AMD Radeon Instinct MI60 vs NVIDIA Tesla P100 PCIe 16 GB Comparison

AMD Radeon Instinct MI60 vs NVIDIA Tesla P100 PCIe 16 GB

The AMD Radeon Instinct MI60 decisively outperforms the NVIDIA Tesla P100 PCIe 16 GB in the available compute benchmark, posting a 16.2% higher Geekbench OpenCL score. With an average benchmark score of 92,466 against 79,605, the MI60 sits in the 93rd percentile of all GPUs, while the P100 sits just behind at the 92nd percentile. The data indicates a clear generational and architectural advantage for the AMD part, driven by a newer process node, higher clock speeds, and double the memory capacity.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon Instinct MI60 has a significantly higher average benchmark score of 92,466, compared to the NVIDIA Tesla P100 PCIe 16 GB's 79,605. This represents a 16.2% advantage for the AMD card in the Geekbench OpenCL test.

Q: How does the MI60 compare to its nearest rivals?

A: The MI60 is 0.9% ahead of the NVIDIA RTX A4500 (91,671) and 1.5% ahead of the RTX A4500 Mobile (91,134). It trails the AMD Radeon Pro VII (97,131) by 4.8% and the AMD Radeon RX 7900M (97,487) by 5.2%.

Q: How does the Tesla P100 compare to its nearest rivals?

A: The P100 is 0.3% ahead of the NVIDIA Tesla P100 PCIe 12 GB (79,396) and 0.8% ahead of the AMD Radeon RX 6850M XT (78,940). It is 0.3% behind the NVIDIA GeForce RTX 5090 (79,842) and 1.7% behind the AMD Radeon Pro Vega 64X (80,959).

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

A: The AMD Radeon Instinct MI60 features 32 GB of HBM2 memory, while the NVIDIA Tesla P100 PCIe 16 GB features 16 GB of HBM2 memory. The MI60 offers double the memory capacity.

Q: What are the release dates for these two accelerators?

A: The AMD Radeon Instinct MI60 was released on 2018-11-17, while the NVIDIA Tesla P100 PCIe 16 GB was released earlier on 2016-06-19. Both cards are now listed as end-of-life products.

Q: Does either card support modern graphics APIs?

A: Yes, both the AMD Radeon Instinct MI60 and the NVIDIA Tesla P100 PCIe 16 GB support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3. Neither card features dedicated ray tracing or tensor cores.

Architecture Differences

The AMD Radeon Instinct MI60 is built on the Vega 20 chip using the GCN 5.1 architecture, fabricated on a 7 nm process node at TSMC. In contrast, the NVIDIA Tesla P100 PCIe 16 GB uses the GP100 chip with the older Pascal architecture, manufactured on a 16 nm process node, also at TSMC. This process difference is substantial, as the MI60 packs 13,230 million transistors into a 331 mm² die, resulting in a transistor density of 40.0M per mm². The P100, despite having more transistors at 15,300 million, uses a much larger 610 mm² die, yielding a lower transistor density of 25.1M per mm².

The MI60 features 4,096 shading units, 256 texture mapping units, and 64 raster operations pipelines. The P100 has fewer of these: 3,584 shading units, 224 TMUs, but a higher count of 96 ROPs. Clock speeds also differ significantly. The MI60 has a base clock of 1200 MHz and a boost clock of 1800 MHz, while the P100 has a base clock of 1190 MHz and a much lower boost clock of 1329 MHz. Memory architecture shows both use a 4096-bit HBM2 bus, but the MI60 runs its memory at 1000 MHz (2 Gbps effective) delivering 1.02 TB/s of bandwidth, whereas the P100 runs at 715 MHz (1430 Mbps effective) for 732.2 GB/s.

Both cards are dual-slot designs and share the same 267 mm length (10.5 inches). The MI60 requires a 700 W suggested PSU and uses 1x 6-pin + 1x 8-pin power connectors, while the P100 requires a 600 W PSU with a single 8-pin connector. Power draw is rated at 300 W for the MI60 and 250 W for the P100. The MI60 uses PCIe 4.0 x16 and has one mini-DisplayPort 1.4a output, whereas the P100 uses the older PCIe 3.0 x16 interface and has no display outputs at all.

Head-to-Head Benchmarks

In the only head-to-head benchmark available, the Geekbench OpenCL test, the AMD Radeon Instinct MI60 wins decisively with a score of 92,488 against the NVIDIA Tesla P100 PCIe 16 GB's 79,605. This yields a delta of 16.2% in favor of the MI60. This is a substantial margin, indicating that the MI60 is not just marginally faster but delivers a significantly higher level of raw compute throughput in this workload.

The MI60's performance advantage is consistent with its architectural strengths. The higher boost clock of 1800 MHz versus 1329 MHz, combined with the newer 7 nm process, allows the AMD card to execute more operations per second. The MI60's FP32 throughput is 14.75 TFLOPS compared to the P100's 9.526 TFLOPS, which translates directly to the observed benchmark lead. Similarly, FP16 performance is 29.49 TFLOPS for the MI60 versus 19.05 TFLOPS for the P100, a 2:1 ratio on both cards.

The MI60 also holds advantages in texture rate, with 460.8 GTexel/s versus 297.7 GTexel/s for the P100. However, the P100 does win in pixel rate, achieving 127.6 GPixel/s compared to the MI60's 115.2 GPixel/s, thanks to its higher ROP count of 96 versus 64. This suggests that while the MI60 is generally faster at compute and texture-heavy tasks, the P100 might have an edge in certain fill-rate-bound scenarios.

Specification Differences

| Specification | AMD Radeon Instinct MI60 | NVIDIA Tesla P100 PCIe 16 GB |

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

| Chip | Vega 20 | GP100 |

| Architecture | GCN 5.1 | Pascal |

| Process Node | 7 nm | 16 nm |

| Transistors | 13,230 million | 15,300 million |

| Die Size | 331 mm² | 610 mm² |

| Transistor Density | 40.0M / mm² | 25.1M / mm² |

| Base Clock | 1200 MHz | 1190 MHz |

| Boost Clock | 1800 MHz | 1329 MHz |

| Memory Clock | 1000 MHz (2 Gbps effective) | 715 MHz (1430 Mbps effective) |

| Memory Size | 32 GB | 16 GB |

| Memory Bandwidth | 1.02 TB/s | 732.2 GB/s |

| Shading Units | 4096 | 3584 |

| TMUs | 256 | 224 |

| ROPs | 64 | 96 |

| Pixel Rate | 115.2 GPixel/s | 127.6 GPixel/s |

| Texture Rate | 460.8 GTexel/s | 297.7 GTexel/s |

| FP32 | 14.75 TFLOPS | 9.526 TFLOPS |

| FP16 | 29.49 TFLOPS (2:1) | 19.05 TFLOPS (2:1) |

| TDP | 300 W | 250 W |

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

| Suggested PSU | 700 W | 600 W |

| Bus Interface | PCIe 4.0 x16 | PCIe 3.0 x16 |

| Display Outputs | 1x mini-DisplayPort 1.4a | No outputs |

| Release Date | 2018-11-17 | 2016-06-19 |

Where Each One Wins

The AMD Radeon Instinct MI60 is the clear winner in raw compute performance. It wins the only head-to-head benchmark with a 16.2% margin in Geekbench OpenCL. Its strengths lie in its substantially higher FP32 (14.75 TFLOPS) and FP16 (29.49 TFLOPS) throughput, which are 54.9% and 54.8% higher than the P100 respectively. The MI60's double the memory capacity (32 GB vs 16 GB) and 39.4% higher memory bandwidth (1.02 TB/s vs 732.2 GB/s) also make it the superior choice for memory-intensive workloads like large model training or rendering high-resolution scenes. Its newer PCIe 4.0 interface doubles the available bus bandwidth for data transfer to and from the host system.

The NVIDIA Tesla P100 PCIe 16 GB, despite losing the overall performance race, does have specific advantages. Its pixel rate is 10.8% higher (127.6 GPixel/s vs 115.2 GPixel/s), driven by its larger ROP count. This could translate to better performance in certain rasterization or pixel-fill-bound tasks. It also consumes 50 W less power (250 W vs 300 W) and requires a lower 600 W suggested PSU versus 700 W, making it more power-efficient in absolute terms. The P100 also has a higher transistor count (15,300 million vs 13,230 million), though on an older, less dense process. The P100's launch MSRP is 5,699 USD, a figure not available for the MI60.

The Verdict

Based strictly on the benchmark data, the AMD Radeon Instinct MI60 is the superior compute accelerator. It delivers a 16.2% higher OpenCL score, has double the memory capacity, provides significantly more memory bandwidth, and offers roughly 55% more FP32 and FP16 throughput. For any compute-centric workload, from scientific simulation to machine learning inference, the MI60 is the stronger choice. Its newer 7 nm architecture and higher boost clock are the primary drivers of this performance gap.

The NVIDIA Tesla P100 PCIe 16 GB, while older and slower in compute, holds niche advantages. Its higher pixel rate and lower power draw make it a potential fit for scenarios where fill rate is the bottleneck or where power constraints are strict. Being an earlier release (2016 vs 2018), it may have a place in legacy systems or specific Pascal-optimized deployments. However, for general data-center compute tasks, the data firmly favors the AMD Radeon Instinct MI60, which sits in a higher performance percentile and outperforms the P100 across nearly every computational metric. The MI60's position among its nearest rivals also shows it is competitive with newer NVIDIA workstation parts, being only 0.9% behind the RTX A4500. The P100, in contrast, is clustered tightly with a mix of older and newer cards, showing its performance level has been surpassed by modern offerings.