AMD Radeon RX 590 vs Intel Arc A750 Comparison

The AMD Radeon RX 590 and Intel Arc A750 occupy adjacent tiers in the database’s aggregate rankings, yet their recorded performance profiles diverge sharply. The RX 590 sits at the 70th percentile of all GPUs, with an average benchmark score of 24,744. The Arc A750 holds the 66th percentile, averaging 20,582 points. That 4,162-point gap (about 20%) in overall averages does not tell the whole story, because the head-to-head data shows a single dominant result: Intel’s card wins the one shared test by a massive margin. This comparison hinges on how much weight that one test carries, and what each card’s strengths imply for different workloads.

Head-to-Head Benchmarks

The only directly comparable measurement in the database is 3DMark Steel Nomad (DX12). Here, the Arc A750 scores 2,612, while the RX 590 manages 1,081. The delta is -58.6% from the AMD card’s perspective, meaning Intel’s part delivers 2.4 times the raw DX12 performance in this synthetic scenario. This is not a marginal win; it is a categorical one. The Arc A750’s score places it in a different performance class for modern API workloads, which is expected given the architectural gulf between the two.

However, the aggregate benchmark set paints a more nuanced picture. The RX 590’s average score of 24,744 comes from two tests: the same Steel Nomad run (1,081) and a Geekbench Metal score of 48,407. The Arc A750’s average of 20,582 is drawn from ten tests, including Geekbench OpenCL (98,554) and Vulkan (85,631), plus a suite of Passmark tests. The RX 590’s Metal score is not directly comparable to any Arc A750 result, so the cross-card comparison relies heavily on the Steel Nomad outlier. If we isolate that shared benchmark, Intel wins 1-0. There are no recorded wins for the RX 590 in any head-to-head test, because no other common test exists in the database.

The Arc A750’s dominance in Steel Nomad is consistent with its other synthetic numbers. Its Geekbench Vulkan score of 85,631 and OpenCL score of 98,554 are far above what the RX 590’s architecture would likely produce (though no direct comparison is recorded). The RX 590’s only non-Steel-Nomad score, Metal at 48,407, is a strong showing for an older card, but Metal is Apple-specific and does not translate to Windows or Linux gaming. In practical terms, the head-to-head data says: if you run DX12 workloads, the Arc A750 is the clear choice, and the RX 590 does not have a countervailing win in any shared test.

Where Each One Wins

The RX 590’s recorded advantages are not found in head-to-head wins; they are inferred from its aggregate percentile and rival deltas. Its nearest rival, the NVIDIA RTX A5000 Mobile, scores 24,763, which is -0.1% relative to the RX 590. That means the RX 590 is statistically tied with a mobile workstation GPU, and it sits 1.2% ahead of the AMD Radeon RX 6600 XT (24,442) and 1.9% ahead of the NVIDIA GeForce GTX 1630 (24,277). These margins are small, but they suggest the RX 590 remains competitive in legacy or driver-optimized titles that do not stress modern APIs. Its 70th percentile ranking indicates it outperforms roughly 70% of all GPUs in the database, which is respectable for a 2018 product.

The Arc A750, by contrast, shows its wins in raw compute and modern API support. Its 66th percentile is lower than the RX 590’s, but that is because its average is dragged down by several Passmark scores that are low for a card of its class: for example, Passmark DirectX 10 (65), DirectX 11 (72), and DirectX 12 (70) are all single-digit or low-double-digit scores, likely reflecting driver overhead or test-specific quirks. However, its Passmark G3D score is 12,534, and its GPU Compute score is 5,368, which are strong. The Arc A750’s Geekbench Vulkan and OpenCL scores are its standout results, indicating excellent compute throughput for non-gaming workloads like rendering or machine learning inference.

Where each one wins comes down to workload era. The RX 590 wins in legacy DirectX 9/10/11 scenarios (if we extrapolate from its higher aggregate percentile and older GCN architecture, which was optimized for those APIs), while the Arc A750 wins decisively in DX12 and compute-heavy tasks. The data shows no recorded win for the RX 590 in any shared test, so its “win” is purely positional: it ranks higher in the overall database, likely due to a broader set of compatible tests and a mature driver stack.

Architecture Differences

The two cards are separated by four years and two fundamental design philosophies. The RX 590 uses GCN 4.0 architecture on a 12 nm process from GlobalFoundries. It packs 5,700 million transistors on a 232 mm² die, yielding a transistor density of 24.6 million per mm². The chip is Polaris 30, and it uses GDDR5 memory with 8 GB capacity, a 256-bit bus, and 256.0 GB/s bandwidth. Its clock speeds are modest: 1,469 MHz base and 1,545 MHz boost. The compute configuration includes 2,304 shading units, 144 texture mapping units, and 32 ROPs. Pixel rate is 49.44 GPixel/s, texture rate is 222.5 GTexel/s, and FP32 throughput is 7.119 TFLOPS. FP16 is identical at 7.119 TFLOPS (1:1), meaning no dedicated half-rate path.

The Arc A750 uses Xe-HPG architecture (generation Alchemist, Arc 7) on a 6 nm TSMC process. It integrates 21,700 million transistors on a 406 mm² die, with a density of 53.4 million per mm², more than double the RX 590’s. The chip is DG2-512. Memory is GDDR6, also 8 GB and 256-bit, but bandwidth doubles to 512.0 GB/s. Clocks run higher: 2,050 MHz base and 2,400 MHz boost. The shader array is larger: 3,584 shading units, 224 TMUs, and 112 ROPs. Pixel rate is 268.8 GPixel/s, texture rate is 537.6 GTexel/s, and FP32 reaches 17.20 TFLOPS. Critically, FP16 is 34.41 TFLOPS (2:1), meaning the Arc A750 has a dedicated fast path for half-precision, useful for AI and some compute workloads. The Arc A750 also includes 28 ray tracing cores, which the RX 590 lacks entirely.

The process node difference (12 nm vs 6 nm) explains the density gap: Intel packs nearly four times the transistors per area. The RX 590’s GCN 4.0 is a veteran architecture with no hardware ray tracing, no mesh shaders, and only DirectX 12 (12_0) support. The Arc A750 supports DirectX 12 Ultimate (12_2), which includes ray tracing, variable rate shading, and mesh shaders. Both cards support OpenGL 4.6, but the Arc A750 supports Vulkan 1.4, while the RX 590 is limited to Vulkan 1.3. The RX 590 uses PCIe 3.0 x16; the Arc A750 uses PCIe 4.0 x16, doubling the bus bandwidth for data transfers.

Power and physical requirements differ too. The RX 590 has a TDP of 175 W, a dual-slot cooler, and a single 8-pin power connector, with a suggested PSU of 450 W. The Arc A750 draws 225 W, also dual-slot, but requires a 6-pin plus an 8-pin connector and a 550 W PSU. Display outputs: the RX 590 offers 1x HDMI 2.0b and 3x DisplayPort 1.4a; the Arc A750 has 1x HDMI 2.1 and 3x DisplayPort 2.0, the latter supporting higher bandwidth for high-refresh monitors.

FAQ

Q: Which card has higher raw FP32 performance?

A: The Intel Arc A750. Its FP32 throughput is 17.20 TFLOPS, compared to the RX 590’s 7.119 TFLOPS. That is roughly 2.4 times the single-precision compute.

Q: Does the RX 590 support ray tracing?

A: No. The RX 590 has no ray tracing cores listed in the database. The Arc A750 includes 28 ray tracing cores, which is a requirement for DirectX 12 Ultimate (12_2) support.

Q: What is the memory bandwidth difference?

A: The Arc A750 has 512.0 GB/s of bandwidth with GDDR6, exactly double the RX 590’s 256.0 GB/s with GDDR5. Both use an 8 GB frame buffer on a 256-bit bus.

Q: Which card has a higher percentile ranking?

A: The RX 590 ranks at the 70th percentile of all GPUs, while the Arc A750 ranks at the 66th percentile. This is despite the Arc A750 winning the only head-to-head test.

Q: Can the Arc A750 use PCIe 4.0?

A: Yes, the Arc A750 supports PCIe 4.0 x16. The RX 590 is limited to PCIe 3.0 x16.

Q: What is the launch MSRP for each?

A: The RX 590 had a launch MSRP of 279 USD. The Arc A750 had a launch MSRP of 289 USD.

The Verdict

The database’s recorded data points to a clear split. For modern DX12 gaming and compute-heavy tasks, the Intel Arc A750 is the superior choice. Its Steel Nomad score of 2,612 versus 1,081 is a 58.6% advantage, and its FP32 compute (17.20 TFLOPS) and memory bandwidth (512.0 GB/s) are double the RX 590’s. The presence of ray tracing cores and DirectX 12 Ultimate support makes it future-proof for current-generation titles. The Arc A750 also offers higher pixel and texture rates: 268.8 GPixel/s and 537.6 GTexel/s, compared to 49.44 GPixel/s and 222.5 GTexel/s.

However, the RX 590’s higher percentile (70 vs 66) and its average score of 24,744 versus 20,582 suggest that in the broader database, it performs more consistently across a wider range of tests. Its nearest rivals are all within 2% of its average score, indicating stable performance across many workloads. The RX 590’s lower power draw (175 W vs 225 W) and single 8-pin connector make it easier to install in older systems with smaller PSUs. Its 2018 release date means mature drivers, while the Arc A750’s 2022 launch came with noted driver maturity issues (reflected in its low Passmark DirectX scores: 65, 72, 70 for DX10, 11, 12).

For users who prioritize legacy API compatibility (DirectX 9/10/11), the RX 590’s GCN architecture and higher aggregate percentile make it a safer pick. For anyone running DX12 games, ray-traced content, or compute workloads like OpenCL or Vulkan compute, the Arc A750 is objectively faster per the recorded data. The Arc A750’s Geekbench Vulkan score (85,631) and OpenCL score (98,554) are far beyond anything the RX 590 could produce, even if no direct comparison exists. The verdict: choose the Arc A750 for future-facing performance and raw throughput; choose the RX 590 if you need a lower-power card that ranks higher in overall database percentile and handles older APIs without driver quirks. The data does not support a universal winner, but for the only shared benchmark, Intel wins decisively.