AMD Radeon PRO V620 vs NVIDIA Tesla V100S PCIe 32 GB Comparison
AMD Radeon PRO V620
Tesla V100S PCIe 32 GB
PERFORMANCE BENCHMARKS
Analysis: AMD Radeon PRO V620 vs NVIDIA Tesla V100S PCIe 32 GB
# Head-to-Head Benchmarks
The data shows a single decisive benchmark matchup between these two cards, and it is not close. In Geekbench OpenCL, the NVIDIA Tesla V100S PCIe 32 GB scores 194415, while the AMD Radeon PRO V620 scores 128580. That is a 51.2% delta in favor of the NVIDIA card—a massive gap that dwarfs any other comparison in this dataset. The V100S also sits in the 98th percentile of all GPUs, while the V620 lands in the 96th percentile, so both are high-end parts, but the NVIDIA card is operating in a different tier entirely.
Looking at the nearest rivals provides additional context. The V100S is only 0.8% behind the NVIDIA RTX 6000D (which scores 195964), a margin that is effectively negligible. It beats the NVIDIA A100 SXM4 40 GB (187147) by 3.9%, the RTX 5000 Ada Generation (184664) by 5.3%, and the A100 SXM4 80 GB (183725) by 5.8%. For the AMD side, the V620 is essentially tied with its closest competitors: it is 0.5% ahead of the Radeon Pro W6800X Duo (135774), 0.8% ahead of the PRO W6800 (135396), and 0.9% ahead of both the NVIDIA A10M (135230) and RTX 4000 Ada Generation (135218). The V620's average benchmark score is 136472, which places it in a cluster of cards scoring between roughly 135,000 and 136,000.
The head-to-head delta of 51.2% is not just a win; it is a rout. The V100S delivers more than 1.5 times the OpenCL performance of the V620. This is the only benchmark available in the data, so the verdict on raw compute performance is unambiguous. The NVIDIA card wins the only head-to-head test, and it wins by a margin that is larger than the difference between the V620 and any of its nearest rivals. The AMD card's Vulkan score of 144364 is not part of the head-to-head comparison because the V100S has no corresponding Vulkan result in the dataset, but it is importantly this Vulkan number is still below the V100S's OpenCL score.
# FAQ
Q: Which card has the higher raw OpenCL benchmark score?
A: The NVIDIA Tesla V100S PCIe 32 GB scores 194415 in Geekbench OpenCL, which is 51.2% higher than the AMD Radeon PRO V620's 128580.
Q: How does the AMD Radeon PRO V620 compare to its nearest competitors?
A: The V620 is statistically tied with its nearest rivals: it is 0.5% ahead of the Radeon Pro W6800X Duo, 0.8% ahead of the PRO W6800, and 0.9% ahead of both the NVIDIA A10M and RTX 4000 Ada Generation.
Q: What is the percentile ranking of each card?
A: The NVIDIA Tesla V100S PCIe 32 GB is in the 98th percentile of all GPUs, while the AMD Radeon PRO V620 is in the 96th percentile.
Q: Does the AMD card have any benchmark advantage over the NVIDIA card?
A: No. The only head-to-head benchmark available is Geekbench OpenCL, and the NVIDIA card wins that test. The AMD card does have a Geekbench Vulkan score of 144364, but there is no corresponding Vulkan score for the NVIDIA card in the data.
Q: Are these cards still in production?
A: No. Both are listed as end-of-life products. The NVIDIA card was released on 2019-11-25, and the AMD card was released on 2021-11-03.
Q: Which card has a higher average benchmark score?
A: The NVIDIA Tesla V100S PCIe 32 GB has an average benchmark score of 194415, which is significantly higher than the AMD Radeon PRO V620's average of 136472.
# Architecture Differences
The architectural divide here is fundamental. The NVIDIA Tesla V100S PCIe 32 GB is built on the Volta architecture with the GV100 chip, fabricated on a 12 nm process at TSMC. The AMD Radeon PRO V620 uses the RDNA 2.0 architecture with the Navi 21 chip, fabricated on a more advanced 7 nm process, also at TSMC. The process node difference is significant: 12 nm versus 7 nm, which explains why the AMD chip packs 26,800 million transistors into a 520 mm² die, while the NVIDIA chip fits 21,100 million transistors into a much larger 815 mm² die. The transistor density tells the story—25.9M transistors per square millimeter for NVIDIA versus 51.5M for AMD.
The NVIDIA card features 640 tensor cores, a key differentiator for AI and deep learning workloads. The AMD card has no tensor cores but instead includes 72 ray tracing cores, which the NVIDIA card lacks. Shading unit counts are close: 5120 for NVIDIA versus 4608 for AMD. The NVIDIA card has 320 texture mapping units and 128 ROPs, while the AMD card has 288 TMUs and 128 ROPs. Clock speeds favor AMD significantly: the V620 boosts to 2200 MHz from a base of 1825 MHz, while the V100S boosts to 1597 MHz from a base of 1245 MHz.
Memory architecture is also completely different. The NVIDIA card uses 32 GB of HBM2 on a 4096-bit bus, delivering 1.13 TB/s of bandwidth. The AMD card uses 32 GB of GDDR6 on a 256-bit bus, delivering 512.0 GB/s—less than half the bandwidth. The memory clocks reflect this: NVIDIA's memory runs at 1107 MHz (2.2 Gbps effective), while AMD's runs at 2000 MHz (16 Gbps effective). The bus width difference is the dominant factor. API support favors AMD for modern graphics: the V620 supports DirectX 12 Ultimate (12_2), while the V100S only reaches DirectX 12 (12_1). Both support OpenGL 4.6 and Vulkan 1.4. The AMD card also uses PCIe 4.0 x16, while the NVIDIA card is limited to PCIe 3.0 x16.
# The Verdict
The data is unambiguous: the NVIDIA Tesla V100S PCIe 32 GB is the superior card for raw compute performance. It wins the only head-to-head benchmark by 51.2%, which is an enormous margin within GPU comparisons. Its 98th percentile ranking places it among the top 2% of all GPUs, and its nearest rivals are the RTX 6000D and A100 SXM4 variants—all of which are within 5.8% of the V100S. The AMD card, by contrast, sits at the 96th percentile and is clustered with mid-tier workstation cards like the PRO W6800 and RTX 4000 Ada Generation.
For compute-heavy workloads such as OpenCL-based processing, the choice is clear: the NVIDIA card delivers more than 1.5 times the performance. The V100S also offers substantially higher memory bandwidth (1.13 TB/s versus 512.0 GB/s), which matters for memory-bound applications. The AMD card does have advantages in architecture modernity—7 nm process, PCIe 4.0, DirectX 12 Ultimate, and ray tracing cores—but these features do not translate into a benchmark win in the available data.
The AMD card is not a poor performer; its average score of 136472 is respectable and it is statistically tied with several capable rivals. But against the V100S, it is outclassed. The verdict is straightforward: for anyone prioritizing raw compute performance as measured by OpenCL, the NVIDIA Tesla V100S PCIe 32 GB is the definitive choice. The AMD card's advantages are in features and efficiency of design, not in delivered performance.
# Specification Differences
| Specification | NVIDIA Tesla V100S PCIe 32 GB | AMD Radeon PRO V620 |
|---|---|---|
| Architecture | Volta | RDNA 2.0 |
| Chip | GV100 | Navi 21 |
| Process Node | 12 nm | 7 nm |
| Transistors | 21,100 million | 26,800 million |
| Die Size | 815 mm² | 520 mm² |
| Transistor Density | 25.9M / mm² | 51.5M / mm² |
| Base Clock | 1245 MHz | 1825 MHz |
| Boost Clock | 1597 MHz | 2200 MHz |
| Memory Type | HBM2 | GDDR6 |
| Memory Bus Width | 4096 bit | 256 bit |
| Memory Bandwidth | 1.13 TB/s | 512.0 GB/s |
| Memory Clock | 1107 MHz (2.2 Gbps effective) | 2000 MHz (16 Gbps effective) |
| Shading Units | 5120 | 4608 |
| TMUs | 320 | 288 |
| ROPs | 128 | 128 |
| Tensor Cores | 640 | None |
| Ray Tracing Cores | None | 72 |
| FP32 Performance | 16.35 TFLOPS | 20.28 TFLOPS |
| FP16 Performance | 32.71 TFLOPS (2:1) | 40.55 TFLOPS (2:1) |
| Pixel Rate | 204.4 GPixel/s | 281.6 GPixel/s |
| Texture Rate | 511.0 GTexel/s | 633.6 GTexel/s |
| TDP | 250 W | 300 W |
| Suggested PSU | 600 W | 700 W |
| Bus Interface | PCIe 3.0 x16 | PCIe 4.0 x16 |
| DirectX Support | 12 (12_1) | 12 Ultimate (12_2) |
| Release Date | 2019-11-25 | 2021-11-03 |
# Where Each One Wins
The NVIDIA Tesla V100S PCIe 32 GB wins in raw compute performance. Its Geekbench OpenCL score of 194415 is 51.2% higher than the AMD card's 128580, making it the clear choice for any workload that relies on OpenCL compute. It also wins decisively on memory bandwidth: 1.13 TB/s versus 512.0 GB/s. For applications that are memory-bandwidth-bound—such as large matrix operations, scientific computing, or data processing—this is a critical advantage. The tensor cores (640 of them) provide dedicated hardware for AI and deep learning workloads that the AMD card simply does not have. The V100S also has a higher percentile ranking (98th versus 96th), indicating stronger overall positioning among all GPUs.
The AMD Radeon PRO V620 wins on architectural modernity and features that do not show up in the OpenCL benchmark. It is built on a 7 nm process versus 12 nm, which explains its higher transistor density (51.5M versus 25.9M per mm²). It has a significantly higher boost clock (2200 MHz versus 1597 MHz) and delivers higher FP32 performance (20.28 TFLOPS versus 16.35 TFLOPS) and FP16 performance (40.55 TFLOPS versus 32.71 TFLOPS). The AMD card also has 72 ray tracing cores, which the NVIDIA card lacks entirely, making it the better choice for ray-traced rendering workloads. It supports DirectX 12 Ultimate (12_2) versus the NVIDIA card's DirectX 12 (12_1), and it uses the faster PCIe 4.0 interface. The AMD card's pixel rate (281.6 GPixel/s) and texture rate (633.6 GTexel/s) are also higher than the NVIDIA card's (204.4 GPixel/s and 511.0 GTexel/s).
In terms of use cases, the NVIDIA card is the winner for compute-centric tasks: OpenCL workloads, AI inference, and memory-bandwidth-heavy applications. The AMD card is the winner for graphics-centric tasks that leverage its ray tracing cores, higher clocks, and modern API support. The AMD card is also smaller and denser, with a die size of 520 mm² versus 815 mm², which may matter for physical integration. However, in the only direct benchmark comparison available, the NVIDIA card wins by a wide margin. The AMD card's advantages are real but situational, and they do not translate into a benchmark victory in this dataset.