AMD Radeon Pro Vega 48 vs NVIDIA CMP 90HX Comparison

NVIDIA CMP 90HX decisively outperforms AMD Radeon Pro Vega 48 in the only shared benchmark, leading by 28.4% in Geekbench OpenCL. The CMP 90HX sits at the 90th percentile of all GPUs, while the Pro Vega 48 ranks at the 88th percentile, a gap that widens further when considering the Pro Vega 48’s average benchmark score across its three available tests. The data shows a clear performance hierarchy, with the CMP 90HX delivering substantially higher raw compute throughput, though the Pro Vega 48 offers distinct advantages in portability and API compatibility.

Head-to-Head Benchmarks

The sole directly comparable benchmark is Geekbench OpenCL, and the result is unambiguous. The NVIDIA CMP 90HX scores 69,000 points, while the AMD Radeon Pro Vega 48 scores 53,757 points. That represents a 28.4% advantage for the NVIDIA part. This is not a marginal victory; it is a commanding lead that places the two cards in different performance tiers despite their relatively close overall percentile rankings.

Contextualizing this score, the CMP 90HX’s nearest rivals in the database are the Intel Arc A770 (68,809 points, 0.3% behind) and the AMD Radeon Instinct MI25 (68,562 points, 0.6% behind). The CMP 90HX also outperforms the AMD Radeon Pro WX 8200 by 1.2% and the NVIDIA Quadro P6000 by 1.4%. This places it in a tight cluster of high-end cards where small percentage differences separate competitors. The 28.4% gap over the Pro Vega 48, therefore, is not just a win — it is a class separation.

The Pro Vega 48’s average benchmark score across all its tests is 60,140, which is 12.8% lower than its own OpenCL score would suggest if it were the only data point. This average pulls in its Geekbench Metal score of 69,010 and its Geekbench Vulkan score of 57,653, both of which are notably higher than its OpenCL result. The Metal score, in particular, nearly matches the CMP 90HX’s OpenCL score, indicating that the Pro Vega 48 is highly optimized for Apple’s Metal API, likely due to its intended use in Mac systems.

When comparing the Pro Vega 48 to its own nearest rivals, the data shows it is competitive within its tier. The Intel Arc Pro A60 scores 60,326 (0.3% higher), and the NVIDIA GeForce RTX 4090 scores 60,347 (0.3% higher). The Pro Vega 48 beats the AMD Radeon PRO V710 by 2.5% and the NVIDIA P102-100 by 2.8%. This suggests that while the Pro Vega 48 is not a top-tier performer in raw OpenCL, it holds its own against a range of professional and consumer cards.

The deltaPct values tell a nuanced story. The CMP 90HX’s rivals are all within 1.4% of its score, meaning it is a marginal leader in a crowded field. The Pro Vega 48’s rivals span a wider range, from 2.8% behind to 0.3% ahead, indicating more variability in its competitive position. The head-to-head delta of 28.4% dwarfs all these intra-tier differences, confirming that the CMP 90HX is fundamentally a faster card for OpenCL workloads.

The Verdict

The data supports a clear verdict: the NVIDIA CMP 90HX is the superior choice for compute-heavy OpenCL workloads. Its 28.4% lead in the shared benchmark is decisive, and its 90th percentile ranking versus the Pro Vega 48’s 88th percentile confirms its higher overall standing. Anyone prioritizing raw OpenCL throughput should select the CMP 90HX without hesitation.

However, the Pro Vega 48 is not without merit. Its Geekbench Metal score of 69,010 is nearly identical to the CMP 90HX’s OpenCL score, suggesting that in Metal-specific environments, the AMD card may be competitive or superior. The Pro Vega 48 also supports Vulkan 1.3, while the CMP 90HX supports Vulkan 1.4, but the AMD card’s higher Metal performance is notable for Mac users. The Pro Vega 48 is an integrated GPU (IGP) with no power connectors, making it suitable for portable devices, whereas the CMP 90HX is a dual-slot card requiring 2x 8-pin connectors and a 700 W power supply.

The CMP 90HX also has a significant hardware advantage in shading units (6,400 vs. 3,072), texture mapping units (200 vs. 192), and raster output pipelines (80 vs. 64). Its FP32 throughput of 21.89 TFLOPS is nearly three times the Pro Vega 48’s 7.373 TFLOPS. These are not minor differences; they are structural advantages that explain the benchmark gap.

For buyers who need a mining GPU with no display outputs, the CMP 90HX is the obvious pick. For those who need a portable, integrated solution with strong Metal performance, the Pro Vega 48 is the only logical choice. The data does not support any other conclusion.

Where Each One Wins

The NVIDIA CMP 90HX wins decisively in OpenCL compute. Its 69,000 score versus 53,757 represents a 28.4% advantage, and its FP32 performance of 21.89 TFLOPS dwarfs the Pro Vega 48’s 7.373 TFLOPS. The CMP 90HX also leads in memory bandwidth (760.3 GB/s vs. 402.4 GB/s) and pixel rate (136.8 GPixel/s vs. 76.80 GPixel/s). For any workload that leverages these resources — scientific computing, rendering, or machine learning — the CMP 90HX is the stronger card.

The AMD Radeon Pro Vega 48 wins in API-specific scenarios. Its Geekbench Metal score of 69,010 is 0.01% higher than the CMP 90HX’s OpenCL score, indicating that in Metal applications, the Pro Vega 48 is on par with or better than the NVIDIA card. Its Vulkan score of 57,653 is lower than the CMP 90HX’s OpenCL score, but the Pro Vega 48 supports Vulkan 1.3, which may be sufficient for many applications. The Pro Vega 48 also has a 2:1 FP16 ratio (14.75 TFLOPS), whereas the CMP 90HX has a 1:1 ratio (21.89 TFLOPS), meaning the AMD card is relatively stronger in FP16 workloads that do not require full FP32 precision.

The Pro Vega 48’s integrated nature is another win. As an IGP with no power connectors, it can be deployed in portable devices where the CMP 90HX’s dual-slot, 320 W design is impossible. The Pro Vega 48’s portable device dependent display outputs also make it flexible for mobile systems, while the CMP 90HX has no outputs at all.

In terms of process technology, the CMP 90HX uses an 8 nm Samsung process with 28,300 million transistors on a 628 mm² die, while the Pro Vega 48 uses a 14 nm GlobalFoundries process with 12,500 million transistors on a 495 mm² die. The CMP 90HX’s transistor density of 45.1M / mm² is nearly double the Pro Vega 48’s 25.3M / mm², explaining its higher performance per unit area.

FAQ

Q: Which GPU has a higher Geekbench OpenCL score?

A: The NVIDIA CMP 90HX scores 69,000, which is 28.4% higher than the AMD Radeon Pro Vega 48’s 53,757.

Q: Does the AMD Radeon Pro Vega 48 outperform the NVIDIA CMP 90HX in any benchmark?

A: The Pro Vega 48’s Geekbench Metal score of 69,010 is slightly higher than the CMP 90HX’s OpenCL score of 69,000, though this is not a direct head-to-head comparison.

Q: What is the average benchmark score for each GPU?

A: The NVIDIA CMP 90HX has an average benchmark score of 69,000, while the AMD Radeon Pro Vega 48 has an average of 60,140 across its three tests.

Q: How do the two GPUs compare in terms of memory bandwidth?

A: The NVIDIA CMP 90HX has 760.3 GB/s of bandwidth, while the AMD Radeon Pro Vega 48 has 402.4 GB/s.

Q: Which GPU has more shading units?

A: The NVIDIA CMP 90HX has 6,400 shading units, compared to 3,072 on the AMD Radeon Pro Vega 48.

Q: Are both GPUs end-of-life products?

A: Yes, both the NVIDIA CMP 90HX and the AMD Radeon Pro Vega 48 have a production status of end-of-life.

Architecture Differences

The NVIDIA CMP 90HX is built on the Ampere architecture using the GA102 chip, manufactured on Samsung’s 8 nm process. It contains 28,300 million transistors on a 628 mm² die, yielding a transistor density of 45.1M / mm². The AMD Radeon Pro Vega 48 uses the GCN 5.0 architecture with the Vega 10 chip, fabricated by GlobalFoundries on a 14 nm process. It has 12,500 million transistors on a 495 mm² die, with a transistor density of 25.3M / mm².

The CMP 90HX features 50 RT cores and 200 tensor cores, which are absent from the Pro Vega 48. These specialized units give the NVIDIA card a significant advantage in ray tracing and AI-accelerated workloads. The CMP 90HX also supports DirectX 12 Ultimate (12_2), while the Pro Vega 48 supports DirectX 12 (12_1). Both support OpenGL 4.6, but the CMP 90HX supports Vulkan 1.4 versus the Pro Vega 48’s Vulkan 1.3.

The memory subsystems are fundamentally different. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus, while the Pro Vega 48 uses 8 GB of HBM2 on a 2048-bit bus. The CMP 90HX’s memory clock is 1188 MHz (19 Gbps effective), whereas the Pro Vega 48’s memory clock is 786 MHz (1572 Mbps effective). Despite the narrower bus, the CMP 90HX achieves higher bandwidth due to faster memory technology.

Specification Differences

The NVIDIA CMP 90HX has a base clock of 1500 MHz and a boost clock of 1710 MHz, while the AMD Radeon Pro Vega 48 has no listed base or boost clocks. The CMP 90HX has a TDP of 320 W, while the Pro Vega 48 has no TDP listed. The CMP 90HX is dual-slot and requires 2x 8-pin power connectors with a suggested 700 W PSU, whereas the Pro Vega 48 is an IGP with no power connectors and no suggested PSU.

The CMP 90HX has a PCIe 1.0 x4 bus interface, while the Pro Vega 48 uses PCIe 3.0 x16. The CMP 90HX has no display outputs, while the Pro Vega 48 has portable device dependent outputs. The CMP 90HX measures 285 mm (11.2 inches) in length and 112 mm (4.4 inches) in height, while the Pro Vega 48 has no listed dimensions.

The CMP 90HX has 200 TMUs and 80 ROPs, versus 192 TMUs and 64 ROPs on the Pro Vega 48. Its FP32 performance is 21.89 TFLOPS, compared to 7.373 TFLOPS. The CMP 90HX’s FP16 performance is also 21.89 TFLOPS (1:1), while the Pro Vega 48’s is 14.75 TFLOPS (2:1). The CMP 90HX was released on 2021-07-27, while the Pro Vega 48 was released on 2019-03-18. Neither card has a listed launch MSRP, predecessor, or successor.