AMD Radeon Pro Vega 20 vs NVIDIA Tesla M60 Comparison

Head-to-Head Benchmarks

The recorded data shows a clear overall advantage for the NVIDIA Tesla M60 in the shared benchmark suite. Across the two workloads where both cards have comparable scores, the Tesla M60 wins both, giving it a 2-0 record in the head-to-head comparison.

The largest gap appears in Geekbench Vulkan, a compute-oriented API test. The Tesla M60 records a score of 31,473, while the AMD Radeon Pro Vega 20 manages 26,410. That works out to a 19.2% advantage for the NVIDIA part, the most decisive margin in this comparison. By contrast, the OpenCL test is closer: the Tesla M60 scores 29,506 against 26,479 for the Radeon Pro Vega 20, a 10.6% lead.

It is worth remembering the AMD card also has a Metal benchmark score of 30,427, but there is no corresponding Metal result for the NVIDIA part in the database. That means the AMD card’s average benchmark score of 27,839 includes a third test, which narrows the aggregate gap. The Tesla M60 averages 30,490 across its two available scores. Even with that discrepancy, the delta between the two cards in the shared tests is consistent: the NVIDIA GPU is ahead in every directly comparable workload.

The percentile ranking reinforces the same picture. The Tesla M60 sits at the 75th percentile among all GPUs in the database, while the Radeon Pro Vega 20 sits at the 73rd. That two-point gap is modest, but the benchmark scores tell a slightly different story. The Tesla M60’s average score of 30,490 places it within 0% of the NVIDIA CMP 70HX (30,476) and 0.2% of the AMD Radeon RX 6700 (30,433). The Radeon Pro Vega 20, with its 27,839 average, sits 0.2% below the AMD Radeon RX 7800M (27,883) and 0.5% below the AMD Radeon Pro W5500X (27,973).

In short, the Tesla M60 is not merely ahead in the two common tests; it is ahead by a meaningful margin, and that advantage holds across both the OpenCL and Vulkan execution paths.

Where Each One Wins

The Tesla M60 wins in raw compute throughput, at least as far as cross-platform API benchmarks go. Its 19.2% advantage in Vulkan suggests that NVIDIA’s Maxwell architecture handles that workload with particular efficiency, and the 10.6% OpenCL lead points to a broadly faster chip in general compute tasks. The M60 also carries 2,048 shading units, 128 TMUs and 64 ROPs, alongside 8 GB of GDDR5 memory on a 256-bit bus, delivering 160.4 GB/s of memory bandwidth. The pixel rate is listed at 75.39 GPixel/s and texture rate at 150.8 GTexel/s, with FP32 throughput of 4.825 TFLOPS.

The Radeon Pro Vega 20, meanwhile, wins in contexts that the shared benchmarks do not capture. It has a Metal score of 30,427, and in the absence of a comparable Metal result for the Tesla M60, that remains an uncontested strength. The card is built for portable Mac systems, listed as an integrated graphics processor (IGP in slot width terms, with display outputs described as portable device dependent). Its power envelope is drastically lower at 100 W versus 300 W for the Tesla M60.

In practical terms, the AMD card also offers a memory bandwidth advantage per watt. The Vega 20 has 4 GB of HBM2 memory on a 1,024-bit bus, yielding 189.4 GB/s, which is actually higher than the Tesla M60’s 160.4 GB/s despite the smaller capacity. That makes the Vega 20 the card that wins in memory bandwidth density, though the M60 wins in raw memory capacity.

Architecture Differences

The two cards come from different architectural generations and different foundries. The NVIDIA Tesla M60 is built on the GM204 chip using Maxwell 2.0 architecture, produced on a 28 nm process at TSMC. It integrates 5,200 million transistors on a die size of 398 mm². That works out to a transistor density of 13.1 million per square millimeter. The card belongs to the Tesla Maxwell generation, with the product line succession going from Tesla Kepler to Tesla Pascal.

The AMD Radeon Pro Vega 20 uses the Vega 12 chip with GCN 5.0 architecture, fabricated on a 14 nm process at GlobalFoundries. The transistor count and die size are not recorded in the database, so no direct density calculation is possible.

Clock speeds differ notably. The Tesla M60 has a base clock of 557 MHz and a boost clock of 1,178 MHz, while the Radeon Pro Vega 20 runs a base of 815 MHz and a boost of 1,283 MHz. The AMD card has the higher clocks, but the NVIDIA card has more execution resources.

The memory subsystems diverge sharply. The M60 uses 8 GB of GDDR5 on a 256-bit bus, at a memory clock of 1,253 MHz (5 Gbps effective), producing 160.4 GB/s of bandwidth. The Vega 20 uses 4 GB of HBM2 on a 1,024-bit bus, at a memory clock of 740 MHz (1,480 Mbps effective), producing 189.4 GB/s. So the Vega 20 has the wider bus and the higher bandwidth, but only half the capacity.

Compute resources tell the opposite story. The Tesla M60 has 2,048 shading units, 128 TMUs and 64 ROPs. The Radeon Pro Vega 20 has 1,280 shading units, 80 TMUs and 32 ROPs. The M60’s pixel rate of 75.39 GPixel/s and texture rate of 150.8 GTexel/s are both higher than the AMD part’s 41.06 GPixel/s and 102.6 GTexel/s. The NVIDIA card’s FP32 output is 4.825 TFLOPS versus the AMD card’s 3.284 TFLOPS.

The Vega 20 does have a notable feature the M60 lacks: FP16 throughput. The Vega 20 lists 6.569 TFLOPS at a 2:1 ratio, while the Tesla M60 has no recorded FP16 figure at all. API support is otherwise similar, DirectX 12 (12_1) and OpenGL 4.6 on both cards, but Vulkan versions differ slightly: the M60 lists Vulkan 1.4 while the Vega 20 lists Vulkan 1.3.

Physical design and power also diverge. The M60 is a dual-slot card requiring a 700 W power supply unit rating, with a single 8-pin power connector, and it has no display outputs. The Vega 20 is an integrated GPU with no listed length and no power connectors, drawing 100 W TDP. The M60’s TDP is 300 W.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA Tesla M60, with an average score of 30,490, against 27,839 for the Radeon Pro Vega 20.

Q: Is the AMD Radeon Pro Vega 20 competitive in any benchmark?

A: Yes, the AMD card has a Geekbench Metal score of 30,427, but the NVIDIA card has no Metal result recorded, so this comparison is not directly available.

Q: Which card has more memory capacity?

A: The NVIDIA Tesla M60, with 8 GB of GDDR5, versus 4 GB of HBM2 on the AMD Radeon Pro Vega 20.

Q: Which card has higher memory bandwidth?

A: The AMD Radeon Pro Vega 20, with 189.4 GB/s, versus 160.4 GB/s for the Tesla M60.

Q: How do the two cards compare in the Vulkan benchmark?

A: The Tesla M60 scores 31,473 versus 26,410 for the Vega 20, a 19.2% advantage for NVIDIA.

Q: What is the power consumption difference?

A: The Tesla M60 is rated at 300 W TDP, while the Radeon Pro Vega 20 is rated at 100 W TDP.

The Verdict

The data points to the NVIDIA Tesla M60 as the faster card in the shared compute benchmarks. Its 10.6% lead in OpenCL and 19.2% lead in Vulkan are not marginal: they indicate a substantial performance advantage in cross-platform compute workloads. The M60 also has more shading units, TMUs, ROPs, and more than double the FP32 throughput. For anyone running general compute tasks or Vulkan-based applications, the Tesla M60 is the stronger choice.

The AMD Radeon Pro Vega 20 is the better option in specific scenarios. It consumes only 100 W TDP compared to 300 W, making it far more power-efficient. It has higher memory bandwidth at 189.4 GB/s despite a smaller 4 GB frame buffer. Its HBM2 memory and integrated form factor suit it for portable or space-constrained Mac systems, where its Metal performance of 30,427 remains a point in its favor. The Vega 20 also offers FP16 compute at 6.569 TFLOPS, which the Tesla M60 does not provide.

There is no scenario in the recorded data where the Vega 20 wins a directly comparable benchmark. But the two cards serve different roles. The Tesla M60 is a high-power, high-throughput accelerator for compute-heavy environments. The Radeon Pro Vega 20 is a low-power, integrated GPU for workstation laptops, with a memory subsystem that punches above its weight and a unique Metal result that has no NVIDIA counterpart in the database.

Specification Differences

| Specification | NVIDIA Tesla M60 | AMD Radeon Pro Vega 20 |

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

| Architecture | Maxwell 2.0 | GCN 5.0 |

| Process node | 28 nm | 14 nm |

| Foundry | TSMC | GlobalFoundries |

| Transistors | 5,200 million | Not recorded |

| Die size | 398 mm² | Not recorded |

| Transistor density | 13.1M / mm² | Not recorded |

| Base clock | 557 MHz | 815 MHz |

| Boost clock | 1,178 MHz | 1,283 MHz |

| Memory clock | 1,253 MHz (5 Gbps effective) | 740 MHz (1,480 Mbps effective) |

| Memory size | 8 GB | 4 GB |

| Memory type | GDDR5 | HBM2 |

| Memory bus | 256 bit | 1,024 bit |

| Memory bandwidth | 160.4 GB/s | 189.4 GB/s |

| Shading units | 2,048 | 1,280 |

| TMUs | 128 | 80 |

| ROPs | 64 | 32 |

| Pixel rate | 75.39 GPixel/s | 41.06 GPixel/s |

| Texture rate | 150.8 GTexel/s | 102.6 GTexel/s |

| FP32 | 4.825 TFLOPS | 3.284 TFLOPS |

| FP16 | Not recorded | 6.569 TFLOPS (2:1) |

| TDP | 300 W | 100 W |

| Slot width | Dual-slot | IGP |

| Power connectors | 1x 8-pin | Not recorded |

| Suggested PSU | 700 W | Not recorded |

| Display outputs | No outputs | Portable device dependent |

| Vulkan version | 1.4 | 1.3 |

| Release date | 2015-08-29 | 2018-11-13 |

| Geekbench OpenCL | 29,506 | 26,679 |

| Geekbench Vulkan | 31,473 | 26,410 |

| Geekbench Metal | Not recorded | 30,427 |

| Average benchmark score | 30,490 | 27,839 |

| Percentile vs all GPUs | 75th | 73rd |