AMD Radeon Instinct MI60 vs NVIDIA Tesla P100 PCIe 12 GB Comparison

AMD
RADEON

AMD Radeon Instinct MI60

CORE STATE Vega 20
VRAM 32 GB
CLOCK SPEED 1800 MHz
TDP 300 W
BUS WIDTH 4096 bit
ARCHITECTURE GCN 5.1
nm
PROCESS 7 nm
LAUNCH DATE 2018
VS
NVIDIA
GEFORCE

Tesla P100 PCIe 12 GB

CORE STATE GP100
VRAM 12 GB
CLOCK SPEED 1329 MHz
TDP 250 W
BUS WIDTH 3072 bit
ARCHITECTURE Pascal
nm
PROCESS 16 nm
LAUNCH DATE 2016

PERFORMANCE BENCHMARKS

geekbench_opencl
92,488
79,396
geekbench_vulkan
92,444
N/A

Analysis: AMD Radeon Instinct MI60 vs NVIDIA Tesla P100 PCIe 12 GB

The AMD Radeon Instinct MI60 is the clear winner in this comparison, outperforming the NVIDIA Tesla P100 PCIe 12 GB by a decisive 16.5% in the available compute benchmark. The MI60’s average benchmark score of 92,466 places it in the 93rd percentile of all GPUs, while the P100’s 79,396 score sits just one percentile lower at 92. This margin is substantial, reflecting fundamental architectural and generational advantages, though the P100 still holds its own in specific rasterization-oriented metrics.

Where Each One Wins

The AMD Radeon Instinct MI60 wins the only direct head-to-head benchmark available, the Geekbench OpenCL test, with a score of 92,488 against the P100’s 79,396. That 16.5% lead is the single biggest differentiator in this matchup and indicates a strong advantage in general-purpose compute workloads, which is the primary use case for both cards. The MI60 also edges out the NVIDIA RTX A4500 by 0.9% and the RTX A4500 Mobile by 1.5% in average benchmark score, showing it remains competitive against newer workstation parts despite being end-of-life.

The NVIDIA Tesla P100 PCIe 12 GB, however, wins in the pixel throughput department. Its pixel rate of 127.6 GPixel/s exceeds the MI60’s 115.2 GPixel/s, a roughly 10.7% advantage. This suggests the P100 is more efficient at fill-rate-bound tasks, such as certain types of rasterization or image processing, even though it loses the overall compute battle. The P100 also has a higher ROP count (96 vs. 64), which directly supports its pixel rate advantage.

Where the MI60 dominates is in raw compute throughput and memory subsystem capabilities. Its FP32 performance of 14.75 TFLOPS is 54.8% higher than the P100’s 9.526 TFLOPS, and its FP16 performance of 29.49 TFLOPS (2:1) is 54.8% higher than the P100’s 19.05 TFLOPS (2:1). The MI60 also delivers nearly double the memory bandwidth (1.02 TB/s vs. 549.1 GB/s), which is critical for large data sets in scientific computing and AI inference. In short, the MI60 wins on compute density and memory throughput, while the P100 wins on pixel fill rate.

Architecture Differences

The two cards represent different generations and design philosophies from their respective manufacturers. The AMD Radeon Instinct MI60 is built on the GCN 5.1 architecture using the Vega 20 chip, manufactured on a 7 nm process at TSMC. This advanced node allows AMD to pack 13,230 million transistors into a die size of just 331 mm², yielding a transistor density of 40.0M per mm². The NVIDIA Tesla P100 PCIe 12 GB, in contrast, uses the older Pascal architecture with the GP100 chip, fabricated on a 16 nm process also at TSMC. The P100 has more transistors overall at 15,300 million, but its die is much larger at 610 mm², resulting in a significantly lower transistor density of 25.1M per mm².

The MI60’s compute configuration is larger in most respects. It features 4,096 shading units, 256 TMUs, and 64 ROPs, compared to the P100’s 3,584 shading units, 224 TMUs, and 96 ROPs. The MI60’s clock speeds are also higher, with a boost clock of 1800 MHz versus the P100’s 1329 MHz, although the base clocks are nearly identical (1200 MHz vs. 1190 MHz). Memory configurations differ sharply: the MI60 has 32 GB of HBM2 on a 4096-bit bus, while the P100 has 12 GB of HBM2 on a 3072-bit bus. This explains the MI60’s 1.02 TB/s bandwidth versus the P100’s 549.1 GB/s.

Interface and power delivery also differ. The MI60 uses PCIe 4.0 x16, while the P100 uses PCIe 3.0 x16. The MI60 requires a 700 W suggested PSU and uses both a 6-pin and 8-pin power connector, whereas the P100 has a 600 W suggested PSU and uses a single 8-pin connector. The P100 has no display outputs, while the MI60 includes a single mini-DisplayPort 1.4a output. Both cards are dual-slot and share the same 267 mm length, but the MI60 has a specified height of 111 mm while the P100’s height is unspecified.

Head-to-Head Benchmarks

The only direct benchmark comparison available is the Geekbench OpenCL test, where the AMD Radeon Instinct MI60 scores 92,488 against the NVIDIA Tesla P100 PCIe 12 GB’s 79,396. This yields a 16.5% delta in favor of AMD. To put this in context, the MI60’s average score of 92,466 is 0.9% higher than the NVIDIA RTX A4500 (91,671) and 1.5% higher than the RTX A4500 Mobile (91,134), but it trails the AMD Radeon Pro VII (97,131) by 4.8% and the AMD Radeon RX 7900M (97,487) by 5.2%. The P100’s average score of 79,396 is essentially tied with the NVIDIA Tesla P100 PCIe 16 GB (79,605, a -0.3% delta) and the NVIDIA GeForce RTX 5090 (79,842, a -0.6% delta), while it leads the AMD Radeon RX 6850M XT (78,940) by 0.6% and trails the AMD Radeon Pro Vega 64X (80,959) by 1.9%.

The 16.5% gap between the MI60 and P100 is larger than any of the deltas seen in their respective nearest rival groups, indicating a clear generational leap. The MI60’s advantage is most pronounced in FP32 and FP16 throughput, where it leads by roughly 55% in both cases. Its memory bandwidth advantage is even more stark: 86% higher than the P100. However, the P100’s pixel rate is 10.8% higher, and its ROP count is 50% higher, suggesting that in specific texture-heavy or fill-rate-limited scenarios, the older NVIDIA card could close the gap.

The Verdict

From the data, the AMD Radeon Instinct MI60 is the superior compute accelerator for general-purpose workloads. Its 16.5% lead in Geekbench OpenCL, combined with 54.8% higher FP32 and FP16 throughput, makes it the obvious choice for tasks like deep learning training, scientific simulations, or any workload that relies on raw floating-point performance. The MI60’s 32 GB memory capacity and 1.02 TB/s bandwidth also provide a substantial advantage for large models or datasets that would exceed the P100’s 12 GB frame buffer and 549.1 GB/s bandwidth.

The NVIDIA Tesla P100 PCIe 12 GB should only be considered if the workload is heavily dependent on pixel fill rate, where its 127.6 GPixel/s output beats the MI60’s 115.2 GPixel/s. The P100 also consumes less power (250 W vs. 300 W) and requires a smaller suggested PSU (600 W vs. 700 W), which may be relevant in power-constrained environments. Its launch MSRP was 4,599 USD, though the MI60 has no listed launch MSRP.

For anyone choosing between these two end-of-life cards, the MI60 is the better investment for compute, while the P100 retains a niche for fill-rate-bound tasks. The MI60’s PCIe 4.0 interface also future-proofs it better than the P100’s PCIe 3.0, though both are legacy products. The data strongly favors AMD in the compute domain, and the 16.5% benchmark delta is the single most important number in this comparison.

FAQ

Q: Which card has a higher average benchmark score?

A: The AMD Radeon Instinct MI60 has an average benchmark score of 92,466, which is 16.5% higher than the NVIDIA Tesla P100 PCIe 12 GB’s average of 79,396.

Q: How does the MI60’s memory bandwidth compare to the P100’s?

A: The MI60 has a memory bandwidth of 1.02 TB/s, which is roughly 86% higher than the P100’s 549.1 GB/s, due to its larger 4096-bit bus and 32 GB of HBM2 versus the P100’s 3072-bit bus and 12 GB of HBM2.

Q: What is the pixel rate difference between the two cards?

A: The NVIDIA Tesla P100 PCIe 12 GB has a higher pixel rate of 127.6 GPixel/s, while the AMD Radeon Instinct MI60 has a pixel rate of 115.2 GPixel/s, giving the P100 a 10.8% advantage in this metric.

Q: Which card has a higher boost clock?

A: The AMD Radeon Instinct MI60 has a boost clock of 1800 MHz, compared to the NVIDIA Tesla P100 PCIe 12 GB’s boost clock of 1329 MHz, a difference of 471 MHz.

Q: Are there any display outputs on either card?

A: The AMD Radeon Instinct MI60 has one mini-DisplayPort 1.4a output, while the NVIDIA Tesla P100 PCIe 12 GB has no display outputs at all.

Q: What is the transistor density of each chip?

A: The MI60’s Vega 20 chip has a transistor density of 40.0M per mm², while the P100’s GP100 chip has a density of 25.1M per mm², reflecting the MI60’s smaller 7 nm process versus the P100’s 16 nm process.

Specification Differences

| Specification | AMD Radeon Instinct MI60 | NVIDIA Tesla P100 PCIe 12 GB |

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

| Chip | Vega 20 | GP100 |

| Architecture | GCN 5.1 | Pascal |

| Generation | Radeon Instinct (MIx) | Tesla Pascal (Pxx) |

| Process Node | 7 nm | 16 nm |

| Foundry | TSMC | TSMC |

| 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 | 12 GB |

| Memory Type | HBM2 | HBM2 |

| Memory Bus Width | 4096 bit | 3072 bit |

| Memory Bandwidth | 1.02 TB/s | 549.1 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 Performance | 14.75 TFLOPS | 9.526 TFLOPS |

| FP16 Performance | 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 |

| Dimensions (Length) | 267 mm (10.5 inches) | 267 mm (10.5 inches) |

| Dimensions (Height) | 111 mm (4.4 inches) | Not specified |

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

| Predecessor | FirePro Data Center | Tesla Maxwell |

| Successor | Not specified | Tesla Volta |

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

| Launch MSRP | Not specified | 4,599 USD |

| Geekbench OpenCL Score | 92,488 | 79,396 |

| Avg Benchmark Score | 92,466 | 79,396 |

| Percentile vs All GPUs | 93 | 92 |

DETAILED SPECIFICATIONS

SPECIFICATION
Instinct MI60
Tesla P100 PCIe 12 GB
Core Specs
Shading Units
4,096
3,584 -12.5%
Shaders
4,096
3,584 -12.5%
TMUs
256
224 -12.5%
ROPs
64
96 +50.0%
Compute Units
64
—
SM Count
—
56
Clocks
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
Memory Size
32 GB
12 GB
VRAM (MB)
32,768
12,288 -62.5%
Memory Type
HBM2
HBM2
Memory Bus
4096 bit
3072 bit
Bandwidth
1.02 TB/s
549.1 GB/s
Cache
L1 Cache
16 KB (per CU)
24 KB (per SM)
L2 Cache
4 MB
3 MB
Performance
Pixel Rate
115.2 GPixel/s
127.6 GPixel/s
Texture Rate
460.8 GTexel/s
297.7 GTexel/s
FP32 (TFLOPS)
14.75 TFLOPS
9.526 TFLOPS
FP64 (TFLOPS)
7.373 TFLOPS (1:2)
4.763 TFLOPS (1:2)
FP16 (TFLOPS)
29.49 TFLOPS (2:1)
19.05 TFLOPS (2:1)
Power
TDP
300 W
250 W
TDP (W)
300
250 -16.7%
Suggested PSU
700 W
600 W
Power Connectors
1x 6-pin + 1x 8-pin
1x 8-pin
Architecture
Architecture
GCN 5.1
Pascal
GPU Name
Vega 20
GP100
Generation
Radeon Instinct (MIx)
Tesla Pascal (Pxx)
Process Size
7 nm
16 nm
Transistors
13,230 million
15,300 million
Die Size
331 mm²
610 mm²
Foundry
TSMC
TSMC
Density
40.0M / mm²
25.1M / mm²
API Support
DirectX
12 (12_1)
12 (12_1)
OpenGL
4.6
4.6
Vulkan
1.3
1.3
OpenCL
2.1
3.0
CUDA
—
6.0
Shader Model
6.7
6.0
Physical
Slot Width
Dual-slot
Dual-slot
Length
267 mm 10.5 inches
267 mm 10.5 inches
Height
111 mm 4.4 inches
—
Outputs
1x mini-DisplayPort 1.4a
No outputs
Bus Interface
PCIe 4.0 x16
PCIe 3.0 x16
Other
Launch Price
—
4,599 USD
Production
End-of-life
End-of-life
Predecessor
FirePro Data Center
Tesla Maxwell
Successor
—
Tesla Volta
View Radeon Instinct MI60 Details View Tesla P100 PCIe 12 GB Details