AMD Radeon Pro W6800X vs NVIDIA CMP 40HX Comparison

Where Each One Wins

The recorded data splits these two cards cleanly by workload and by intended environment. The AMD Radeon Pro W6800X wins the only direct head-to-head benchmark available in the database, the Geekbench OpenCL test, with a 33.3% margin over the NVIDIA CMP 40HX. That single win, combined with the overall benchmark averages, gives AMD the clear lead in general compute throughput, especially in workloads that scale with memory capacity, shading units, and raw FP32 output.

The NVIDIA CMP 40HX, by contrast, does not win any recorded benchmark in this comparison. Its role in the database is defined by its Geekbench Vulkan score of 77,879, a result that indicates its compute capability is skewed toward APIs that the AMD card was not directly tested on in this pairing. The CMP 40HX was designed as a mining-oriented GPU, and the data reflects that specialization: it has no display outputs, a narrow PCIe interface, and a smaller memory pool. None of those traits translate into a benchmark win against the W6800X, but they do define a use case where the card can still function: headless compute tasks where the host system does not rely on the GPU for rendering or display.

The AMD Radeon Pro W6800X also holds a commanding lead in the average benchmark score across all recorded tests. Its average is 160,671, while the CMP 40HX averages 85,637. That is a gap of roughly 87.6% in favor of the AMD card, and it places the W6800X at the 97th percentile of all GPUs in the database, versus the 93rd percentile for the CMP 40HX. The percentile difference, while modest in rank, masks a very large absolute performance gap. The W6800X is positioned among professional-grade accelerators like the NVIDIA A100 PCIe 40 GB and the AMD Radeon PRO W7800, while the CMP 40HX sits near workstation cards like the AMD Radeon PRO W6600.

For users choosing between these two, the data points in one direction: the AMD card wins in every measured category where both cards have results. The NVIDIA card is not without merit, but its merits are narrow. It consumes less power, requires a smaller power supply, and fits in a dual-slot chassis, which makes it easier to deploy in systems where physical space and power delivery are constrained. But in raw performance, the W6800X is the winner by every metric in the database.

Architecture Differences

The two GPUs come from different architectural generations and are built for different purposes. The AMD Radeon Pro W6800X uses the Navi 21 chip on the RDNA 2.0 architecture, fabricated on a 7 nm process at TSMC. The NVIDIA CMP 40HX uses the TU106 chip on the Turing architecture, fabricated on a 12 nm process, also at TSMC. The process node difference is substantial: 7 nm versus 12 nm, and it shows in transistor density. The AMD chip packs 26,800 million transistors into a 520 mm² die, yielding a density of 51.5 million transistors per square millimeter. The NVIDIA chip contains 10,800 million transistors on a 445 mm² die, for a density of 24.3 million per square millimeter. That means the AMD chip crams more than twice as many transistors into roughly the same physical area.

The compute resources differ accordingly. The W6800X has 3,840 shading units, 240 texture mapping units, and 96 ROPs. The CMP 40HX has 2,304 shading units, 144 TMUs, and 64 ROPs. The AMD card also carries 60 ray tracing cores, while the NVIDIA card has 36 RT cores and 288 tensor cores. The tensor core count is notable: the CMP 40HX includes tensor hardware that the W6800X lacks entirely, but the benchmark data does not include any tensor-specific tests, so its practical advantage is not measured here.

Memory configuration is another major divider. The W6800X ships with 32 GB of GDDR6 on a 256-bit bus, delivering 512.0 GB/s of bandwidth. The CMP 40HX has 8 GB of GDDR6 on a 256-bit bus, delivering 448.0 GB/s. The bus width is identical, but the AMD card has four times the capacity and roughly 14.3% more bandwidth. Clock speeds also favor AMD: the W6800X runs at a base of 1800 MHz and boosts to 2087 MHz, while the CMP 40HX runs at 1470 MHz base and 1650 MHz boost. Memory clocks differ as well, with the AMD card rated at 2000 MHz (16 Gbps effective) versus 1750 MHz (14 Gbps effective) on the NVIDIA card.

The physical and electrical profiles diverge sharply. The W6800X is a quad-slot card using Apple MPX power and interface connections, with a 200 W TDP and a suggested 550 W power supply. The CMP 40HX is a dual-slot card with a single 8-pin power connector, a 185 W TDP, and a suggested 450 W power supply. The NVIDIA card is shorter and narrower: 229 mm long and 111 mm tall versus 267 mm long and 120 mm tall for the AMD card. The NVIDIA card also has no display outputs, while the AMD card provides 1x HDMI 2.1 and 4x Thunderbolt outputs. The CMP 40HX uses a PCIe 1.0 x4 bus interface, which is a severe bottleneck for data transfer, whereas the AMD card uses the Apple MPX interface.

Both cards support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both are end-of-life products. The W6800X launched on 2021-08-02 with a launch MSRP of 2,799 USD. The CMP 40HX launched earlier, on 2021-02-24, with a launch MSRP of 699 USD. The AMD card was released roughly five months later, but the price gap is enormous and reflects the difference in memory capacity, compute resources, and target market.

Head-to-Head Benchmarks

The database contains exactly one direct head-to-head benchmark between these two cards, and it is decisive. In Geekbench OpenCL, the AMD Radeon Pro W6800X scores 124,498, while the NVIDIA CMP 40HX scores 93,395. That is a 33.3% advantage for the AMD card. This is not a marginal difference. In compute workloads that stress raw FP32 throughput and memory bandwidth, the W6800X is a full third faster than the CMP 40HX.

The FP32 numbers from the specification data corroborate this benchmark result. The W6800X is rated at 16.03 TFLOPS, more than double the 7.603 TFLOPS of the CMP 40HX. The texture rate tells a similar story: 500.9 GTexel/s for the AMD card versus 237.6 GTexel/s for the NVIDIA card. Pixel rate is also lopsided, at 200.4 GPixel/s versus 105.6 GPixel/s. Every throughput metric in the database points the same direction.

The CMP 40HX has its own benchmark entries, but they are not directly comparable to the W6800X because the AMD card was not tested on the same workloads in this dataset. The CMP 40HX scores 93,395 in Geekbench OpenCL and 77,879 in Geekbench Vulkan. The Vulkan score is lower than its OpenCL score by roughly 16.6%, which suggests the card is better optimized for OpenCL-style compute in this configuration. But even its stronger result, the OpenCL score, falls 33.3% short of the AMD card.

The nearest rival data for each card provides additional context. The W6800X's average score of 160,671 places it within 1.1% of the NVIDIA A100 PCIe 40 GB, within 2.6% of the AMD Radeon PRO W7800, within 2.8% of the NVIDIA RTX A5500, and within 3.3% of the NVIDIA RTX 4500 Ada Generation. That is a tight cluster of professional-class GPUs, and the W6800X sits right in the middle of it. The CMP 40HX, with an average score of 85,637, sits in a lower tier. It trails the AMD Radeon PRO W7600 by 1.7% and the NVIDIA Quadro GP100 by 2.1%, while leading the AMD Radeon PRO W6600 by 4.4% and the AMD Radeon Pro Vega 64X by 5.8%. The contrast between the two rival clusters is stark: the W6800X trades blows with A100-class hardware, while the CMP 40HX competes with mid-range workstation GPUs.

The wins tally also reads clearly. The AMD card records 1 win in the head-to-head section; the NVIDIA card records 0.

FAQ

Q: Which card is faster in compute workloads?

The AMD Radeon Pro W6800X is faster in the only shared benchmark, winning Geekbench OpenCL by 33.3%. Its score of 124,498 beats the CMP 40HX's 93,395 by a wide margin.

Q: Does the NVIDIA card have more memory bandwidth?

No. The W6800X leads with 512.0 GB/s of bandwidth, while the CMP 40HX has 448.0 GB/s. The AMD card also has 32 GB of memory versus 8 GB on the NVIDIA card.

Q: Can the NVIDIA card output video?

No. The CMP 40HX has no display outputs. The AMD W6800X provides 1x HDMI 2.1 and 4x Thunderbolt outputs.

Q: Why does the NVIDIA card have a lower average benchmark score?**

The CMP 40HX averages 85,637 across recorded benchmarks, which is far below the W6800X's 160,671 average. The gap reflects differences in shading units, clocks, and memory capacity.

Q: Are both cards still in production?

No. Both are listed as end-of-life products in the database.

Q: Which card uses more power?

The AMD card has a 200 W TDP and suggests a 550 W power supply. The NVIDIA card has a 185 W TDP and suggests a 450 W power supply.

Specification Differences

The two cards differ in nearly every specification category. The AMD Radeon Pro W6800X uses the Navi 21 chip on RDNA 2.0, while the NVIDIA CMP 40HX uses TU106 on Turing. The process nodes are 7 nm and 12 nm respectively. Transistor counts are 26,800 million versus 10,800 million, on die sizes of 520 mm² versus 445 mm², giving transistor densities of 51.5M per mm² versus 24.3M per mm².

Clock speeds: the W6800X runs at 1800 MHz base and 2087 MHz boost, while the CMP 40HX runs at 1470 MHz base and 1650 MHz boost. Memory clocks are 2000 MHz (16 Gbps effective) versus 1750 MHz (14 Gbps effective). Memory capacity is 32 GB versus 8 GB, both GDDR6, both on 256-bit buses, but with bandwidth of 512.0 GB/s versus 448.0 GB/s.

Compute resources: the W6800X has 3,840 shading units, 240 TMUs, 96 ROPs, and 60 RT cores. The CMP 40HX has 2,304 shading units, 144 TMUs, 64 ROPs, 36 RT cores, and 288 tensor cores. The AMD card has no tensor cores; the NVIDIA card has no tensor core benchmark results in this dataset.

Pixel rate is 200.4 GPixel/s versus 105.6 GPixel/s. Texture rate is 500.9 GTexel/s versus 237.6 GTexel/s. FP32 is 16.03 TFLOPS versus 7.603 TFLOPS. FP16 is 32.06 TFLOPS versus 15.21 TFLOPS, both at 2:1 ratio.

Power and physical specs: TDP is 200 W versus 185 W. The W6800X is a quad-slot card with Apple MPX power and interface; the CMP 40HX is dual-slot with a single 8-pin connector and PCIe 1.0 x4. Suggested PSU is 550 W versus 450 W. Length is 267 mm versus 229 mm, height is 120 mm versus 111 mm, and the NVIDIA card is 35 mm wide. Display outputs are 1x HDMI 2.1 plus 4x Thunderbolt versus no outputs. The AMD card launched 2021-08-02 with a launch MSRP of 2,799 USD; the NVIDIA card launched 2021-02-24 with a launch MSRP of 699 USD. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both are end-of-life.

The Verdict

The data selects a clear winner for anyone comparing compute performance: the AMD Radeon Pro W6800X. It wins the only head-to-head benchmark by 33.3%, holds a 160,671 versus 85,637 average score advantage, and sits at the 97th percentile of all GPUs compared to the 93rd percentile for the CMP 40HX. Its nearest rivals are the NVIDIA A100 PCIe 40 GB and AMD Radeon PRO W7800, putting it in a performance class that the CMP 40HX does not approach.

The NVIDIA CMP 40HX should be selected only in very specific circumstances. It draws 185 W versus 200 W, fits in a dual-slot chassis, uses a standard 8-pin connector, and requires a 450 W power supply rather than 550 W. For a headless compute node with limited space and power budget, those traits matter. But the PCIe 1.0 x4 interface is a severe constraint, and the card's 8 GB memory pool is a quarter the size of the W6800X's 32 GB. The data does not show any workload where the CMP 40HX beats the W6800X.

For professional graphics, rendering, or any task that benefits from large memory and high FP32 throughput, the Radeon Pro W6800X is the correct choice. For a low-power, compact, headless compute card where benchmark performance is secondary to physical and electrical constraints, the CMP 40HX has a role. In every measured metric, the AMD card leads.