AMD Radeon HD 8690M vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

FAQ

Q: How do the two GPUs compare in raw benchmark performance?

A: The AMD Radeon HD 8690M scores 6,137 in Geekbench OpenCL, while the NVIDIA RTX PRO 6000 Blackwell Server scores 5,996 in 3DMark Steel Nomad DX12. Because these are different test suites, the data cannot be directly compared; each GPU is benchmarked in its own context.

Q: What are the nearest rivals for each GPU?

A: The HD 8690M sits within 1.2% of the AMD Radeon HD 6950, and within 1% of the NVIDIA RTX A400 and GeForce MX230. The RTX PRO 6000 Blackwell Server is within 0.2% of the AMD FirePro W4100 and NVIDIA Quadro K4000M, and within 0.1% of the GTX 770M and RX 6400.

Q: Which GPU has the larger memory configuration?

A: The RTX PRO 6000 Blackwell Server has 96 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s bandwidth. The HD 8690M has 2 GB of GDDR5 on a 64-bit bus, with 32.00 GB/s bandwidth.

Q: What process nodes do the two GPUs use?

A: The HD 8690M uses a 28 nm process at TSMC with 690 million transistors on a 56 mm² die. The RTX PRO 6000 Blackwell Server uses a 5 nm process at TSMC with 92,200 million transistors on a 750 mm² die.

Q: What is the production status of each GPU?

A: The HD 8690M is end-of-life, released in 2013. The RTX PRO 6000 Blackwell Server is active, released in 2025.

Q: What API levels does each GPU support?

A: The HD 8690M supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. The RTX PRO 6000 Blackwell Server supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Architecture Differences

The architectural gap between these two GPUs spans over a decade of GPU design evolution. The AMD Radeon HD 8690M is built on the GCN 1.0 architecture, using the Sun chip, and belongs to the Solar System generation (HD 8600M series). It is fabricated on a 28 nm process at TSMC, integrating 690 million transistors into a 56 mm² die. The transistor density sits at 12.3M per mm². This is a legacy mobile GPU with a 64-bit memory interface and GDDR5 memory.

The NVIDIA RTX PRO 6000 Blackwell Server, in contrast, uses the Blackwell 2.0 architecture with the GB202 chip, belonging to the Server Blackwell (Bxx) generation. It is built on a 5 nm process at TSMC, packing 92,200 million transistors into a 750 mm² die, achieving a transistor density of 122.9M per mm². That is a 10x improvement in density per square millimeter.

The compute resources differ dramatically. The HD 8690M has 320 shading units, 20 texture mapping units, and 8 ROPs. The RTX PRO 6000 Blackwell Server has 24,064 shading units, 752 TMUs, and 192 ROPs. Additionally, the NVIDIA part includes 188 ray tracing cores and 752 tensor cores, while the AMD part has none of these dedicated accelerators.

Memory architecture is another fundamental divide. The HD 8690M uses 2 GB of GDDR5 with a 64-bit bus, yielding 32.00 GB/s of bandwidth. The RTX PRO 6000 Blackwell Server uses 96 GB of GDDR7 on a 512-bit bus, producing 1.79 TB/s of bandwidth — a 56x increase in bandwidth capacity.

The clock speeds also reflect different design philosophies. The HD 8690M runs at a base of 925 MHz and boost of 975 MHz with memory at 1000 MHz (4 Gbps effective). The RTX PRO 6000 Blackwell Server runs at 1590 MHz base and 2617 MHz boost, with memory at 1750 MHz (28 Gbps effective).

Power and physical specifications are equally divergent. The HD 8690M has no listed TDP, slot width, or power connector specifications, reflecting its mobile design. The RTX PRO 6000 Blackwell Server has a 600 W TDP, is dual-slot, uses a single 16-pin power connector, and requires a 1000 W suggested PSU. It measures 267 mm in length, 111 mm in height, and 40 mm in width.

The bus interface also differs: the HD 8690M uses PCIe 3.0 x8, while the RTX PRO 6000 Blackwell Server uses PCIe 5.0 x16. Display outputs are unspecified for the AMD part, while the NVIDIA part has 4x DisplayPort 2.1b.

Where Each One Wins

The AMD Radeon HD 8690M wins in the context of legacy mobile computing. Its 28 nm GCN architecture and 2 GB GDDR5 memory were designed for thin-and-light laptops in the 2013 era. The benchmark data shows it achieving a Geekbench OpenCL score of 6,137, which places it at the 36th percentile of all GPUs. Its nearest rival, the Intel Iris Pro Graphics 6200, is only 0.3% behind, indicating that this GPU is competitive with its integrated graphics contemporaries.

The HD 8690M is best suited for basic 3D acceleration, video playback, and light productivity workloads on older systems. Its 624.0 GFLOPS of FP32 performance and 19.50 GTexel/s texture rate are adequate for 1080p gaming at reduced settings from its era, but it lacks ray tracing and tensor cores entirely. The 32.00 GB/s memory bandwidth is a limiting factor for modern workloads.

The NVIDIA RTX PRO 6000 Blackwell Server wins in every conceivable modern compute category. Its 126.0 TFLOPS of FP32 and FP16 (1:1) performance, combined with 188 RT cores and 752 tensor cores, make it a server-class accelerator for AI inference, ray-traced rendering, and high-performance scientific computing. The 96 GB of GDDR7 memory with 1.79 TB/s bandwidth enables massive datasets to reside on the GPU without host memory transfers.

The RTX PRO 6000 Blackwell Server's 3DMark Steel Nomad DX12 score of 5,996 places it at the 34th percentile of all GPUs, with nearest rivals like the GTX 770M and RX 6400 within 0.1%. This suggests that in this specific test, the GPU's performance is comparable to mid-range consumer parts, though the test itself is likely not representative of the GPU's server-oriented capabilities.

The pixel rate difference is stark: the RTX PRO 6000 Blackwell Server achieves 502.5 GPixel/s versus 7.800 GPixel/s for the HD 8690M. The texture rate is 1,968.0 GTexel/s versus 19.50 GTexel/s. For rendering tasks, the NVIDIA part is in a completely different performance class.

Specification Differences

• Process node: 28 nm (HD 8690M) vs 5 nm (RTX PRO 6000 Blackwell Server)
• Transistors: 690 million vs 92,200 million
• Die size: 56 mm² vs 750 mm²
• Transistor density: 12.3M / mm² vs 122.9M / mm²
• Base clock: 925 MHz vs 1590 MHz
• Boost clock: 975 MHz vs 2617 MHz
• Memory clock: 1000 MHz (4 Gbps effective) vs 1750 MHz (28 Gbps effective)
• Memory size: 2 GB vs 96 GB
• Memory type: GDDR5 vs GDDR7
• Memory bus width: 64 bit vs 512 bit
• Memory bandwidth: 32.00 GB/s vs 1.79 TB/s
• Shading units: 320 vs 24,064
• TMUs: 20 vs 752
• ROPs: 8 vs 192
• RT cores: None vs 188
• Tensor cores: None vs 752
• Pixel rate: 7.800 GPixel/s vs 502.5 GPixel/s
• Texture rate: 19.50 GTexel/s vs 1,968.0 GTexel/s
• FP32 performance: 624.0 GFLOPS vs 126.0 TFLOPS
• FP16 performance: Not listed vs 126.0 TFLOPS (1:1)
• TDP: Not listed vs 600 W
• Slot width: Not listed vs Dual-slot
• Power connectors: Not listed vs 1x 16-pin
• Suggested PSU: Not listed vs 1000 W
• Bus interface: PCIe 3.0 x8 vs PCIe 5.0 x16
• Display outputs: Not listed vs 4x DisplayPort 2.1b
• DirectX support: 12 (11_1) vs 12 Ultimate (12_2)
• Vulkan support: 1.2.170 vs 1.4
• Dimensions: Not listed vs 267 mm x 111 mm x 40 mm
• Production status: End-of-life vs Active
• Release date: 2013-02-28 vs 2025-03-17

Head-to-Head Benchmarks

The fact pack contains no direct head-to-head benchmark entries, and the win counts are zero for both GPUs. However, the individual benchmark results provide context for how each GPU performs in its respective test environment.

The AMD Radeon HD 8690M delivers a Geekbench OpenCL score of 6,137. This places it at the 36th percentile of all GPUs. Its closest competitor is the AMD Radeon HD 6950, which scores 6,210, putting the HD 8690M 1.2% behind. The Intel Iris Pro Graphics 6200 scores 6,117 (0.3% behind), while the NVIDIA RTX A400 scores 6,078 (1% behind) and the GeForce MX230 scores 6,077 (1% behind). These narrow margins indicate that the HD 8690M is clustered tightly with its direct peers in OpenCL compute performance.

The NVIDIA RTX PRO 6000 Blackwell Server achieves a 3DMark Steel Nomad DX12 score of 5,996, placing it at the 34th percentile of all GPUs. Its nearest rivals are tightly packed: the GeForce GTX 770M scores 6,000 (0.1% ahead), the RX 6400 scores 6,001 (0.1% ahead), the FirePro W4100 scores 5,987 (0.2% behind), and the Quadro K4000M scores 5,986 (0.2% behind). This clustering suggests that in this particular DX12 test, the RTX PRO 6000 Blackwell Server performs on par with older mid-range GPUs, despite its massive architectural advantages.

The performance delta between the two GPUs in their respective benchmarks is not directly comparable due to different test suites. However, the architectural specifications reveal the scale of difference. The RTX PRO 6000 Blackwell Server has 75 times more shading units, 37.6 times more TMUs, and 24 times more ROPs than the HD 8690M. Its FP32 throughput of 126.0 TFLOPS is 202 times higher than the HD 8690M's 624.0 GFLOPS.

The memory bandwidth differential is even more pronounced: 1.79 TB/s versus 32.00 GB/s, a 56-fold difference. The RTX PRO 6000 Blackwell Server also opens capabilities that the HD 8690M cannot physically support, including ray tracing acceleration, tensor core operations, and FP16 compute at 1:1 ratio.

Given the release dates — 2013 for the HD 8690M and 2025 for the RTX PRO 6000 Blackwell Server — the performance gap reflects not just a generational leap but a fundamental shift in GPU purpose. The HD 8690M was a mobile graphics solution; the RTX PRO 6000 Blackwell Server is a server-class accelerator designed for professional workloads. The benchmark scores, while superficially similar (6,137 vs 5,996), measure entirely different capabilities in different software environments. The data shows that percentile rankings for both GPUs hover in the mid-30s, indicating that raw benchmark scores in this range are common across many GPU generations.