NVIDIA Quadro RTX 8000 vs NVIDIA Tesla M60 Comparison

The NVIDIA Quadro RTX 8000 decisively outperforms the NVIDIA Tesla M60 in every direct benchmark comparison, with the data showing a generational leap rather than a marginal improvement. In the two head-to-head tests available, the RTX 8000 secures both wins, delivering a 71% higher score in Geekbench OpenCL and a 74.3% higher score in Geekbench Vulkan. The Tesla M60, an older Maxwell-based part, trails by such a wide margin that its only remaining advantages are in its lower power draw and simpler power connector requirements.

Head-to-Head Benchmarks

The benchmark results paint a clear picture of dominance for the Quadro RTX 8000. In Geekbench OpenCL, the RTX 8000 scores 101,883 against the Tesla M60’s 29,506, a delta of -71% from the M60’s perspective. This means the RTX 8000 is roughly 3.45 times faster in raw compute workloads that leverage OpenCL. The gap is even more pronounced in Geekbench Vulkan, where the RTX 8000 achieves 122,637 versus the M60’s 31,473, representing a -74.3% delta. This indicates that the RTX 8000’s advantage extends to modern graphics APIs, where its Turing architecture’s dedicated hardware for ray tracing and tensor operations can be fully utilized.

The RTX 8000 also has more benchmark data available, including Passmark tests where it scores 19,799 in G3D, 9,992 in GPU compute, and 866 in G2D. These additional metrics, while not directly comparable to the M60 due to missing data, further solidify its position as a high-performance workstation card. The average benchmark score for the RTX 8000 is 28,421, which is slightly lower than the M60’s 30,490 average, but this discrepancy is due to the inclusion of legacy DirectX 10, 11, and 9 tests in the RTX 8000’s suite, which score low (137, 188, and 211 respectively) and drag down its average. In modern workloads, the RTX 8000’s superiority is unambiguous.

Architecture Differences

The architectural gap between these two NVIDIA professional cards is vast, stemming from a three-year release difference. The Tesla M60 is built on the Maxwell 2.0 architecture using the GM204 chip, fabricated on a 28 nm process at TSMC. It contains 5,200 million transistors on a 398 mm² die, yielding a transistor density of 13.1M per mm². In contrast, the Quadro RTX 8000 uses the Turing architecture with the TU102 chip, built on a 12 nm process, also at TSMC. This newer process allows for 18,600 million transistors on a 754 mm² die, resulting in a density of 24.7M per mm² — nearly double the density of the M60.

The core configurations differ dramatically. The M60 has 2,048 shading units, 128 texture mapping units, and 64 raster output pipelines. The RTX 8000 nearly doubles the shading units to 4,608, more than doubles the TMUs to 288, and increases ROPs to 96. Critically, the RTX 8000 introduces 72 RT cores and 576 tensor cores, which are entirely absent from the Maxwell-based M60. These dedicated cores enable hardware-accelerated ray tracing and AI-based tensor operations, features that the M60 cannot support at all.

Memory architecture is another major divider. The M60 comes with 8 GB of GDDR5 memory on a 256-bit bus, providing 160.4 GB/s of bandwidth. The RTX 8000 offers 48 GB of GDDR6 memory on a 384-bit bus, delivering 672.0 GB/s — over four times the bandwidth. This makes the RTX 8000 vastly more capable for large datasets, high-resolution textures, and complex scientific visualizations. The M60’s memory clock is 1253 MHz (5 Gbps effective), while the RTX 8000’s is 1750 MHz (14 Gbps effective). The RTX 8000 also supports DirectX 12 Ultimate (12_2), whereas the M60 only reaches DirectX 12 (12_1), and the RTX 8000 has display outputs (4x DisplayPort 1.4a and 1x USB Type-C) while the M60 has none, reflecting its server-oriented design.

FAQ

Q: Which card has a higher average benchmark score?

A: The Tesla M60 has a higher average benchmark score of 30,490 compared to the RTX 8000’s 28,421. However, this is misleading because the RTX 8000’s average includes legacy DirectX 9, 10, and 11 tests, which score very low, while the M60 only has Geekbench results.

Q: What is the performance difference in Vulkan workloads?

A: The RTX 8000 leads by 74.3% in Geekbench Vulkan, scoring 122,637 versus the M60’s 31,473. This indicates the RTX 8000 is roughly 3.9 times faster in Vulkan-based applications.

Q: Does the Tesla M60 support hardware ray tracing?

A: No. The M60 is based on Maxwell 2.0 architecture and has no RT cores. The RTX 8000, based on Turing, includes 72 RT cores specifically for ray tracing workloads.

Q: How much memory does each card have?

A: The Tesla M60 has 8 GB of GDDR5 memory, while the Quadro RTX 8000 has 48 GB of GDDR6 memory. The RTX 8000 also has a wider 384-bit bus and over four times the memory bandwidth.

Q: What are the power requirements for each card?

A: The Tesla M60 has a TDP of 300 W and requires a 700 W power supply with a single 8-pin connector. The RTX 8000 has a lower TDP of 260 W, requires a 600 W power supply, and uses one 6-pin plus one 8-pin connector.

Q: Which card is newer and what process node is it built on?

A: The RTX 8000 was released in 2018 and is built on a 12 nm process. The Tesla M60 was released in 2015 and uses a 28 nm process. The RTX 8000’s smaller node contributes to its higher transistor count and density.

The Verdict

The data is unequivocal: the Quadro RTX 8000 is the superior performer in every measurable benchmark. Its Geekbench OpenCL and Vulkan scores are more than triple those of the Tesla M60, and its architectural features — including RT cores, tensor cores, and 48 GB of memory — make it a far more capable workstation card. The RTX 8000 also requires less power (260 W vs 300 W) and a smaller power supply (600 W vs 700 W), making it more efficient despite its massive performance lead.

However, the Tesla M60 is not without merit. Its average benchmark score is higher due to the lack of legacy test results, and it holds a 0% delta against its nearest rival, the NVIDIA CMP 70HX, showing it remains competitive within its niche. For users with workloads that rely purely on older compute APIs or who need a server card with no display outputs, the M60 could still serve a purpose. But for anyone needing modern features, high memory capacity, or high-end performance, the RTX 8000 is the clear choice, despite its launch MSRP of 9,999 USD.

Specification Differences

| Specification | NVIDIA Tesla M60 | NVIDIA Quadro RTX 8000 |

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

| Architecture | Maxwell 2.0 | Turing |

| Process Node | 28 nm | 12 nm |

| Transistors | 5,200 million | 18,600 million |

| Die Size | 398 mm² | 754 mm² |

| Transistor Density | 13.1M / mm² | 24.7M / mm² |

| Base Clock | 557 MHz | 1395 MHz |

| Boost Clock | 1178 MHz | 1770 MHz |

| Memory Size | 8 GB | 48 GB |

| Memory Type | GDDR5 | GDDR6 |

| Memory Bus | 256 bit | 384 bit |

| Memory Bandwidth | 160.4 GB/s | 672.0 GB/s |

| Shading Units | 2048 | 4608 |

| TMUs | 128 | 288 |

| ROPs | 64 | 96 |

| RT Cores | None | 72 |

| Tensor Cores | None | 576 |

| Pixel Rate | 75.39 GPixel/s | 169.9 GPixel/s |

| Texture Rate | 150.8 GTexel/s | 509.8 GTexel/s |

| FP32 Performance | 4.825 TFLOPS | 16.31 TFLOPS |

| FP16 Performance | N/A | 32.62 TFLOPS (2:1) |

| TDP | 300 W | 260 W |

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

| Suggested PSU | 700 W | 600 W |

| Display Outputs | No outputs | 4x DisplayPort 1.4a, 1x USB Type-C |

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

| Release Date | 2015-08-29 | 2018-08-12 |

Where Each One Wins

The Quadro RTX 8000 wins in virtually all modern compute and graphics scenarios. Its 16.31 TFLOPS of FP32 performance versus the M60’s 4.825 TFLOPS makes it ideal for scientific simulations, AI training, and complex 3D rendering. The 48 GB memory capacity is essential for handling massive datasets or rendering scenes that exceed 8 GB, which would be impossible on the M60. The RT and tensor cores provide hardware acceleration for ray-traced workflows and deep learning inference, features that the M60 simply cannot offer. The RTX 8000 also wins in power efficiency, delivering over three times the performance while consuming 40 W less power.

The Tesla M60’s wins are limited to niche scenarios. Its higher average benchmark score (30,490 vs 28,421) is an artifact of the RTX 8000’s legacy test results, but it does indicate that the M60 is still competitive in older DirectX 9, 10, and 11 workloads when compared to the RTX 8000’s low scores in those specific tests. The M60 also has a simpler power connector requirement (single 8-pin) and no display outputs, which could be advantageous in server environments where a card is used purely for compute and display output is handled separately. For users with legacy applications that do not benefit from Turing’s advanced features and who require a dual-slot card with no video outputs, the M60 remains a functional, if dated, option.