AMD Radeon R9 M380 vs NVIDIA Quadro RTX 4000 Comparison

The Quadro RTX 4000 and Radeon R9 M380 occupy very different corners of the GPU landscape: a single-slot workstation card built on NVIDIA's Turing architecture and a mobile-class AMD part from the GCN 2.0 era. The database records only one common benchmark between them, and it is not close. What follows is a breakdown of that head-to-head result, the architectural gulf underneath it, and what the recorded data says about where each card makes sense.

Head-to-Head Benchmarks

The only test with recorded results for both GPUs is Geekbench OpenCL, and it produces a landslide. The Quadro RTX 4000 scores 74540; the Radeon R9 M380 scores 12565. That is a 493.2 percent gap in NVIDIA's favor, one of the more lopsided deltas in the database. In practical terms, the RTX 4000 delivers roughly six times the OpenCL compute throughput of the M380.

Context sharpens the picture. The RTX 4000's overall average benchmark score sits at 17789, placing it in the 61st percentile against all GPUs in the database. Its nearest rivals by average score include the GeForce RTX 4060 (17639, a 0.9 percent gap), the Radeon HD 7790 (17666, 0.7 percent), the Radeon 780M (17588, 1.1 percent), and the Radeon Pro 560 (17551, 1.4 percent). The M380's average score is 15521, good for the 58th percentile, with rivals including the GeForce GTX 1080 Ti (15548, a mere 0.2 percent apart), the Radeon Pro W5500 (15679, 1 percent), the RTX 2060 (15290, 1.5 percent), and the GTX 580 (15283, 1.6 percent).

The percentile gap of just three points is misleading, since the two cards were measured on largely different test suites. The RTX 4000 carries a full set of Passmark DirectX 9 through DirectX 12 results, a 3DMark Steel Nomad DX12 score of 1873, a Geekbench Vulkan score of 78844, and a Passmark GPU compute score of 6176. The M380's database entry includes a Geekbench Metal score of 18476 alongside its OpenCL result. Where the measurements overlap, the outcome is decisive: one win for the RTX 4000, zero for the M380.

Architecture Differences

These GPUs are separated by three years and two process-node generations. The Quadro RTX 4000 arrived on November 12, 2018, using the TU104 chip on TSMC's 12 nm process with 13,600 million transistors packed into a 545 mm² die, yielding a density of 25.0M per mm². The Radeon R9 M380 dates to May 4, 2015, built on the Strato chip at 28 nm with 2,080 million transistors across a 160 mm² die at 13.0M per mm². Roughly six and a half times the transistor budget explains much of the performance spread.

The core counts diverge just as sharply. The RTX 4000 fields 2304 shading units, 144 TMUs, 64 ROPs, 36 RT cores, and 288 tensor cores. The M380 has 768 shading units, 48 TMUs, 16 ROPs, and no RT or tensor hardware at all. Clocks tell a similar story: 1005 MHz base and 1545 MHz boost for the Quadro versus 900 MHz base and 1000 MHz boost for the Radeon.

Memory is another major differentiator. The RTX 4000 pairs 8 GB of GDDR6 on a 256-bit bus running at 13 Gbps effective, for 416.0 GB/s of bandwidth. The M380 offers 4 GB of GDDR5 on a 128-bit bus at 6 Gbps effective, delivering 96.00 GB/s. That is less than a quarter of the Quadro's bandwidth, and half the capacity.

Theoretical throughput figures compound the gap. The RTX 4000 manages 7.119 TFLOPS FP32, 14.24 TFLOPS FP16 at a 2:1 ratio, 98.88 GPixel/s of pixel fill, and 222.5 GTexel/s of texture fill. The M380 records 1.536 TFLOPS FP32, 16.00 GPixel/s, and 48.00 GTexel/s, with no FP16 figure in the database. Software support follows hardware: the Quadro reports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the M380 tops out at DirectX 12 (12_0) and Vulkan 1.2.170, though both list OpenGL 4.6.

Physical and power characteristics differ as well. The RTX 4000 is a single-slot desktop card measuring 241 mm long and 111 mm tall, drawing 160 W through one 8-pin connector, with a suggested 450 W PSU and outputs of 3x DisplayPort 1.4 plus USB Type-C. The M380's database entry records no TDP, slot width, power connector, display output, or dimension data, consistent with its mobile-oriented identity within the Gem System (R9 M300) generation. Both are end-of-life, both use PCIe 3.0 x16, and both were fabbed at TSMC. The RTX 4000 launched at 899 USD MSRP; no launch MSRP is recorded for the M380.

Where Each One Wins

On recorded data, the Quadro RTX 4000 wins everywhere the two can be directly compared. Its OpenCL result of 74540 against 12565 makes it the only viable choice of the pair for GPU compute workloads, and its dedicated tensor and RT cores, absent entirely on the M380, point it at ray-traced rendering and accelerated inference tasks. Its Passmark G3D score of 15117 and its Passmark DirectX results (205 for DX9, 128 for DX11, 108 for DX10, 52 for DX12) round out a broad workload profile the M380 simply does not have entries for.

The M380's strongest recorded showing is its Geekbench Metal score of 18476, a result that reflects its place in Apple-ecosystem systems given the Metal API. That is its lone notable database entry beyond OpenCL, and it is the only dimension where the card demonstrates competitive output in its own context, sitting within 1.5 percent of cards like the RTX 2060 and GTX 1080 Ti by average score. For legacy mobile scenarios where a compact, modest GCN part was the shipping configuration, the M380 filled that role; the data shows it as a fixed constraint of that hardware class rather than a contender against a full workstation card.

FAQ

Q: What is the score difference in the shared benchmark?

A: In Geekbench OpenCL, the Quadro RTX 4000 scores 74540 versus 12565 for the Radeon R9 M380, a 493.2 percent lead for NVIDIA.

Q: How much more memory bandwidth does the RTX 4000 have?

A: 416.0 GB/s versus 96.00 GB/s, thanks to GDDR6 on a 256-bit bus at 13 Gbps effective compared to GDDR5 on a 128-bit bus at 6 Gbps effective.

Q: Which card supports ray tracing hardware?

A: Only the Quadro RTX 4000, with 36 RT cores and 288 tensor cores. The M380 has neither.

Q: How do their average benchmark scores and percentiles compare?

A: The RTX 4000 averages 17789 at the 61st percentile; the M380 averages 15521 at the 58th percentile.

Q: Do they support the same DirectX feature level?

A: No. The RTX 4000 reports DirectX 12 Ultimate (12_2); the M380 reports DirectX 12 (12_0).

Q: How much larger is the RTX 4000's die and transistor count?

A: 545 mm² with 13,600 million transistors versus 160 mm² with 2,080 million transistors.

The Verdict

The data leaves no ambiguity. In the single directly comparable benchmark, the Quadro RTX 4000 beats the Radeon R9 M380 by 493.2 percent, and every architectural measurement, from FP32 throughput (7.119 versus 1.536 TFLOPS) to memory bandwidth to feature support, points the same direction. Buyers or system evaluators needing compute, modern API support through DirectX 12 Ultimate and Vulkan 1.4, dedicated ray tracing and tensor hardware, or 8 GB of fast GDDR6 should land on the RTX 4000 without hesitation. It is the workstation-class part of the two by every recorded metric.

The M380's case rests entirely on context rather than performance. It exists within mobile Gem System hardware, posts its best result in Metal, and belongs to a generation that predates Turing by three years and two process nodes. Within its own historical class, its average score of 15521 keeps it within striking distance of cards like the GTX 1080 Ti and RTX 2060 by the database's rival clustering. But as a head-to-head matchup, this is not a contest: one win for the RTX 4000, zero for the M380, and a 493.2 percent gap in the only test they share.