AMD Radeon Pro 575X vs AMD Radeon RX 590 GME Comparison

# The Verdict

The data splits cleanly along generational lines. The Radeon Pro 575X is a 2019-era mobile/workstation part built on the Ellesmere die with GCN 4.0, while the RX 590 GME is a 2020 desktop Polaris 20 chip on the same fundamental architecture. Across the recorded benchmarks, the RX 590 GME takes both head-to-head matchups, and its average benchmark score sits about 20% higher than the Pro 575X. However, the Pro 575X posts a better percentile ranking against the entire GPU database (82nd vs 77th), suggesting it is the more consistent performer relative to its peer group. For a user selecting strictly from the recorded data, the RX 590 GME is the faster card, but the Pro 575X holds a stronger relative standing among all tested GPUs.

Architecture Differences

Both cards are built by AMD on GlobalFoundries' 14 nm process and share a GCN-family design, but they are not identical silicon. The Pro 575X uses the Ellesmere chip, while the RX 590 GME uses Polaris 20. The die size is effectively the same (232 mm² for the Pro, 232 mm² for the GME). Transistor counts are nearly identical, with the Pro at 5,700 million and the GME at 5,700 million, yielding a density of 24.6M per mm² versus 24.6M per mm². The core count scales in the same direction as the clock speeds: the Pro 575X has 2048 shading units and 144 texture mapping units, while the RX 590 GME has 2304 shading units and 144 TMUs. Raster operation pipelines are identical at 32 per part. The Pro 575X's memory runs at a 1695 MHz data rate with an effective throughput of 6.8 Gbps, while the GME's memory is rated at 2000 MHz with 8 Gbps effective. The bus width is the same 256-bit on both, but the GME's total bandwidth reaches 256.0 GB/s versus 217.0 GB/s for the Pro. Memory size and type also differ: 4 GB of GDDR5 for the Pro, 8 GB of GDDR5 for the GME.

The Pro 575X uses a 150 W thermal design power target and requires no external power connector, whereas the RX 590 GME is rated for 175 W and needs a single 8-pin connector. The Pro has a PCIe 3.0 x16 interface and its display outputs are portable-device dependent, but the GME provides 1x HDMI 2.0b and 3x DisplayPort 1.4a. API support overlaps: DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3 are supported by both, but the GME also lists Vulkan 1.3 while the Pro's pack lists the same. The Pro was released in March 2019 and is now end-of-life; the GME came in March 2020 and is also end-of-life. No successor or predecessor is listed for the Pro, while the GME's predecessor is Arctic Islands and its successor is Vega.

Where Each One Wins

Looking at the recorded head-to-head results, the RX 590 GME is the clear winner. In Geekbench OpenCL it scores 36,294 versus 34,773 for the Pro 575X, a 4.2% advantage. The gap widens dramatically in Geekbench Vulkan, where the GME posts 60,508 against the Pro's 37,919, a 37.3% lead. Those are the only two direct comparisons in the database, and the GME wins both.

Outside of direct matchups, the Pro 575X actually posts a higher percentile rank (82) than the GME (77), and its average benchmark score of 39,116 is slightly higher than the GME's 32,601. That is a curious inversion: the Pro looks better relative to all GPUs even though it loses the direct comparisons. The GME's raw speed in Vulkan shows a large synthetic advantage, and its higher memory capacity (8 GB vs 4 GB) and bandwidth (256.0 GB/s vs 217.0 GB/s) explain much of the gap in compute-heavy tests. The Pro's lower TDP (150 W vs 175 W) and no external power requirement make it the easier card to fit in a portable chassis, but the GME's additional memory and faster clocked memory give it the edge in sustained workloads.

The Pro 575X is the pick if the workload is light and the user prefers a card that lands in the database's top percentile of all GPUs. The RX 590 GME is the pick if the workload is Vulkan-heavy or involves large textures, where its 8 GB frame buffer and 37.3% higher Vulkan score come into play.

FAQ

Q: Is the Radeon Pro 575X faster than the RX 590 GME?

A: In direct head-to-head tests, no. The RX 590 GME wins both recorded matchups: OpenCL by 4.2% and Vulkan by 37.3%. However, the Pro 575X sits in a higher overall percentile (82nd vs 77th), meaning it outperforms a larger fraction of the database.

Q: Which GPU has more cores?

A: The RX 590 GME has 2304 shading units and 144 texture mapping units, while the Pro 575X has 2048 and 128, respectively.

Q: Do they use the same memory type?

A: Yes, both use GDDR5. The Pro 575X packs 4 GB on a 256-bit bus at 1695 MHz (6.8 Gbps effective), the RX 590 GME uses 8 GB on the same 256-bit bus at 2000 MHz (8 Gbps effective).

Q: Which card is older?

A: The Radeon Pro 575X launched on March 17, 2019 and is now end-of-life. The RX 590 GME launched on March 8, 2020, and is also end-of-life. The RX GME's predecessor is Arctic Islands; no successor is listed.

Q: Which card is more power hungry?

A: The RX 590 GME has a 175W board power and needs one 8-pin power connector. The Pro 575X is rated at 150W and draws no external power connector.

Head-to-Head Benchmarks

Two recorded benchmark tests settle the question. The first, Geekbench OpenCL, scores the Pro 575X at 34,773 and the RX 590 GME at 36,294, a win for the RX 590 GME by 4.2%. The second, Geekbench Vulkan, is a much larger win for the RX 590 GME: 60,508 to 37,919, a 37.3% margin. In average benchmark terms, the Pro 575X scores 39,116 and the RX 590 GME scores 32,601, so the Pro actually averages 6% higher despite losing both direct matchups.

The RX 590 GME's wins are tied to its hardware: a wider memory bus is present in both (256-bit), but the GME's memory runs at 2000 MHz versus 1695 MHz and its bandwidth is 256.0 GB/s versus 217.0 GB/s. That translates directly to the larger Vulkan gap, which is a synthetic workload. The Pro 575X doesn't have any of that, so it wins the percentile battle: 82% of all GPUs versus 77%.

Specification Differences

| Field | AMD Radeon Pro 575X | AMD Radeon RX 590 GME |

| --- | --- | --- |

| Chip | Ellesmere | Polaris 20 |

| Architecture | GCN 4.0 | GCN 4.0 |

| Process | 14 nm | 14 nm |

| Transistors | 5,700 million | 5,700 million |

| Die size | 232 mm² | 232 mm² |

| Density | 24.6M / mm² | 24.6M / mm² |

| Shading units | 2048 | 2304 |

| TMUs | 128 | 144 |

| ROPs | 32 | 32 |

| Memory | 4 GB GDDR5 | 8 GB GDDR5 |

| Memory clock | 1695 MHz (6.8 Gbps effective) | 2000 MHz (8 Gbps effective) |

| Bus width | 256-bit | 256-bit |

| Bandwidth | 217.0 GB/s | 256.0 GB/s |

| FP32 | 4.489 TFLOPS | 6.543 TFLOPS |

| FP16 | 4.489 TFLOPS (1:1) | 6.543 TFLOPS (1:1) |

| TDP | 150 W | 175 W |

| Power connectors | None | 1x 8-pin |

| Bus interface | PCIe 3.0 x16 | PCIe 3.0 x16 |

| Display outputs | Portable Device Dependent | 1x HDMI 2.0b, 3x DisplayPort 1.4a |

| APIs | DirectX 12 (12_0), OpenGL 4.6, Vulkan 1.3 | DirectX 12 (12_0), OpenGL 4.6, Vulkan 1.3 |

| Release | 2019-03-17 | 2020-03-08 |

| Status | End-of-life | End-of-life |