Intel Arc Pro A60 vs NVIDIA Tesla M40 Comparison

The Verdict

The data clearly favors the Intel Arc Pro A60 across the recorded benchmarks. The Intel card wins both head-to-head tests, with a 62% lead in Geekbench OpenCL and a 28.2% lead in Geekbench Vulkan. For compute-oriented professional workloads that rely on OpenCL or Vulkan, the Arc Pro A60 is the stronger performer. The NVIDIA Tesla M40 retains relevance only in legacy compute environments where Maxwell architecture support is required, particularly for CUDA-based tasks that do not appear in these specific benchmark results. The Arc Pro A60 also demands less system power, with a 130 W TDP versus the Tesla M40's 250 W TDP, and it suggests a 300 W power supply compared to the Tesla M40's 600 W recommendation. The Tesla M40 is end-of-life and lacks display outputs, making it a poor choice for any workstation that needs visual output. The Arc Pro A60 is active in production and includes four DisplayPort 2.0 outputs. The choice is straightforward: for current professional workloads, the Arc Pro A60 wins on performance, efficiency, and usability. The Tesla M40 is only defensible for specific legacy deployments where its compute capabilities are already integrated into existing infrastructure.

Architecture Differences

The two GPUs come from different architectural generations and process nodes. The Intel Arc Pro A60 uses the Xe-HPG architecture on the DG2-256 chip, fabricated on a 6 nm process at TSMC. The NVIDIA Tesla M40 uses Maxwell 2.0 architecture on the GM200 chip, fabricated on a 28 nm process, also at TSMC. This process gap is significant: the Arc Pro A60 packs 11,500 million transistors into a 269 mm² die, yielding a transistor density of 42.8 million per mm². The Tesla M40 has 8,000 million transistors on a much larger 601 mm² die, giving a density of only 13.3 million per mm². The Intel chip fits more than three times the transistors per area.

Clock behavior differs substantially. The Arc Pro A60 has a base clock of 900 MHz and a boost clock of 2050 MHz. The Tesla M40 has a base clock of 948 MHz and a boost clock of 1112 MHz. The Intel boost clock is nearly double that of the Tesla card. Memory technology also differs: the Arc Pro A60 uses 12 GB of GDDR6 on a 192-bit bus, with memory clocked at 2000 MHz (16 Gbps effective) and bandwidth of 384.0 GB/s. The Tesla M40 also has 12 GB, but it is GDDR5 on a 384-bit bus, with memory at 1502 MHz (6 Gbps effective) and bandwidth of 288.4 GB/s. The Intel card achieves higher bandwidth with a narrower bus thanks to faster memory.

Compute resources are arranged differently. The Arc Pro A60 has 2048 shading units, 128 TMUs, and 64 ROPs, plus 16 ray tracing cores. The Tesla M40 has 3072 shading units, 192 TMUs, and 96 ROPs, with no ray tracing cores. Higher raw counts on the Tesla card do not translate to better benchmark performance. Feature support also diverges: the Arc Pro A60 supports DirectX 12 Ultimate (12_2), while the Tesla M40 only supports DirectX 12 (12_1). Both support OpenGL 4.6 and Vulkan 1.4. The Arc Pro A60 uses PCIe 4.0 x16, while the Tesla M40 uses PCIe 3.0 x16.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The Intel Arc Pro A60 has an average benchmark score of 60326, placing it in the 88th percentile of all GPUs. The NVIDIA Tesla M40 has an average score of 41897, placing it in the 83rd percentile.

Q: How large is the performance gap in OpenCL?

A: In Geekbench OpenCL, the Intel Arc Pro A60 scores 63485 versus the Tesla M40's 39192, a difference of 62%. This is the largest margin between the two cards in any recorded test.

Q: Does the Tesla M40 support ray tracing?

A: No. The Tesla M40 has no ray tracing cores listed in the database. The Intel Arc Pro A60 includes 16 ray tracing cores.

Q: What are the power supply requirements for each card?

A: The Intel Arc Pro A60 has a 130 W TDP and suggests a 300 W power supply. The NVIDIA Tesla M40 has a 250 W TDP and suggests a 600 W power supply.

Q: Can the Tesla M40 drive displays directly?

A: No. The Tesla M40 has no display outputs. The Intel Arc Pro A60 includes 4x DisplayPort 2.0 outputs.

Q: Which GPU is still in production?

A: The Intel Arc Pro A60 has an active production status and was released on 2023-06-05. The NVIDIA Tesla M40 is end-of-life and was released on 2015-11-09.

Specification Differences

| Specification | Intel Arc Pro A60 | NVIDIA Tesla M40 |

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

| Architecture | Xe-HPG | Maxwell 2.0 |

| Process node | 6 nm | 28 nm |

| Transistors | 11,500 million | 8,000 million |

| Die size | 269 mm² | 601 mm² |

| Transistor density | 42.8M / mm² | 13.3M / mm² |

| Base clock | 900 MHz | 948 MHz |

| Boost clock | 2050 MHz | 1112 MHz |

| Memory type | GDDR6 | GDDR5 |

| Memory bus width | 192 bit | 384 bit |

| Memory bandwidth | 384.0 GB/s | 288.4 GB/s |

| Shading units | 2048 | 3072 |

| TMUs | 128 | 192 |

| ROPs | 64 | 96 |

| Ray tracing cores | 16 | None |

| FP32 performance | 8.397 TFLOPS | 6.832 TFLOPS |

| FP16 performance | 16.79 TFLOPS (2:1) | Not listed |

| TDP | 130 W | 250 W |

| Slot width | Single-slot | Dual-slot |

| Power connectors | Not listed | 8-pin EPS |

| Suggested PSU | 300 W | 600 W |

| Bus interface | PCIe 4.0 x16 | PCIe 3.0 x16 |

| Display outputs | 4x DisplayPort 2.0 | No outputs |

| DirectX support | 12 Ultimate (12_2) | 12 (12_1) |

| Production status | Active | End-of-life |

| Release date | 2023-06-05 | 2015-11-09 |

| Predecessor | Not listed | Tesla Kepler |

| Successor | Not listed | Tesla Pascal |

Head-to-Head Benchmarks

The Geekbench OpenCL test shows the most dramatic divergence. The Intel Arc Pro A60 scores 63485, while the NVIDIA Tesla M40 scores 39192. This 62% advantage is substantial and indicates that the Intel architecture handles OpenCL compute workloads far more efficiently. The Tesla M40's higher shading unit count (3072 versus 2048) does not compensate for its older architecture and slower clocks. The Intel card's boost clock of 2050 MHz, combined with its 8.397 TFLOPS FP32 output, drives this result. The Tesla M40's FP32 throughput is 6.832 TFLOPS, which is roughly 19% lower, yet the benchmark gap is much larger at 62%. This suggests architectural efficiency differences beyond raw compute throughput.

The Geekbench Vulkan test narrows the gap but still favors Intel. The Arc Pro A60 scores 57166, and the Tesla M40 scores 44602, a 28.2% difference. Vulkan support is listed as 1.4 for both cards, so the disparity comes from hardware capabilities rather than API version. The Arc Pro A60's modern Xe-HPG architecture with dedicated ray tracing cores likely contributes to better Vulkan performance. The Tesla M40, despite having more TMUs (192 versus 128) and more ROPs (96 versus 64), cannot match the Intel card's newer memory subsystem. The Tesla M40's texture rate is 213.5 GTexel/s, and its pixel rate is 106.8 GPixel/s. The Intel card achieves 262.4 GTexel/s and 131.2 GPixel/s, respectively. These throughput advantages align with the Vulkan results.

The Intel card wins 2 benchmarks, and the Tesla M40 wins 0. No recorded test favors the NVIDIA card. The average benchmark scores reflect this trend: 60326 for Intel versus 41897 for NVIDIA. The percentile rankings reinforce the gap, with Intel at the 88th percentile and NVIDIA at the 83rd percentile. The nearest rival data for the Intel card shows it sits between the AMD Radeon Pro Vega 48 (60140, 0.3% slower) and the AMD Radeon Pro W6600M (61896, 2.5% faster). The Tesla M40's nearest rivals include the NVIDIA Tesla M40 24 GB (41707, 0.5% slower) and the AMD Radeon Pro 5300 (40870, 2.5% slower).

Where Each One Wins

The Intel Arc Pro A60 wins in every recorded benchmark category. Its OpenCL advantage of 62% makes it the clear choice for compute-heavy workloads that use OpenCL. Its Vulkan advantage of 28.2% covers graphics and compute tasks that leverage the Vulkan API. The card's higher FP32 throughput (8.397 TFLOPS versus 6.832 TFLOPS) and FP16 support (16.79 TFLOPS at 2:1 ratio) indicate strong performance for single-precision and half-precision compute. The Tesla M40 has no listed FP16 capability. The Arc Pro A60 also wins on power efficiency, with a 130 W TDP versus 250 W, and a suggested PSU of 300 W versus 600 W. This makes the Intel card suitable for workstations with modest power supplies and single-slot installations.

The NVIDIA Tesla M40 has no benchmark wins, but the data still defines its niche. It has a higher base clock (948 MHz versus 900 MHz), more shading units (3072 versus 2048), more TMUs (192 versus 128), more ROPs (96 versus 64), and a wider memory bus (384 bit versus 192 bit). For code that was written specifically for Maxwell architecture and CUDA, the Tesla M40 may still function in legacy pipelines. The database does not include CUDA benchmarks, so any CUDA advantage remains unquantified. The Tesla M40's 12 GB of GDDR5 memory matches the Arc Pro A60's 12 GB of GDDR6, so memory capacity is equal. The Tesla M40 is dual-slot and has no display outputs, which limits its deployment to compute-only servers. Its end-of-life status and 250 W TDP make it less attractive for new builds. The Intel card is the winner for any new deployment where OpenCL or Vulkan performance matters. The Tesla M40 is only relevant for maintaining existing Maxwell-based systems.