NVIDIA GeForce GTX 670 vs NVIDIA RTX PRO 6000D Blackwell Max-Q Comparison

FAQ

Q: What are the architectural generations of the GTX 670 and RTX PRO 6000D Blackwell Max-Q?

A: The GTX 670 is based on the Kepler architecture using the GK104 chip, part of the GeForce 600 generation. The RTX PRO 6000D Blackwell Max-Q uses the Blackwell 2.0 architecture with the GB202 chip, belonging to the Blackwell PRO W (x000) generation.

Q: How do the memory subsystems compare between the two cards?

A: The GTX 670 has 2 GB of GDDR5 memory on a 256-bit bus, delivering 192.3 GB/s of bandwidth. The RTX PRO 6000D Blackwell Max-Q features 96 GB of GDDR7 memory on a 512-bit bus, providing 1.79 TB/s of bandwidth, which is substantially larger in both capacity and throughput.

Q: What is the difference in transistor count and manufacturing process?

A: The GTX 670 contains 3,540 million transistors on a 28 nm process with a die size of 294 mm². The RTX PRO 6000D Blackwell Max-Q packs 92,200 million transistors on a 5 nm process with a die size of 750 mm². The transistor density jumps from 12.0M per mm² on the older card to 122.9M per mm² on the newer one.

Q: Which card has a higher FP32 compute performance?

A: The RTX PRO 6000D Blackwell Max-Q delivers 110.1 TFLOPS of FP32 performance. The GTX 670 offers 2.634 TFLOPS. The newer card is roughly 42 times faster in raw single-precision compute.

Q: Do these cards support the same modern graphics APIs?

A: No. The GTX 670 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The RTX PRO 6000D Blackwell Max-Q supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it more feature-complete for current and future titles.

Architecture Differences

The architectural gap between these two NVIDIA cards is vast, spanning over a decade of GPU evolution. The GTX 670 uses the Kepler architecture, built on a 28 nm TSMC process, with 3,540 million transistors packed into a 294 mm² die. The RTX PRO 6000D Blackwell Max-Q uses the Blackwell 2.0 architecture on a 5 nm TSMC process, with 92,200 million transistors across a 750 mm² die. This represents a 26-fold increase in transistor count and a 10-fold increase in transistor density per square millimeter.

The compute cores differ dramatically in scale. The GTX 670 has 1,344 shading units, 112 texture mapping units, and 32 ROPs. The RTX PRO 6000D Blackwell Max-Q has 24,064 shading units, 752 TMUs, and 192 ROPs. The newer card also introduces specialized hardware that the Kepler chip lacks entirely: 188 ray tracing cores and 752 tensor cores. These features enable hardware-accelerated ray tracing and AI workloads, which the GTX 670 cannot perform at the hardware level.

Clock speeds have also increased substantially. The GTX 670 runs at a 915 MHz base clock with a 980 MHz boost. The RTX PRO 6000D Blackwell Max-Q operates at 1,590 MHz base and 2,288 MHz boost, a significant increase in operating frequency. The memory architecture diverges completely: the older card uses 2 GB of GDDR5 with a 256-bit interface, while the newer card uses 96 GB of GDDR7 on a 512-bit bus. Memory bandwidth jumps from 192.3 GB/s to 1.79 TB/s, a nearly tenfold improvement.

The interface and connectivity also reflect generational progress. The GTX 670 uses PCIe 3.0 x16 and offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The RTX PRO 6000D Blackwell Max-Q uses PCIe 5.0 x16 and provides 4x DisplayPort 2.1b outputs. Power delivery changes from dual 6-pin connectors on the older card to a single 16-pin connector on the newer one. The GTX 670 has a TDP of 170 W with a suggested 450 W PSU, while the RTX PRO 6000D Blackwell Max-Q has a 300 W TDP and a suggested 700 W PSU.

The production status could not be more different. The GTX 670 is marked as end-of-life, released on May 9, 2012, with its successor being the GeForce 700 series. The RTX PRO 6000D Blackwell Max-Q is active, released on March 17, 2025, and has no successor listed. The physical dimensions are similar in height at 111 mm, but the newer card is longer at 267 mm versus 241 mm and slightly thicker at 40 mm versus 38 mm.

Head-to-Head Benchmarks

Direct head-to-head benchmark comparisons between these two cards are not available in the database, as the recorded data shows no shared test workloads. However, the individual benchmark scores and their relationships to other GPUs provide useful context for their relative performance tiers.

The GTX 670 has three recorded benchmark scores. In Geekbench Metal, it scores 7,424. In Geekbench OpenCL, it scores 15,341. In Geekbench Vulkan, it scores 15,553. Its average benchmark score across these tests is 12,773. The database places this card at the 52nd percentile among all GPUs. Its nearest rivals in the database include the NVIDIA GeForce GTX 590 with an average score of 12,830, which is 0.4% higher, and the AMD Radeon Pro 455 at 12,831, also 0.4% higher. The AMD Radeon RX 7600M XT scores 12,710, putting it 0.5% lower than the GTX 670, and the AMD FirePro W5100 scores 12,847, which is 0.6% higher. This clustering indicates the GTX 670 sits in a competitive middle tier for its era, with performance within a single percentage point of its nearest rivals.

The RTX PRO 6000D Blackwell Max-Q has only one recorded benchmark score: 11,088 in 3DMark Steel Nomad DX12. Its average benchmark score is 11,088, placing it at the 50th percentile among all GPUs. Its nearest rivals in the database are telling. The NVIDIA RTX PRO 6000 Blackwell Max-Q scores identically at 11,088 with a delta of 0%. The AMD Radeon RX 550 scores 11,075, just 0.1% lower. The NVIDIA GeForce GTX 1650 SUPER scores 11,047, 0.4% lower. The AMD FirePro W4300 scores 11,225, which is 1.2% higher than the RTX PRO 6000D Blackwell Max-Q.

These rival relationships reveal something unexpected. Despite having vastly superior specifications on paper, the RTX PRO 6000D Blackwell Max-Q records a lower average benchmark score than the GTX 670 in the database. The GTX 670's average of 12,773 exceeds the RTX PRO 6000D Blackwell Max-Q's average of 11,088. The newer card also sits slightly lower in the overall percentile ranking, at 50th versus the GTX 670's 52nd percentile.

The interpretation of these numbers requires caution. The benchmark suites differ between the two cards. The GTX 670's scores come from Geekbench workloads, which may favor older architectures in certain compute tasks. The RTX PRO 6000D Blackwell Max-Q's score comes from a single 3DMark Steel Nomad DX12 test, which is a modern DirectX 12 workload designed for current GPUs. The data does not support a direct performance equivalence claim, but it does show that the newer card's raw score in its tested workload is lower than the older card's average across its tested workloads.

The Verdict

The data presents a clear generational divide, but not in the direction one might expect. The GTX 670, despite being from 2012 and marked as end-of-life, records a higher average benchmark score of 12,773 compared to the RTX PRO 6000D Blackwell Max-Q's 11,088. The older card also holds a slightly higher percentile ranking at 52nd versus 50th. For users who prioritize raw benchmark scores in the recorded tests, the GTX 670 appears stronger.

However, the RTX PRO 6000D Blackwell Max-Q offers capabilities that the GTX 670 cannot match on any level. The newer card has 96 GB of GDDR7 memory versus 2 GB of GDDR5, a 512-bit bus versus 256-bit, and 1.79 TB/s of bandwidth versus 192.3 GB/s. It supports ray tracing and tensor cores, which the Kepler architecture lacks entirely. Its FP32 compute of 110.1 TFLOPS dwarfs the 2.634 TFLOPS of the GTX 670. It also supports modern APIs including DirectX 12 Ultimate and Vulkan 1.4, while the older card maxes out at DirectX 12 (11_0) and Vulkan 1.2.175.

The choice depends on the workload. The data suggests the GTX 670 may perform adequately in legacy Geekbench-style compute tasks, where its 52nd percentile ranking places it near contemporaries like the GTX 590 and Radeon Pro 455. The RTX PRO 6000D Blackwell Max-Q targets a different category entirely: professional workloads requiring massive memory capacity, hardware ray tracing, and AI acceleration. Its 3DMark Steel Nomad score of 11,088 places it near the RTX PRO 6000 Blackwell Max-Q and GTX 1650 SUPER, but its feature set extends far beyond what those scores indicate.

For modern gaming, content creation, and AI workloads, the RTX PRO 6000D Blackwell Max-Q is the only viable choice given its architectural support for current APIs and specialized cores. For users running legacy applications that rely on older compute paths, the GTX 670's recorded scores suggest it remains functional. The production status is decisive: the GTX 670 is end-of-life, while the RTX PRO 6000D Blackwell Max-Q is active and supported.

Specification Differences

| Specification | NVIDIA GeForce GTX 670 | NVIDIA RTX PRO 6000D Blackwell Max-Q |

|---|---|---|

| Architecture | Kepler | Blackwell 2.0 |

| Chip | GK104 | GB202 |

| Generation | GeForce 600 | Blackwell PRO W (x000) |

| Process Node | 28 nm | 5 nm |

| Transistors | 3,540 million | 92,200 million |

| Die Size | 294 mm² | 750 mm² |

| Transistor Density | 12.0M / mm² | 122.9M / mm² |

| Base Clock | 915 MHz | 1590 MHz |

| Boost Clock | 980 MHz | 2288 MHz |

| Memory Clock | 1502 MHz, 6 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 2 GB | 96 GB |

| Memory Type | GDDR5 | GDDR7 |

| Memory Bus Width | 256 bit | 512 bit |

| Memory Bandwidth | 192.3 GB/s | 1.79 TB/s |

| Shading Units | 1344 | 24064 |

| TMUs | 112 | 752 |

| ROPs | 32 | 192 |

| RT Cores | None | 188 |

| Tensor Cores | None | 752 |

| Pixel Rate | 27.44 GPixel/s | 439.3 GPixel/s |

| Texture Rate | 109.8 GTexel/s | 1,720.6 GTexel/s |

| FP32 | 2.634 TFLOPS | 110.1 TFLOPS |

| FP16 | Not listed | 110.1 TFLOPS (1:1) |

| TDP | 170 W | 300 W |

| Power Connectors | 2x 6-pin | 1x 16-pin |

| Suggested PSU | 450 W | 700 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 5.0 x16 |

| Display Outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | 4x DisplayPort 2.1b |

| DirectX | 12 (11_0) | 12 Ultimate (12_2) |

| Vulkan | 1.2.175 | 1.4 |

| Length | 241 mm | 267 mm |

| Width | 38 mm | 40 mm |

| Production Status | End-of-life | Active |

| Release Date | 2012-05-09 | 2025-03-17 |

| Launch MSRP | 399 USD | 8,565 USD |