AMD Radeon PRO W6800

The AMD Radeon PRO W6800 is a Navi 21 GPU built on RDNA 2.0 architecture, manufactured by TSMC on a 7 nm process with 26,800 million transistors on a 520 mm² die, and it belongs to the Radeon Pro Navi (Navi II Series) generation. It launched on June 7, 2021 with a launch MSRP of 2,249 USD, and its current production status is end-of-life. The card carries 32 GB of GDDR6 memory on a 256-bit bus, with a 2000 MHz memory clock, 16 Gbps effective, and 512.0 GB/s of bandwidth. Its compute configuration includes 3840 shading units, 240 TMUs, 96 ROPs, and 60 ray-tracing cores, with a base clock of 1575 MHz and a boost clock of 2322 MHz. The predecessor in the data is the Radeon Pro Vega.

Power and Cooling — TDP, PSU recommendation, connector requirements

The Radeon PRO W6800 is rated for a 250 W TDP. The suggested PSU recommendation in the data is 600 W. Power is delivered through one 6-pin and one 8-pin PCIe power connector. The card occupies a dual-slot width, and its dimensions are 267 mm in length, 120 mm in height, and 50 mm in width. It connects to the motherboard via a PCIe 4.0 x16 bus interface.

The underlying Navi 21 die is built on TSMC's 7 nm process and packs 26,800 million transistors across 520 mm², resulting in a transistor density of 51.5M / mm². The base clock of 1575 MHz and boost clock of 2322 MHz define the operating range for the 250 W TDP. At those clocks, the card produces a pixel rate of 222.9 GPixel/s and a texture rate of 557.3 GTexel/s. No game clock field is listed in the specification data, so the card's sustained behavior is described by the base and boost figures only.

How It Compares

The W6800's average benchmark score is 133588, which places it in the 97th percentile among all GPUs in the database. Its four nearest rivals are separated by narrow average-score gaps, and the deltaPct values show how close the field is.

The AMD Radeon RX 9070 GRE reaches an average score of 134417. The database lists a deltaPct of -0.6, meaning the W6800 sits 0.6% behind the RX 9070 GRE in aggregate benchmark performance. Given how small that delta is, the two cards are effectively level in this data set.

The NVIDIA RTX 4000 Ada Generation averages 135218. The deltaPct is -1.2, so the W6800 trails this rival by 1.2%. That is a slightly larger gap than the one to the RX 9070 GRE, but it remains a narrow difference in the benchmark results.

The NVIDIA A10M averages 135230, the highest average score among the four nearest rivals. The deltaPct against the A10M is also -1.2, placing the W6800 1.2% behind the A10M. The W6800's average of 133588 therefore sits close to the A10M's result despite the A10M having the top score in this group.

The NVIDIA GeForce RTX 3090 Ti averages 131911. The deltaPct is 1.3, which means the W6800 leads the RTX 3090 Ti by 1.3% in average benchmark score. This is the only nearest rival in the group that the W6800 outperforms, and it brackets the W6800's average from below.

Ray Tracing and Feature Set — RT/tensor cores, API support from facts

The Radeon PRO W6800 includes 60 ray-tracing cores. No tensor core count is listed in the specification data, so the card cannot be characterized on tensor-core hardware from this information. The architecture is RDNA 2.0, and the supported APIs are DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. These API entries indicate that the card is equipped for modern graphics workloads within the bounds of the listed feature set.

The shading complex contains 3840 shading units, 240 TMUs, and 96 ROPs. Floating-point throughput is 17.83 TFLOPS for FP32 and 35.67 TFLOPS for FP16 at a 2:1 ratio. The memory subsystem provides 32 GB of GDDR6 on a 256-bit bus with 512.0 GB/s of bandwidth. Display output is handled by 6x mini-DisplayPort 1.4a connectors. The card's boost clock reaches 2322 MHz, and its fillrates stand at 222.9 GPixel/s and 557.3 GTexel/s.

FAQ

Q: What GPU chip and architecture does the Radeon PRO W6800 use?

A: It uses the Navi 21 chip with the RDNA 2.0 architecture, fabricated by TSMC on a 7 nm process, with 26,800 million transistors and a 520 mm² die size.

Q: How much memory does the W6800 have?

A: It has 32 GB of GDDR6 memory on a 256-bit bus, with a 2000 MHz memory clock / 16 Gbps effective and 512.0 GB/s of bandwidth.

Q: What is the TDP and power-supply requirement?

A: The TDP is 250 W, and the suggested PSU is 600 W. Power is supplied through one 6-pin and one 8-pin PCIe power connector.

Q: What APIs are supported?

A: The card supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: How many ray-tracing cores does it have, and are tensor cores listed?

A: It has 60 ray-tracing cores. No tensor core count appears in the specification data.

Q: What display outputs are available?

A: The card provides 6x mini-DisplayPort 1.4a outputs.

Benchmark Performance

The Geekbench Metal score for the W6800 is 171137, the Geekbench OpenCL score is 120399, and the Geekbench Vulkan score is 109228. Among these three published runs, Metal produces the highest result, while Vulkan produces the lowest. The average benchmark score across the data is 133588, and the card is placed in the 97th percentile of all GPUs in the database.

Comparing to the nearest rivals, the exact deltaPct values are -0.6 versus the AMD Radeon RX 9070 GRE, -1.2 versus the NVIDIA RTX 4000 Ada Generation, -1.2 versus the NVIDIA A10M, and +1.3 versus the NVIDIA GeForce RTX 3090 Ti. The rival average scores are 134417 for the RX 9070 GRE, 135218 for the RTX 4000 Ada Generation, 135230 for the A10M, and 131911 for the RTX 3090 Ti. These figures show the W6800 in a tight cluster with its immediate competition: it is essentially tied with the RX 9070 GRE, slightly behind the RTX 4000 Ada Generation and A10M, and ahead of the RTX 3090 Ti.

The spread between the lowest and highest rival averages runs from 131911 to 135230. The W6800's average of 133588 lands inside that range. The Metal result of 171137 is notably higher than the OpenCL and Vulkan results, which means the card's aggregate position is pulled upward by its strongest API-specific run. The 97th percentile placement reinforces that the card sits near the top of the database distribution while still being narrowly edged out by two of its nearest rivals in average score.