NVIDIA GeForce GTX 970 vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

The Verdict

The database comparison between the NVIDIA GeForce GTX 970 and the NVIDIA RTX PRO 6000 Blackwell Server is not a contest in any traditional sense. The RTX PRO 6000 Blackwell Server wins the only shared benchmark, 3DMark Steel Nomad DX12, by a margin of 93.8 percent. The GTX 970 posts a score of 369 while the RTX PRO 6000 reaches 5996. That is a 16.2 times higher result, and no other benchmark appears in both records, so the verdict rests on this single measurement.

The GTX 970 is an end-of-life consumer product from the GeForce 900 generation, built on the Maxwell 2.0 architecture with the GM204 chip. Its average benchmark score across all recorded tests is 7157, placing it at the 39th percentile of all GPUs in the database. Its nearest rivals, all within roughly one percent of its average score, include the Intel Iris Pro Graphics P580, the NVIDIA GeForce GTX 560 SE, the AMD Radeon Vega 8 Mobile, and the NVIDIA GeForce GTX 750. This places the GTX 970 in the lower middle of the performance distribution, a position consistent with its age and specification level.

The RTX PRO 6000 Blackwell Server, in contrast, is an active product from the Server Blackwell generation, built on the Blackwell 2.0 architecture with the GB202 chip. Its average benchmark score, based solely on the 3DMark Steel Nomad DX12 result, is 5996, which puts it at the 34th percentile of all GPUs. Its nearest rivals, again within a fraction of a percent, are the NVIDIA GeForce GTX 770M, the AMD Radeon RX 6400, the AMD FirePro W4100, and the NVIDIA Quadro K4000M. The percentile ranking is surprising given the hardware specifications, but the database records only one benchmark for this server-class part, and that test heavily favors modern architectures.

Who should pick which is clear from the data. The GTX 970 suits anyone running legacy DirectX 9 through DirectX 12 workloads that are represented in its benchmark suite, particularly OpenCL and Vulkan compute tasks where it records strong scores. The RTX PRO 6000 Blackwell Server is the choice for modern DirectX 12 Ultimate workloads, specifically the kind of ray-traced, high-fidelity rendering that 3DMark Steel Nomad DX12 measures. For a buyer choosing between these two today, the RTX PRO 6000 Blackwell Server delivers the only modern performance evidence in the database, while the GTX 970 offers a broader set of legacy benchmark results but at a far lower absolute level.

Where Each One Wins

The GTX 970 wins in benchmark coverage. It has eleven recorded tests spanning DirectX 9, DirectX 10, DirectX 11, DirectX 12, Metal, OpenCL, Vulkan, 2D, 3D, and compute workloads. Its strongest results include a Passmark G3D score of 9638, a Geekbench OpenCL score of 29982, a Geekbench Vulkan score of 20005, and a Geekbench Metal score of 13595. It also records a Passmark GPU Compute score of 4073 and a Passmark G2D score of 764. The DirectX legacy scores are modest: Passmark DirectX 9 at 143, DirectX 11 at 71, DirectX 10 at 46, and DirectX 12 at 41. The 3DMark Steel Nomad DX12 score of 369 is the lowest among its recorded results, indicating that the card struggles with modern DirectX 12 rendering loads.

The RTX PRO 6000 Blackwell Server wins in absolute performance where it matters most. Its single recorded benchmark, 3DMark Steel Nomad DX12, yields a score of 5996, which is 5627 points higher than the GTX 970's result on the same test. This is the only direct head-to-head measurement in the database, and it is decisive. The RTX PRO 6000 also wins on architectural capability: it supports DirectX 12 Ultimate (12_2), while the GTX 970 supports DirectX 12 (12_1). The RTX PRO 6000 includes 188 ray tracing cores and 752 tensor cores, while the GTX 970 has none of either. The RTX PRO 6000 provides 96 GB of GDDR7 memory on a 512 bit bus with 1.79 TB/s of bandwidth, versus 4 GB of GDDR5 on a 256 bit bus with 224.4 GB/s of bandwidth.

The data suggests a use-case split. For legacy compute and older graphics APIs, the GTX 970 has a proven track record in the database, with high OpenCL and Vulkan scores relative to its peers. For modern DirectX 12 Ultimate workloads, particularly those that leverage ray tracing and tensor operations, the RTX PRO 6000 Blackwell Server is the only sensible choice based on the recorded evidence. The GTX 970 is a general-purpose older card; the RTX PRO 6000 is a server-class accelerator with a single, very strong modern benchmark result.

Architecture Differences

The two GPUs come from different eras of NVIDIA design. The GTX 970 uses the GM204 chip on the Maxwell 2.0 architecture, fabricated on a 28 nm process at TSMC. It contains 5,200 million transistors on a die size of 398 mm², yielding a transistor density of 13.1 million per square millimeter. The RTX PRO 6000 Blackwell Server uses the GB202 chip on the Blackwell 2.0 architecture, fabricated on a 5 nm process, also at TSMC. It contains 92,200 million transistors on a die size of 750 mm², yielding a transistor density of 122.9 million per square millimeter. The RTX PRO 6000 packs roughly 17.7 times more transistors into less than twice the die area, a density improvement of about 9.4 times.

The compute resources differ by an order of magnitude. The GTX 970 has 1664 shading units, 104 texture mapping units, and 56 render output units. The RTX PRO 6000 has 24064 shading units, 752 texture mapping units, and 192 render output units. That is 14.5 times more shading units, 7.2 times more TMUs, and 3.4 times more ROPs. The RTX PRO 6000 also adds dedicated hardware that the GTX 970 lacks entirely: 188 ray tracing cores and 752 tensor cores. The pixel rate jumps from 65.97 GPixel/s on the GTX 970 to 502.5 GPixel/s on the RTX PRO 6000, a 7.6 times increase. The texture rate jumps from 122.5 GTexel/s to 1,968.0 GTexel/s, a 16.1 times increase.

Memory architecture is radically different. The GTX 970 uses 4 GB of GDDR5 with a 256 bit bus and 224.4 GB/s of bandwidth. The RTX PRO 6000 uses 96 GB of GDDR7 with a 512 bit bus and 1.79 TB/s of bandwidth. That is 24 times more memory capacity and roughly 8 times more bandwidth. The memory clock on the GTX 970 is 1753 MHz with 7 Gbps effective, while the RTX PRO 6000 runs at 1750 MHz with 28 Gbps effective, four times the effective data rate.

Clock speeds also favor the newer card. The GTX 970 has a base clock of 1050 MHz and a boost clock of 1178 MHz. The RTX PRO 6000 has a base clock of 1590 MHz and a boost clock of 2617 MHz. The boost clock is 1439 MHz higher, a 122 percent increase. Floating point performance scales accordingly: the GTX 970 delivers 3.920 TFLOPS of FP32, while the RTX PRO 6000 delivers 126.0 TFLOPS of FP32, a 32.1 times difference. The RTX PRO 6000 also delivers 126.0 TFLOPS of FP16 with a 1:1 ratio, while the GTX 970 has no recorded FP16 capability.

The process node difference explains much of the gap. The 5 nm node allows for far higher transistor density, enabling the 92,200 million transistor count on a 750 mm² die. The 28 nm node of the GTX 970 caps its density at 13.1 million transistors per square millimeter. The result is a GPU that is architecturally more advanced in every measurable way, from shading throughput to memory bandwidth to dedicated ray tracing and tensor hardware.

FAQ

Q: Which GPU has a higher average benchmark score in the database?

A: The GTX 970 has an average benchmark score of 7157 across eleven recorded tests. The RTX PRO 6000 Blackwell Server has an average benchmark score of 5996 from a single recorded test. The GTX 970's average is 1161 points higher, but this reflects the broader test suite, not superior performance in modern workloads.

Q: What is the only benchmark both GPUs share?

A: The only shared benchmark is 3DMark Steel Nomad DX12. The GTX 970 scores 369, while the RTX PRO 6000 Blackwell Server scores 5996. The RTX PRO 6000 wins with a 93.8 percent margin, meaning its score is approximately 16.2 times higher.

Q: Does the RTX PRO 6000 support ray tracing?

A: Yes, the RTX PRO 6000 Blackwell Server includes 188 ray tracing cores and 752 tensor cores. The GTX 970 has no ray tracing cores and no tensor cores, and its DirectX support is limited to 12 (12_1) rather than the 12 Ultimate (12_2) of the RTX PRO 6000.

Q: How much memory does each GPU have?

A: The GTX 970 has 4 GB of GDDR5 memory on a 256 bit bus with 224.4 GB/s of bandwidth. The RTX PRO 6000 Blackwell Server has 96 GB of GDDR7 memory on a 512 bit bus with 1.79 TB/s of bandwidth. The RTX PRO 6000 has 24 times the capacity and roughly 8 times the bandwidth.

Q: What are the power requirements for each GPU?

A: The GTX 970 has a TDP of 148 W and uses two 6-pin power connectors with a suggested power supply of 300 W. The RTX PRO 6000 Blackwell Server has a TDP of 600 W and uses one 16-pin power connector with a suggested power supply of 1000 W. Both are dual-slot cards with identical physical dimensions.

Q: Which GPU is still in production?

A: The GTX 970 is marked as end-of-life in the database. The RTX PRO 6000 Blackwell Server is marked as active. The GTX 970 was released on September 18, 2014, while the RTX PRO 6000 was released on March 17, 2025.

Head-to-Head Benchmarks

The only head-to-head benchmark in the database is 3DMark Steel Nomad DX12. The GTX 970 records a score of 369, and the RTX PRO 6000 Blackwell Server records a score of 5996. The winner is the RTX PRO 6000 with a delta of 93.8 percent. This means the RTX PRO 6000's score is 5627 points higher, or 16.2 times the GTX 970's score.

To put the GTX 970's result in context, its 3DMark Steel Nomad DX12 score of 369 is far below its other benchmark results. Its Passmark G3D score is 9638, its Geekbench OpenCL score is 29982, and its Geekbench Vulkan score is 20005. The Steel Nomad test is a modern DirectX 12 workload that stresses GPU features the GTX 970 lacks, such as ray tracing and advanced shading. The GTX 970's DirectX 12 support is limited to the 12_1 feature level, while the RTX PRO 6000 supports 12 Ultimate (12_2).

The RTX PRO 6000's score of 5996 in Steel Nomad aligns with its massive compute resources: 24064 shading units, 752 TMUs, 192 ROPs, 188 ray tracing cores, and 752 tensor cores. Its FP32 throughput of 126.0 TFLOPS is 32.1 times the GTX 970's 3.920 TFLOPS. Its pixel rate of 502.5 GPixel/s is 7.6 times the GTX 970's 65.97 GPixel/s. Its texture rate of 1,968.0 GTexel/s is 16.1 times the GTX 970's 122.5 GTexel/s. The memory bandwidth of 1.79 TB/s is roughly 8 times the GTX 970's 224.4 GB/s.

The delta of 93.8 percent is the largest in any comparison in the database for these two parts. The GTX 970's nearest rivals, such as the Intel Iris Pro Graphics P580 and the NVIDIA GeForce GTX 560 SE, are within 0.2 percent of its average score. The RTX PRO 6000's nearest rivals, such as the NVIDIA GeForce GTX 770M and the AMD Radeon RX 6400, are within 0.1 percent of its average score. Neither card's nearest rivals approach the performance gap seen in the Steel Nomad test, which underscores how differently these two GPUs handle modern DirectX 12 rendering.

Specification Differences

The table below lists only the fields where the two GPUs differ, based on the database records.

| Specification | GeForce GTX 970 | RTX PRO 6000 Blackwell Server |

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

| Architecture | Maxwell 2.0 | Blackwell 2.0 |

| Chip | GM204 | GB202 |

| Generation | GeForce 900 | Server Blackwell (Bxx) |

| Process Node | 28 nm | 5 nm |

| Transistors | 5,200 million | 92,200 million |

| Die Size | 398 mm² | 750 mm² |

| Transistor Density | 13.1M / mm² | 122.9M / mm² |

| Base Clock | 1050 MHz | 1590 MHz |

| Boost Clock | 1178 MHz | 2617 MHz |

| Memory Clock | 1753 MHz, 7 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 4 GB | 96 GB |

| Memory Type | GDDR5 | GDDR7 |

| Memory Bus Width | 256 bit | 512 bit |

| Memory Bandwidth | 224.4 GB/s | 1.79 TB/s |

| Shading Units | 1664 | 24064 |

| TMUs | 104 | 752 |

| ROPs | 56 | 192 |

| RT Cores | None | 188 |

| Tensor Cores | None | 752 |

| Pixel Rate | 65.97 GPixel/s | 502.5 GPixel/s |

| Texture Rate | 122.5 GTexel/s | 1,968.0 GTexel/s |

| FP32 | 3.920 TFLOPS | 126.0 TFLOPS |

| FP16 | Not recorded | 126.0 TFLOPS (1:1) |

| TDP | 148 W | 600 W |

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

| Suggested PSU | 300 W | 1000 W |

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

| Display Outputs | 1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.2 | 4x DisplayPort 2.1b |

| DirectX Support | 12 (12_1) | 12 Ultimate (12_2) |

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

| Release Date | 2014-09-18 | 2025-03-17 |

| Predecessor | GeForce 700 | Server Hopper |

| Successor | GeForce 10 | Server Rubin |

| Launch MSRP | 329 USD | Not recorded |

The two GPUs share identical physical dimensions: 267 mm in length, 111 mm in height, and 40 mm in width, both dual-slot. Both use TSMC as the foundry, both support OpenGL 4.6 and Vulkan 1.4, and both have a 267 mm board length. The GTX 970 was released on September 18, 2014, with a launch MSRP of 329 USD. The RTX PRO 6000 Blackwell Server was released on March 17, 2025, with no recorded launch MSRP.