AMD Radeon HD 8790M vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

NVIDIA’s RTX PRO 6000 Blackwell Server and AMD’s Radeon HD 8790M occupy opposite ends of the hardware spectrum, separated by a decade of architectural evolution and a massive gulf in intended use. The data reveals a stark contrast: the RTX PRO 6000 delivers modern server-grade compute with a single benchmark score of 5996 in 3DMark Steel Nomad DX12, while the HD 8790M, an end-of-life mobile part from 2013, manages an average score of 5691 across OpenCL and Vulkan tests. The benchmark delta is modest on paper, but the underlying specifications and performance contexts could not be more different.

Head-to-Head Benchmarks

The head-to-head benchmark list is empty, so direct comparisons rely on the aggregate scores and nearest rival data. The RTX PRO 6000’s sole benchmark, 3DMark Steel Nomad DX12, yields a score of 5996. This places it in the 34th percentile of all GPUs, with nearest rivals including the NVIDIA GeForce GTX 770M (6000, -0.1%), AMD Radeon RX 6400 (6001, -0.1%), AMD FirePro W4100 (5987, +0.2%), and NVIDIA Quadro K4000M (5986, +0.2%). The data shows a tightly clustered set of scores, meaning the RTX PRO 6000 is statistically indistinguishable from these older or lower-tier parts in this specific test, despite its modern architecture and massive compute resources.

The HD 8790M, by contrast, has two benchmark entries: Geekbench OpenCL (5017) and Geekbench Vulkan (6365), resulting in an average of 5691. This lands it in the 33rd percentile, just one point below the RTX PRO 6000. Its nearest rivals include Intel Iris Pro Graphics P6300 (5712, -0.4%), NVIDIA GeForce GTX 670MX (5721, -0.5%), NVIDIA GeForce GTX 550 Ti (5731, -0.7%), and NVIDIA Quadro M500M (5604, +1.6%). Interestingly, the HD 8790M’s Vulkan score of 6365 is higher than the RTX PRO 6000’s Steel Nomad score, but these are different workloads; the Vulkan test likely favors the older card’s driver optimizations or the specific scene complexity, while Steel Nomad is a modern DX12 benchmark that stresses features the HD 8790M does not even support fully.

The delta between the two average scores is 305 points, with the RTX PRO 6000 leading by approximately 5.4%. However, this gap is misleading. The RTX PRO 6000’s score comes from a single DX12 test, while the HD 8790M’s average blends two different APIs. Benchmark results indicate that in raw percentile terms, the two are nearly equivalent—34th versus 33rd—suggesting that the HD 8790M’s legacy drivers and simpler architecture can still hold its own in certain synthetic workloads, despite being outclassed in every measurable specification.

Architecture Differences

The architectural divide is enormous. The RTX PRO 6000 uses the GB202 chip on a 5 nm TSMC process, packing 92,200 million transistors into a 750 mm² die, yielding a transistor density of 122.9M per mm². It runs on Blackwell 2.0 architecture, part of the Server Blackwell (Bxx) generation, with a base clock of 1590 MHz and a boost of 2617 MHz. The HD 8790M, meanwhile, uses the Mars chip on a 28 nm process, with just 950 million transistors on a 77 mm² die (12.3M per mm²), based on GCN 1.0 architecture from the Solar System (HD 8700M) generation, clocked at 850 MHz base and 900 MHz boost.

Memory systems diverge sharply. The RTX PRO 6000 offers 96 GB of GDDR7 on a 512-bit bus, delivering 1.79 TB/s bandwidth; the HD 8790M has 2 GB of GDDR5 on a 128-bit bus, with 64.00 GB/s. The RTX PRO 6000 features 24,064 shading units, 752 TMUs, 192 ROPs, 188 RT cores, and 752 tensor cores, while the HD 8790M has 384 shading units, 24 TMUs, and 8 ROPs, with no RT or tensor cores. Pixel and texture rates reflect this: 502.5 GPixel/s and 1,968.0 GTexel/s for the RTX PRO 6000 versus 7.200 GPixel/s and 21.60 GTexel/s for the HD 8790M. FP32 compute is 126.0 TFLOPS versus 691.2 GFLOPS—a 182-fold difference.

The RTX PRO 6000 is built for PCIe 5.0 x16, while the HD 8790M uses an MXM-A (3.0) interface, indicating its mobile laptop origins. The RTX PRO 6000 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the HD 8790M supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. Process node, transistor count, memory bandwidth, and feature set all point to the RTX PRO 6000 being a purpose-built server accelerator, whereas the HD 8790M is a legacy mobile GPU designed for modest 2013-era laptops.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The RTX PRO 6000 has an average score of 5996, while the HD 8790M averages 5691, giving the NVIDIA part a 305-point lead, which is roughly 5.4% higher.

Q: Does the HD 8790M outperform the RTX PRO 6000 in any benchmark?

A: Yes. The HD 8790M scores 6365 in Geekbench Vulkan, which is higher than the RTX PRO 6000’s 5996 in 3DMark Steel Nomad DX12. However, these are different tests, so this is not a direct head-to-head win.

Q: How do the two compare in terms of manufacturing process?

A: The RTX PRO 6000 uses a 5 nm TSMC process, whereas the HD 8790M uses a 28 nm TSMC process. This contributes to the RTX PRO 6000’s much higher transistor density of 122.9M per mm² versus 12.3M per mm².

Q: What are the memory differences?

A: The RTX PRO 6000 has 96 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The HD 8790M has 2 GB of GDDR5 on a 128-bit bus with 64.00 GB/s bandwidth.

Q: Which GPU has more shading units?

A: The RTX PRO 6000 has 24,064 shading units, compared to the HD 8790M’s 384. This is a 63-fold difference in raw shader count.

Q: What is the production status of each?

A: The RTX PRO 6000 is marked as Active in production, while the HD 8790M is End-of-life. The RTX PRO 6000 was released in 2025, and the HD 8790M in 2013.

The Verdict

The data points to an obvious conclusion: the RTX PRO 6000 Blackwell Server is the superior part for any modern, compute-intensive workload. Its FP32 throughput of 126.0 TFLOPS dwarfs the HD 8790M’s 691.2 GFLOPS, and its 96 GB of GDDR7 memory with 1.79 TB/s bandwidth is in a different class entirely. The HD 8790M, with its 2 GB GDDR5 and 64.00 GB/s bandwidth, cannot handle datasets or textures of any significant size. Benchmark percentiles are close (34th vs 33rd), but this is an artifact of the limited test set; the RTX PRO 6000’s single DX12 score does not reflect its full capability, and the HD 8790M’s Vulkan score is likely a legacy driver quirk.

For users needing server-grade acceleration, ray tracing, or tensor operations, the RTX PRO 6000 is the only choice, given its 188 RT cores and 752 tensor cores, which the HD 8790M lacks entirely. The HD 8790M’s 8 ROPs and 24 TMUs are insufficient for even moderate resolutions, while the RTX PRO 6000’s 192 ROPs and 752 TMUs provide 502.5 GPixel/s and 1,968.0 GTexel/s. The RTX PRO 6000 is also actively produced and supports modern APIs like DirectX 12 Ultimate and Vulkan 1.4, ensuring long-term software compatibility. The HD 8790M is end-of-life, with no future driver support, and its Vulkan 1.2.170 support is already outdated. The verdict is unambiguous: the RTX PRO 6000 is the only viable option for current or future workloads; the HD 8790M is a historical artifact.

Specification Differences

• Process Node: 5 nm (RTX PRO 6000) vs 28 nm (HD 8790M)
• Transistors: 92,200 million vs 950 million
• Die Size: 750 mm² vs 77 mm²
• Transistor Density: 122.9M / mm² vs 12.3M / mm²
• Base Clock: 1590 MHz vs 850 MHz
• Boost Clock: 2617 MHz vs 900 MHz
• Memory Size: 96 GB vs 2 GB
• Memory Type: GDDR7 vs GDDR5
• Memory Bus Width: 512 bit vs 128 bit
• Memory Bandwidth: 1.79 TB/s vs 64.00 GB/s
• Shading Units: 24,064 vs 384
• TMUs: 752 vs 24
• ROPs: 192 vs 8
• RT Cores: 188 vs None
• Tensor Cores: 752 vs None
• Pixel Rate: 502.5 GPixel/s vs 7.200 GPixel/s
• Texture Rate: 1,968.0 GTexel/s vs 21.60 GTexel/s
• FP32 Compute: 126.0 TFLOPS vs 691.2 GFLOPS
• TDP: 600 W vs Not specified
• Slot Width: Dual-slot vs MXM Module
• Power Connectors: 1x 16-pin vs None
• Bus Interface: PCIe 5.0 x16 vs MXM-A (3.0)
• Display Outputs: 4x DisplayPort 2.1b vs Portable Device Dependent
• DirectX Support: 12 Ultimate (12_2) vs 12 (11_1)
• Vulkan Support: 1.4 vs 1.2.170
• Production Status: Active vs End-of-life
• Release Date: 2025-03-17 vs 2013-03-31

Where Each One Wins

The RTX PRO 6000 wins decisively in every compute and rendering metric that matters for modern workloads. Its FP32 performance is 126.0 TFLOPS, making it suitable for AI training, scientific simulation, or high-end 3D rendering—tasks that require massive parallel throughput. The 96 GB memory capacity and 1.79 TB/s bandwidth enable handling of large models or datasets that would exhaust the HD 8790M’s 2 GB instantly. The presence of 188 RT cores and 752 tensor cores means the RTX PRO 6000 can accelerate ray-traced scenes and matrix operations, features entirely absent from the HD 8790M. With a 600 W TDP and dual-slot design, it is built for server racks, not laptops.

The HD 8790M wins only in the narrow context of legacy compatibility and portability. Its MXM-A (3.0) interface and lack of power connectors make it a drop-in upgrade for older mobile workstations. Its 28 nm process and 77 mm² die mean it draws minimal power, though no TDP is listed. In the Geekbench Vulkan test, it scored 6365, outperforming the RTX PRO 6000’s Steel Nomad score, but this is a synthetic anomaly rather than a practical advantage. The HD 8790M’s 4 Gbps effective memory clock and 64.00 GB/s bandwidth are fine for 2013-era games at low settings, but the RTX PRO 6000’s 28 Gbps effective memory clock and 1.79 TB/s bandwidth are 28 times faster. For any modern application, the RTX PRO 6000 is the clear winner; the HD 8790M’s only "win" is its ability to run on portable devices and its higher single-test Vulkan score, which does not translate to real-world superiority.