AMD Radeon PRO V620 vs NVIDIA Quadro GP100 Comparison

The AMD Radeon PRO V620 and the NVIDIA Quadro GP100 are both end-of-life workstation accelerators, but they represent very different generations of GPU design. The recorded data shows a single head-to-head benchmark result, the Geekbench OpenCL test, where the AMD Radeon PRO V620 scores 128,580 points against 87,445 points for the NVIDIA Quadro GP100. That is a 47% advantage for the AMD part. The AMD card also holds a higher average benchmark score of 136,472 compared to 87,445 for the NVIDIA, placing it in the 96th percentile of all GPUs in the database, while the NVIDIA sits at the 93rd percentile. The following analysis breaks down these results, the architectural differences behind them, and the practical implications for workstation use.

Head-to-Head Benchmarks

The only direct comparison available in the database is the Geekbench OpenCL workload. This test measures raw compute throughput across a wide range of operations, including integer math, floating-point math, and memory bandwidth exercises. The AMD Radeon PRO V620 produced a score of 128,580, while the NVIDIA Quadro GP100 managed 87,445. The delta is 47% in favor of the AMD card. This is not a marginal victory; it is a substantial gap that indicates the V620 has a decisive compute advantage in this particular benchmark.

Looking at the broader context, the AMD Radeon PRO V620’s average benchmark score across all recorded tests is 136,472. That average is higher than its OpenCL score alone, suggesting that its performance in other workloads is even stronger. The closest rivals to the V620 in the database include the AMD Radeon Pro W6800X Duo with an average score of 135,774, which is just 0.5% behind, and the AMD Radeon PRO W6800 at 135,396, a 0.8% gap. The NVIDIA A10M and NVIDIA RTX 4000 Ada Generation also sit nearby, with scores of 135,230 and 135,218 respectively, each about 0.9% behind the V620. This cluster of scores indicates that the V620 is at the top of its immediate performance tier, but the margin over these rivals is thin, within a single percentage point.

The NVIDIA Quadro GP100, in contrast, has an average benchmark score of 87,445, identical to its OpenCL result because that is the only test recorded for it. Its nearest rivals are a different group. The AMD Radeon PRO W7600 scores 87,108, just 0.4% below the GP100, and the NVIDIA CMP 40HX scores 85,637, a 2.1% deficit. However, the GP100 falls behind the NVIDIA RTX A4500 Mobile and NVIDIA RTX A4500, which score 91,134 and 91,671 respectively, meaning the GP100 is 4% and 4.6% slower than those two parts. The data shows a clear performance hierarchy: the V620 belongs to a higher-performing class, while the GP100 is a mid-tier competitor among its peers.

The 47% delta in the OpenCL test is the headline number here. It is worth remembering the GP100 was released in 2016, while the V620 came later in 2021. The five-year gap in release dates explains a significant portion of the performance disparity, as the V620 benefits from newer architecture and manufacturing processes. But the benchmark result stands on its own: in a compute-heavy OpenCL workload, the AMD Radeon PRO V620 is nearly 1.5 times faster than the NVIDIA Quadro GP100.

Architecture Differences

The architectural gap between these two GPUs is substantial and directly explains the benchmark results. The AMD Radeon PRO V620 is built on the RDNA 2.0 architecture, using the Navi 21 chip, and manufactured on a 7 nm process at TSMC. The NVIDIA Quadro GP100 uses the older Pascal architecture, with the GP100 chip, built on a 16 nm process, also at TSMC. The shift from 16 nm to 7 nm is a major generational leap, allowing for significantly higher transistor density and power efficiency.

The transistor counts reflect this difference. The AMD chip contains 26,800 million transistors on a die size of 520 mm², resulting in a transistor density of 51.5 million per mm². The NVIDIA chip has 15,300 million transistors on a larger die of 610 mm², yielding a density of only 25.1 million per mm². The AMD part packs nearly twice as many transistors into a smaller physical area, which is a direct consequence of the newer process node. The higher density allows the V620 to implement far more compute units and specialized hardware.

Clock speeds also tell a story. The AMD Radeon PRO V620 has a base clock of 1825 MHz and a boost clock of 2200 MHz. The NVIDIA Quadro GP100 operates at a base of 1304 MHz and a boost of 1443 MHz. The AMD card runs at a significantly higher frequency, which contributes to its compute advantage. Combined with more shading units, the V620’s raw throughput is far higher.

The shading unit count is the largest single architectural difference. The AMD Radeon PRO V620 has 4,608 shading units, 288 texture mapping units, and 128 raster output pipelines. The NVIDIA Quadro GP100 has 3,584 shading units, 224 TMUs, and 96 ROPs. The AMD card has roughly 28% more shading units, 29% more TMUs, and 33% more ROPs. This translates directly to higher pixel and texture rates: the V620 achieves 281.6 GPixel/s and 633.6 GTexel/s, while the GP100 manages 138.5 GPixel/s and 323.2 GTexel/s. The AMD card is roughly double the GP100 in both fill-rate metrics.

Memory architecture is another critical divergence. The AMD Radeon PRO V620 uses 32 GB of GDDR6 memory on a 256-bit bus, providing a bandwidth of 512.0 GB/s. The NVIDIA Quadro GP100 uses 16 GB of HBM2 memory on a massive 4096-bit bus, delivering 732.2 GB/s of bandwidth. Despite having twice the memory capacity, the AMD card actually has lower memory bandwidth. The GP100’s HBM2 implementation offers 43% more bandwidth, which is a notable advantage for memory-bound workloads. However, the V620 compensates with its higher compute throughput and larger memory pool.

The AMD Radeon PRO V620 also includes 72 ray tracing cores, a feature entirely absent from the NVIDIA Quadro GP100, which has no ray tracing hardware. The V620 supports DirectX 12 Ultimate with feature level 12_2, while the GP100 only supports DirectX 12 with feature level 12_1. Both cards support OpenGL 4.6, but the V620 supports Vulkan 1.4 compared to Vulkan 1.3 on the GP100. The V620 also supports PCIe 4.0 x16, while the GP100 is limited to PCIe 3.0 x16, which affects data transfer speeds with the host system.

Power consumption differs as well. The AMD Radeon PRO V620 has a TDP of 300 W and requires two 8-pin power connectors, with a suggested 700 W power supply. The NVIDIA Quadro GP100 has a TDP of 235 W, needs only one 8-pin connector, and suggests a 550 W power supply. The AMD card draws more power, but it delivers substantially more compute performance per the benchmark data. The GP100 offers a single DVI and four DisplayPort 1.4a outputs, while the V620 has no display outputs at all, indicating its intended use as a compute-only accelerator.

Where Each One Wins

The benchmark data shows a clear winner in compute workloads, but the architectural differences suggest distinct use cases. The AMD Radeon PRO V620 wins decisively in raw compute performance. Its OpenCL score of 128,580 versus 87,445 for the GP100, a 47% lead, makes it the better choice for general-purpose GPU compute tasks such as machine learning inference, scientific simulations, and rendering pipelines that rely heavily on FP32 arithmetic. The V620’s FP32 throughput is rated at 20.28 TFLOPS, almost double the GP100’s 10.34 TFLOPS. For FP16 workloads, the V620 offers 40.55 TFLOPS against 20.69 TFLOPS for the GP100, again a 2:1 advantage.

The AMD card also wins on memory capacity. With 32 GB of GDDR6, it can hold larger datasets and models in memory compared to the GP100’s 16 GB. This is critical for workloads like large-scale data analytics, complex 3D scene rendering, or training larger neural networks, where fitting data into VRAM is often the limiting factor. The V620’s higher shading unit count, TMUs, and ROPs also make it more capable in rasterization-heavy tasks, even though it lacks display outputs, meaning it would be used purely as a render or compute node.

The NVIDIA Quadro GP100 has one clear advantage from the recorded data: memory bandwidth. Its HBM2 memory provides 732.2 GB/s, which is 43% higher than the V620’s 512.0 GB/s. For workloads that are heavily bandwidth-bound, such as certain sparse matrix operations, fluid dynamics simulations, or large data shuffling, the GP100 may hold its own despite lower compute throughput. The GP100 also has a lower TDP at 235 W versus 300 W, and requires only a single 8-pin power connector, making it easier to integrate into existing systems with less demanding power delivery.

The GP100 also has display outputs, with one DVI and four DisplayPort 1.4a connections, whereas the V620 has none. This means the GP100 can function as a display-capable workstation card, whereas the V620 is strictly for headless compute. For users who need a single card that can both drive monitors and accelerate compute workloads, the GP100 offers that flexibility. The V620 would require a separate display adapter.

In terms of software support, the V620 supports newer API versions, including DirectX 12 Ultimate and Vulkan 1.4, which may be relevant for applications leveraging the latest graphics features. The GP100’s Vulkan 1.3 and DirectX 12_1 support are older but still functional for many legacy applications. The V620’s ray tracing cores are a significant feature for any workload that uses hardware-accelerated ray tracing, such as photorealistic rendering in DCC tools, while the GP100 has no such capability.

FAQ

Q: Which GPU has the higher average benchmark score in the database?

A: The AMD Radeon PRO V620 has an average benchmark score of 136,472, while the NVIDIA Quadro GP100 has an average of 87,445. The V620 also ranks in the 96th percentile of all GPUs, compared to the GP100’s 93rd percentile.

Q: What is the performance difference in the Geekbench OpenCL test?

A: The AMD Radeon PRO V620 scores 128,580, and the NVIDIA Quadro GP100 scores 87,445. The V620 is 47% faster in this test.

Q: How do the memory capacities compare?

A: The AMD Radeon PRO V620 has 32 GB of GDDR6 memory, while the NVIDIA Quadro GP100 has 16 GB of HBM2 memory. The V620 has twice the capacity but lower bandwidth: 512.0 GB/s versus 732.2 GB/s.

Q: Which card has higher memory bandwidth?

A: The NVIDIA Quadro GP100 has higher memory bandwidth at 732.2 GB/s, thanks to its 4096-bit HBM2 interface. The AMD Radeon PRO V620 has 512.0 GB/s over a 256-bit GDDR6 bus.

Q: What are the transistor counts and process nodes?

A: The AMD Radeon PRO V620 uses 26,800 million transistors on a 7 nm process, while the NVIDIA Quadro GP100 uses 15,300 million transistors on a 16 nm process. The V620’s die is 520 mm², and the GP100’s die is 610 mm².

Q: Which card supports ray tracing?

A: Only the AMD Radeon PRO V620 has ray tracing cores, with 72 of them. The NVIDIA Quadro GP100 has no ray tracing hardware.

Q: What are the power requirements?

A: The AMD Radeon PRO V620 has a TDP of 300 W and needs two 8-pin power connectors with a suggested 700 W power supply. The NVIDIA Quadro GP100 has a TDP of 235 W, requires one 8-pin connector, and suggests a 550 W power supply.

The Verdict

The data points to the AMD Radeon PRO V620 as the superior compute accelerator in this pairing. Its 47% lead in the OpenCL benchmark, combined with its 96th percentile ranking versus 93rd for the GP100, makes it the clear choice for tasks that demand high FP32 and FP16 throughput. The V620’s 20.28 TFLOPS of FP32 performance and 40.55 TFLOPS of FP16 performance dwarf the GP100’s 10.34 and 20.69 TFLOPS respectively. For any workload centered on raw compute, such as AI inference, scientific computing, or batch rendering, the V620 is the better performer.

The V620 also wins on memory capacity with 32 GB versus 16 GB, which is essential for large datasets or models that must reside entirely in VRAM. Its newer architecture, 7 nm process, higher clock speeds, and ray tracing support give it a generational advantage that the GP100 cannot overcome. The only areas where the GP100 leads are memory bandwidth, power draw, and display connectivity. Its 732.2 GB/s bandwidth is higher, and its 235 W TDP with a single 8-pin connector makes it easier to power. It also offers display outputs, which the V620 lacks entirely.

For a user who needs a compute card with maximum performance and has the power delivery and cooling to handle 300 W, the AMD Radeon PRO V620 is the data-backed recommendation. For a user with a legacy system that has limited power headroom, or who needs a single card to drive displays while also doing compute, the NVIDIA Quadro GP100 has those specific advantages. But the benchmark record is unambiguous: in head-to-head compute performance, the AMD Radeon PRO V620 is the faster card by a wide margin. The GP100’s only path to relevance is its bandwidth and display capabilities, which are niche benefits in the face of a 47% compute deficit.