AMD Radeon R9 M390X vs NVIDIA GeForce RTX 2070 SUPER Comparison

# AMD Radeon R9 M390X vs NVIDIA GeForce RTX 2070 SUPER

The AMD Radeon R9 M390X and NVIDIA GeForce RTX 2070 SUPER represent two very different eras of GPU design, separated by four years of architectural evolution. The R9 M390X is a 28nm GCN 3.0 part built for mobile MXM modules, while the RTX 2070 SUPER is a 12nm Turing desktop card with dedicated ray tracing and tensor cores. Their average benchmark scores sit remarkably close—20,662 for the AMD versus 20,282 for the NVIDIA—yet the only shared benchmark tells a dramatically different story. This analysis digs into what those numbers mean, where each card excels, and why the data paints a more complex picture than raw averages suggest.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon R9 M390X posts an average benchmark score of 20,662, which is 380 points higher than the NVIDIA GeForce RTX 2070 SUPER’s 20,282. This places the AMD card at the 66th percentile of all GPUs, while the NVIDIA sits at the 65th percentile.

Q: How do the two cards compare in the Geekbench Vulkan benchmark?

A: The RTX 2070 SUPER scores 90,637 in Geekbench Vulkan, while the R9 M390X scores 16,951. The NVIDIA card leads by 81.3% in this test, which is the only head-to-head benchmark available in the data.

Q: What are the nearest performance rivals for each card according to average score?

A: For the R9 M390X, the closest rival is the AMD Radeon RX 5600 XT (20,713, just 0.2% higher). For the RTX 2070 SUPER, the nearest rival is the NVIDIA Quadro M4000M (20,480, 1% lower). Both cards also sit near the Intel Arc A750 and B570, as well as the NVIDIA RTX 3070 Mobile.

Q: What memory configurations do the two GPUs use?

A: The R9 M390X has 4 GB of GDDR5 memory on a 256-bit bus, delivering 160.0 GB/s bandwidth. The RTX 2070 SUPER has 8 GB of GDDR6 on a 256-bit bus, providing 448.0 GB/s bandwidth.

Q: Which GPU has more shading units and texture mapping units?

A: The RTX 2070 SUPER has 2,560 shading units and 160 TMUs, compared to the R9 M390X’s 2,048 shading units and 128 TMUs. The NVIDIA card also has 64 ROPs versus 32 on the AMD.

Q: What is the release date and production status of each card?

A: The R9 M390X was released on May 4, 2015, and the RTX 2070 SUPER on July 8, 2019. Both are marked as end-of-life products in the data.

Architecture Differences

The architectural gap between these two GPUs is vast, beginning with manufacturing process. The R9 M390X uses a 28 nm process at TSMC, while the RTX 2070 SUPER is built on a 12 nm process, also at TSMC. This translates to a major transistor difference: the AMD chip, codenamed Amethyst, packs 5,000 million transistors on a 366 mm² die, yielding a density of 13.7 million transistors per mm². The NVIDIA TU104 chip contains 13,600 million transistors on a 545 mm² die, achieving 25.0 million per mm²—nearly double the density.

The R9 M390X is built on GCN 3.0 architecture, part of the Gem System generation (R9 M300 series), with a predecessor in the Solar System and successor in Polaris Mobile. The RTX 2070 SUPER belongs to the GeForce 20-series, using Turing architecture with a GeForce 10 predecessor and GeForce 30 successor. Turing introduces features entirely absent from GCN 3.0: 40 ray tracing cores and 320 tensor cores, dedicated hardware for real-time ray tracing and AI-accelerated workloads. The R9 M390X has no such units.

Compute capabilities differ sharply. The R9 M390X delivers 2.961 TFLOPS of FP32 and the same for FP16 at a 1:1 ratio. The RTX 2070 SUPER offers 9.062 TFLOPS FP32 and 18.12 TFLOPS FP16 at a 2:1 ratio, meaning its FP16 throughput is more than six times higher. Pixel and texture rates follow suit: 23.14 GPixel/s and 92.54 GTexel/s for the AMD, versus 113.3 GPixel/s and 283.2 GTexel/s for the NVIDIA.

API support also diverges. The R9 M390X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The RTX 2070 SUPER supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, the latter being a significantly newer specification. The NVIDIA card’s DirectX 12 Ultimate tier includes features like mesh shaders and variable rate shading, which the older AMD architecture cannot expose.

Where Each One Wins

Based on the available data, the R9 M390X wins the average benchmark score battle, but that metric aggregates different test suites. The AMD card’s benchmark results come from Geekbench Metal (24,373) and Geekbench Vulkan (16,951), giving it an average of 20,662. The NVIDIA card’s results span many more tests—including 3DMark Steel Nomad DX12 (1,651), Geekbench OpenCL (83,358), Geekbench Vulkan (90,637), and multiple Passmark tests—with its average of 20,282 pulled down by low Passmark scores in legacy DX10 (132) and DX12 (67) tests.

The R9 M390X wins on total average score and percentile placement (66th vs 65th). It also holds a narrower performance delta to its nearest rival (0.2% behind the RX 5600 XT) compared to the RTX 2070 SUPER’s 1% gap to the Quadro M4000M. In raw compute tests like Geekbench Vulkan, however, the RTX 2070 SUPER dominates decisively, scoring 90,637 versus 16,951—a 5.3x advantage. The NVIDIA card also shines in Geekbench OpenCL with 83,358 and in Passmark G3D with 18,169, indicating strong general-purpose compute and Direct3D performance.

The RTX 2070 SUPER also wins on specification sheets across the board: more memory, higher bandwidth, faster clocks, more shading units, higher pixel/texture rates, and dedicated RT and tensor cores. Its TDP of 215 W versus 75 W reflects a desktop-oriented design, while the R9 M390X is an MXM module for portable devices.

Specification Differences

The most obvious difference is form factor: the R9 M390X is an MXM Module with no power connectors and portable-device-dependent display outputs, while the RTX 2070 SUPER is a Dual-slot desktop card measuring 267 mm in length, 116 mm in height, and 35 mm in width, requiring a 1x 6-pin plus 1x 8-pin power connection and a 550 W suggested PSU.

Clock speeds differ substantially. The R9 M390X lists no base or boost clock, only a memory clock of 1250 MHz (5 Gbps effective). The RTX 2070 SUPER has a base clock of 1605 MHz and boost of 1770 MHz, with memory at 1750 MHz (14 Gbps effective). Memory capacity and type also diverge: 4 GB GDDR5 versus 8 GB GDDR6, with bandwidth of 160.0 GB/s versus 448.0 GB/s.

The RTX 2070 SUPER has 2,560 shading units, 160 TMUs, 64 ROPs, 40 RT cores, and 320 tensor cores. The R9 M390X has 2,048 shading units, 128 TMUs, 32 ROPs, and no RT or tensor cores. Compute rates: the AMD offers 2.961 TFLOPS FP32 and FP16, while the NVIDIA offers 9.062 TFLOPS FP32 and 18.12 TFLOPS FP16. Pixel rate is 23.14 GPixel/s for AMD versus 113.3 GPixel/s for NVIDIA; texture rate is 92.54 GTexel/s versus 283.2 GTexel/s.

Power consumption is a huge differentiator: 75 W TDP for the R9 M390X versus 215 W for the RTX 2070 SUPER. The NVIDIA card also supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the AMD supports DirectX 12 (12_0) and Vulkan 1.2.170. The RTX 2070 SUPER has a launch MSRP of 499 USD; the R9 M390X has no listed launch MSRP.

Head-to-Head Benchmarks

The single head-to-head benchmark in the data is Geekbench Vulkan, and it is a landslide. The RTX 2070 SUPER scores 90,637, while the R9 M390X scores 16,951, resulting in a delta of -81.3% for the AMD card. This means the NVIDIA card is approximately 5.35 times faster in this test, a gap that dwarfs the 1.9% difference in average benchmark scores.

This divergence invites investigation. The Geekbench Vulkan test likely stresses modern compute workloads that benefit from the RTX 2070 SUPER’s newer architecture, higher clock speeds, and greater memory bandwidth. The R9 M390X’s 160 GB/s bandwidth and 2.961 TFLOPS FP32 simply cannot compete with the NVIDIA card’s 448 GB/s and 9.062 TFLOPS. The 40 RT cores and 320 tensor cores may also accelerate certain Vulkan compute paths, though the test is not explicitly ray tracing or AI-focused.

Other benchmark results reinforce the NVIDIA card’s compute strength. Geekbench OpenCL gives the RTX 2070 SUPER 83,358, and Passmark G3D gives 18,169. The R9 M390X’s Geekbench Metal score of 24,373 suggests it performs better in Apple’s Metal API than in Vulkan, hinting that its architecture may be more optimized for certain legacy or mobile workloads. The RTX 2070 SUPER’s Passmark scores vary widely—from 67 in DX12 to 223 in DX9—indicating inconsistent performance across different DirectX versions, possibly due to driver overhead or test-specific quirks.

The Verdict

The data supports a clear split: the RTX 2070 SUPER is the superior compute and gaming GPU by a wide margin, while the R9 M390X wins only on the narrow metric of average benchmark score. If the priority is raw performance in modern APIs like Vulkan or OpenCL, the NVIDIA card is the obvious choice—its 90,637 Vulkan score versus 16,951 for the AMD is a 5.3x gap that no other metric can offset.

However, the R9 M390X’s 75 W TDP and MXM form factor make it a compelling option for portable or low-power systems, where the RTX 2070 SUPER’s 215 W TDP and dual-slot desktop design would be impractical. The AMD card also edges out the NVIDIA in percentile ranking (66th vs 65th) and average score (20,662 vs 20,282), suggesting that in aggregate, across diverse test suites, it holds its own surprisingly well.

For a desktop user seeking maximum performance, the RTX 2070 SUPER is the data-backed choice, despite its higher power draw and 499 USD launch MSRP. For a mobile or embedded application where power efficiency and compact size matter more than peak compute, the R9 M390X remains relevant. The benchmarks reveal that while the NVIDIA card dominates in modern compute, the AMD card’s consistency across its limited test set keeps it competitive in overall averages—a reminder that no single benchmark tells the whole story.