AMD Radeon RX 590 vs NVIDIA Quadro RTX 8000 Comparison

The NVIDIA Quadro RTX 8000 and the AMD Radeon RX 590 occupy two very different corners of the GPU landscape, and the recorded data makes the split immediately obvious. Both are end-of-life parts released within months of each other, but one is a 48 GB workstation flagship with a launch MSRP of 9,999 USD, while the other is a mainstream Polaris card with a launch MSRP of 279 USD. Despite that gulf, the aggregate database scores land closer than the specifications would suggest: 28421 for the Quadro versus 24744 for the RX 590, a gap of roughly 13 percent. That proximity is worth questioning, because it says as much about how the averages are computed as it does about real-world performance.

Head-to-Head Benchmarks

There are no overlapping test entries between these two cards in the database, so a direct per-test comparison cannot be made. What the data does offer is each card's individual results, and the picture they paint is instructive.

The Quadro RTX 8000's compute credentials dominate its record. It posts 101883 in Geekbench OpenCL and 122637 in Geekbench Vulkan, figures consistent with its 16.31 TFLOPS of FP32 throughput and its 672.0 GB/s of memory bandwidth. The Passmark results add context: 19799 in G3D, 9992 in GPU compute, and 866 in G2D. Curiously, its DirectX-specific Passmark results are tiny by comparison, 211 for DirectX 9, 188 for DirectX 11, 137 for DirectX 10, and just 79 for DirectX 12. Those numbers reflect a workload profile the card was never tuned for, and they drag on the average. The percentile field places it at the 74th percentile versus all GPUs in the database, and its average of 28421 sits within a percent of cards like the AMD Radeon R9 M295X (28580, a delta of -0.6 percent), the AMD FirePro S7150 (28117, +1.1 percent), the AMD Radeon RX 570 (28766, -1.2 percent), and the NVIDIA GeForce GTX 980 Ti (28020, +1.4 percent).

The RX 590's record is far thinner: 1081 in 3DMark Steel Nomad DirectX 12 and 48407 in Geekbench Metal. Its average of 24744 puts it at the 70th percentile, effectively tied with the NVIDIA RTX A5000 Mobile (24763, -0.1 percent), the Intel Arc A350M (24647, +0.4 percent), the AMD Radeon RX 6600 XT (24442, +1.2 percent), and the NVIDIA GeForce GTX 1630 (24277, +1.9 percent). The interesting anomaly here is the RX 590 averaging near an RX 6600 XT while carrying only 7.119 TFLOPS of FP32, less than half the Quadro's. Benchmark averages clearly weight different test mixes for different cards, and any cross-card conclusion drawn from the two averages alone should be treated cautiously. On raw capability, the Quadro's FP32 advantage is unambiguous: 16.31 TFLOPS against 7.119 TFLOPS, more than double.

Where Each One Wins

The Quadro RTX 8000 wins everywhere the silicon is the deciding factor. Its pixel rate of 169.9 GPixel/s is well over three times the RX 590's 49.44 GPixel/s. Its texture rate of 509.8 GTexel/s dwarfs the 222.5 GTexel/s of the Polaris part. Memory bandwidth is 672.0 GB/s versus 256.0 GB/s, and capacity is 48 GB versus 8 GB, a sixfold difference that matters enormously for large datasets, rendering scenes, and GPU compute workloads that spill beyond a mainstream frame buffer. The card also carries 72 RT cores and 576 tensor cores, hardware the RX 590 simply does not have, so ray tracing and accelerated inference workloads fall entirely to the NVIDIA side.

The RX 590's wins are practical rather than computational. Its TDP is 175 W against the Quadro's 260 W, and the suggested PSU is 450 W versus 600 W, so it fits into smaller, less demanding systems. It needs only a single 8-pin connector where the Quadro requires a 6-pin plus an 8-pin. It is shorter, 241 mm against 267 mm. Its API support is current for its class, with DirectX 12 feature level 12_0, Vulkan 1.3, and OpenGL 4.6. Notably, its FP16 rate equals its FP32 rate at 7.119 TFLOPS, a 1:1 ratio, whereas the Quadro doubles FP16 to 32.62 TFLOPS, so even in the half-precision case the advantage stays with the Turing card.

Architecture Differences

These two parts represent opposite design philosophies of the same era. The Quadro RTX 8000 uses the TU102 chip on NVIDIA's Turing architecture, part of the Quadro Turing (Tx000) generation, fabricated at TSMC on a 12 nm process. The RX 590 uses the Polaris 30 chip on AMD's GCN 4.0 architecture, from the Polaris (RX 500) generation, fabricated at GlobalFoundries, also on 12 nm. The node parity makes the silicon divergence more striking: the TU102 packs 18,600 million transistors on a 754 mm² die, while Polaris 30 has 5,700 million transistors on a 232 mm² die. Transistor density is nearly identical, 24.7M per mm² versus 24.6M per mm², so the performance gap comes from die scale and architecture, not packing efficiency.

The execution resource counts tell the story. The Quadro has 4608 shading units, 288 TMUs, and 96 ROPs, against 2304, 144, and 32 respectively on the RX 590. The dedicated RT cores and tensor cores on the NVIDIA side have no counterpart in GCN 4.0, which is a generational feature gap rather than a segment gap. Succession differs too: the Quadro line moved to Workstation Ampere, while the RX 590 was followed by Vega. Their predecessors were Quadro Volta and Arctic Islands.

Clocking strategies differ in shape. The Quadro runs a 1395 MHz base with a 1770 MHz boost, a wide range that suits bursty professional workloads. The RX 590 runs a tighter 1469 MHz base and 1545 MHz boost. Note that the AMD card actually has the higher base clock, yet delivers less than half the FP32 throughput, confirming that resources, not frequency, decide the outcome.

Specification Differences

The fields where these cards diverge are numerous. Memory is the starkest contrast: 48 GB of GDDR6 on a 384-bit bus at 672.0 GB/s for the Quadro, versus 8 GB of GDDR5 on a 256-bit bus at 256.0 GB/s for the RX 590. Memory clocks are 1750 MHz (14 Gbps effective) versus 2000 MHz (8 Gbps effective). FP16 differs in both magnitude and ratio, 32.62 TFLOPS at 2:1 versus 7.119 TFLOPS at 1:1.

Power and cooling differ: 260 W TDP against 175 W, one 6-pin plus one 8-pin connector against a single 8-pin, and a 600 W suggested PSU against 450 W. Dimensions differ at 267 mm versus 241 mm in length; the database records a 111 mm height for the Quadro and none for the RX 590. Both are dual-slot cards on PCIe 3.0 x16, so the platform footprint is shared.

Display output is a differentiator: the Quadro offers 4x DisplayPort 1.4a plus 1x USB Type-C, while the RX 590 offers 1x HDMI 2.0b plus 3x DisplayPort 1.4a. API support is close, with both at OpenGL 4.6, but the Quadro lists DirectX 12 Ultimate (12_2) and Vulkan 1.4 against the RX 590's DirectX 12 (12_0) and Vulkan 1.3. Release dates were months apart in the same year, August for the Quadro and November for the RX 590, and both are now end-of-life.

FAQ

Q: Which card has more raw compute performance?

A: The Quadro RTX 8000, decisively. It delivers 16.31 TFLOPS FP32 and 32.62 TFLOPS FP16 against the RX 590's 7.119 TFLOPS in both cases.

Q: How do their database averages compare?

A: The Quadro averages 28421 at the 74th percentile, while the RX 590 averages 24744 at the 70th percentile, a gap of roughly 13 percent despite the much larger hardware disparity.

Q: Is there direct head-to-head benchmark data between them?

A: No. The database contains no shared test entries for these two cards, so comparisons rely on each card's individual results, percentiles, and nearest rivals.

Q: Which card supports ray tracing hardware?

A: Only the Quadro RTX 8000, with 72 RT cores and 576 tensor cores. The RX 590 has neither.

Q: Which is more power efficient?

A: The RX 590 draws less: 175 W TDP and a 450 W suggested PSU, versus 260 W and 600 W for the Quadro.

Q: How much memory does each card have?

A: The Quadro has 48 GB of GDDR6 with 672.0 GB/s of bandwidth; the RX 590 has 8 GB of GDDR5 with 256.0 GB/s.

The Verdict

The data frames this as a workload decision, not a close contest. Anyone doing GPU compute, large-memory work, or anything touching ray tracing and tensor hardware should be looking at the Quadro RTX 8000: double the FP32 throughput, six times the memory, more than double the bandwidth, and dedicated RT and tensor cores the RX 590 lacks entirely. Its Geekbench OpenCL and Vulkan results of 101883 and 122637 back that up.

The RX 590 makes its case on footprint and efficiency: lower TDP, one power connector, shorter card, and a smaller suggested PSU. Its benchmark average of 24744 puts it within striking distance of far newer hardware in its rival cluster, including an RX 6600 XT within 1.2 percent. For lighter graphics work where the 8 GB buffer suffices, it remains a recorded performer. For everything demanding, the Turing workstation card wins on every measured axis of capability.