPS |

| FP16 | 31.52 TFLOPS (1:1) | 104.8 TFLOPS (1:1) |

| TDP | 150 W | 575 W |

| Slot Width | Single-slot | Dual-slot |

| Power Connectors | 8-pin EPS | 1x 16-pin |

| Suggested PSU | 450 W | 950 W |

| Bus Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |

| Display Outputs | No outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

| Length | 267 mm (10.5 inches) | 304 mm (12 inches) |

| Height | 112 mm (4.4 inches) | 137 mm (5.4 inches) |

| Width | Not specified | 48 mm (1.9 inches) |

| Production Status | End-of-life | Active |

| Release Date | 2021-04-11 | 2025-01-29 |

| Predecessor | Tesla Turing | GeForce 40 |

| Successor | Server Ada | GeForce 60 |

Head-to-Head Benchmarks

The database records two direct benchmark comparisons between these cards. Both are Geekbench tests, and the RTX 5090 D wins both.

In Geekbench OpenCL, the RTX 5090 D scores 310,674 against the A10G's 158,063. The delta is 49.1% in favor of the RTX 5090 D. This test measures general compute performance, and the result aligns with the raw specification gap: the RTX 5090 D has more than twice the shading units, a much higher boost clock, and significantly higher FP32 throughput (104.8 TFLOPS versus 31.52 TFLOPS).

In Geekbench Vulkan, the gap widens. The RTX 5090 D scores 376,915, while the A10G manages 145,863. The delta is 61.3% in favor of the RTX 5090 D. Vulkan is a low-level API that can expose hardware differences more directly, and the RTX 5090 D's newer architecture, higher clock speeds, and faster memory subsystem all contribute to this larger lead.

The A10G does not win any head-to-head benchmark in the database. Its wins are relative to its own peer group: it is 1.1% ahead of the Tesla V100 PCIe 32 GB and 9.3% ahead of the AMD Instinct MI100. But against the RTX 5090 D, the data is unambiguous. The RTX 5090 D is faster in every recorded comparison.

The average benchmark score comparison is trickier. The A10G averages 151,963, while the RTX 5090 D averages 77,712. This is because the RTX 5090 D has a much broader benchmark suite, including PassMark DirectX 9, 10, 11, and 12 tests that score in the hundreds, pulling the average down. The A10G only has two Geekbench results, both in the 140,000 to 160,000 range. The average score is not a reliable measure of relative performance here; the head-to-head tests are more meaningful.

The Verdict

The data points to two very different use cases. The NVIDIA A10G is a server accelerator designed for environments where power draw and physical space are at a premium. Its 150 W TDP, single-slot design, and 8-pin EPS connector make it easy to integrate into existing server infrastructure. Its 97th percentile ranking shows it still competes well in its class, and its 24 GB of memory is enough for many inference and training workloads. The A10G is end-of-life, but for a build that prioritizes efficiency and density, it remains a viable option.

The NVIDIA GeForce RTX 5090 D is the performance leader by a wide margin. In the head-to-head benchmarks, it beats the A10G by 49.1% in OpenCL and 61.3% in Vulkan. It has more memory (32 GB), faster memory (GDDR7 at 1.79 TB/s), and a newer architecture (Blackwell 2.0 on a 5 nm process). It also has display outputs, making it suitable for desktop workstations. However, it demands a 575 W TDP and a 950 W suggested PSU, and it is dual-slot with a 16-pin connector.

If the choice is for a compute node in a rack with strict power limits, the A10G is the sensible pick. If the choice is for raw compute throughput, modern API support, or any kind of interactive or rendering workload, the RTX 5090 D is the clear winner. The RTX 5090 D also has a launch MSRP of 2,299 USD, which is the only price data available in the database.

The RTX 5090 D is a newer, faster, and more capable card on nearly every metric. The A10G is a specialized server part that trades performance for efficiency. Choose based on the environment, not just the score.